US12001939B2

US12001939B2 - Artificial intelligence (AI)-based guidance for an ultrasound device to improve capture of echo image views

Info

Publication number: US12001939B2
Application number: US17/219,670
Authority: US
Inventors: II James Otis Hare; Su Ping Carolyn LAM; Yoran Hummel; Mathias IVERSEN; Andrie OCHTMAN
Original assignee: EkoAi Pte Ltd
Current assignee: EkoAi Pte Ltd
Priority date: 2018-12-11
Filing date: 2021-03-31
Publication date: 2024-06-04
Also published as: US20210264238A1

Abstract

Artificial intelligence (AI)-based guidance for an ultrasound device to improve capture of echo image views comprises receiving a plurality of the echo images captured by a probe of the ultrasound device, the ultrasound device including a display and a user interface (UI) that displays the echo images to a user. One or more neural networks process the echo images to continuously attempt to automatically classify the echo images by view type and generates corresponding classification confidence scores. The view type of the echo images are simultaneously displayed in the UI along with the echo images. Feedback indications are displayed in the UI of the ultrasound device to the user, where the feedback indications include which directions to move the probe of the ultrasound device so the probe can be placed in a correct position to capture and successfully classify the echo images.

Description

CROSS-REFERENCE TO RELATED APPLICATION

This application is a continuation in-part of patent application Ser. No. 16/833,001, filed Mar. 27, 2020, which issued on Apr. 12, 2022 as U.S. Pat. No. 11,301,996, which is a continuation in-part of patent application Ser. No. 16/216,929, filed Dec. 11, 2018, which issued on Apr. 28, 2020 as U.S. Pat. No. 10,631,828, both assigned to the assignee of the present application and incorporated herein by reference.

BACKGROUND

Cardiovascular disease including heart failure is a major health problem accounting for about 30% of human deaths worldwide. Heart failure is also the leading cause of hospitalization in adults over the age of 65 years. Echocardiography is an important diagnostic aid in cardiology for the morphological and functional assessment of the heart. In a typical patient echocardiogram (echo) examination, a clinician called a sonographer places an ultrasound device against the patient's chest to capture a number of 2D images/videos of the patients' heart. Reflected sound waves reveal the inner structure of the heart walls and the velocities of blood flows. The ultrasound device position is varied during an echo exam to capture different anatomical sections as 2D slices of the heart from different viewpoints or views. The clinician has the option of adding to these 2D images a waveform captured from various possible modalities including continuous wave Doppler, m-mode, pulsed wave Doppler and pulsed wave tissue Doppler. The 2D images/videos and Doppler modality images are typically saved in DICOM (Digital Imaging and Communications in Medicine) formatted files. Although the type of modality is partially indicated in the metadata of the DICOM file, the ultrasound device position in both the modality and 2D views, which is the final determinant of which cardiac structure has been imaged, is not.

After the patient examination, a clinician/technician goes through the DICOM files, manually annotates heart chambers and structures like the left ventricle (LV) and takes measurements of those structures. The process is reliant on the clinicians' training to recognize the view in each image and make the appropriate measurements. In a follow up examination, a cardiologist reviews the DICOM images and measurements, compares them to memorized guideline values and make a diagnosis based on the interpretation made from the echocardiogram.

The current workflow process for analyzing DICOM images, measuring cardiac structures in the images and determining, predicting and prognosticating heart disease is highly manual, time-consuming and error-prone. Because the workflow process is so labor intensive, more than 95% of the images available in a typical patient echocardiographic study are never annotated or quantified. The view angle or Doppler modality type by which an image was captured is typically not labelled, which means the overwhelming majority of stored DICOMs from past patient studies and clinical trials do not possess the basic structure and necessary identification of labels to allow for machine learning on this data.

BRIEF SUMMARY

The disclosed embodiments provide methods and systems for artificial intelligence (AI)-based guidance for an ultrasound device to improve capture of echo image views. Aspects of exemplary embodiment include at least one processor receiving a plurality of the echo images captured by the ultrasound device, the ultrasound device including a display and a user interface (UI) that displays the echo images to a user, the echo images comprising 2D images and Doppler modality images of a heart. One or more neural networks process the echo images to continuously attempt to automatically classify the echo images by view type and generates corresponding classification confidence scores. The view type of the echo images are simultaneously displayed in the UI along with the echo images. Feedback indications are displayed in the UI of the ultrasound device to the user, where the feedback indications include which directions to move a probe of the ultrasound device so the probe can be placed in a correct position to capture and successfully classify the echo images.

According to the method and system disclosed herein, the disclosed embodiments provide an automated clinical workflow that diagnoses heart disease, while enhancing functionality of the mobile ultrasound device to improve the capture of echo image views.

BRIEF DESCRIPTION OF SEVERAL VIEWS OF THE DRAWINGS

FIGS. 1A-1C are diagrams illustrating embodiments of a system for implementing an automated clinical workflow diagnoses heart disease based on both cardiac biomarker measurements and AI recognition of 2D and Doppler modality Echocardiographic images for automated measurements and the diagnosis, prediction and prognosis of heart disease. FIG. 1D illustrates a block diagram of an example point-of-care (POC) device for measuring cardiac biomarkers.

FIG. 2 illustrates architectural layers of the echo workflow engine.

FIG. 3 is a flow diagram illustrating one embodiment of a process for performed by the echo workflow engine to diagnose heart disease based on both cardiac biomarker measurements and AI recognition of both 2D and Doppler modality echo images to perform automated measurements and the diagnosis, prediction and prognosis of heart disease.

FIG. 4A is a flow diagram illustrating details of the process for automatically recognizing and analyze both 2D and Doppler modality echo images to perform automated measurements and the diagnosis, prediction and prognosis of heart disease according to one embodiment.

FIG. 4B is a flow diagram illustrating advanced functions of the echo workflow engine.

FIG. 5A is diagram illustrating an example 2D echo image.

FIG. 5B is diagram illustrating an example Doppler modality image.

FIGS. 6A-6K are diagrams illustrating some example view types automatically classified by the echo workflow engine.

FIG. 7 is a diagram illustrating an example 2D image segmented to produce annotations indicating cardiac chambers, and an example Doppler modality image segmented to produce a mask and trace waveform

FIGS. 8A and 8B are diagrams illustrating examples of finding systolic/diastolic end points.

FIGS. 9A and 9B are diagrams illustrating processing of an imaging window in a 2D echo image and the automated detection of out of sector annotations.

FIGS. 10A and 10B are diagrams graphically illustrating structural measurements automatically generated from annotations of cardiac chambers in a 2D image, and velocity measurements automatically generated from annotations of waveforms in a Doppler modality.

FIG. 11 is a diagram graphically illustrating measurements of global longitudinal strain that were automatically generated from the annotations of cardiac chambers in 2D images.

FIG. 12A is a diagram graphically illustrating an example set of best measurement data based on largest volume cardiac chambers and the saving of the best measurement data to a repository.

FIG. 12B is a diagram graphically illustrating the input of automatically derived measurements from a patient with normal LV EF measurements into a set of rules to determine a conclusion that the patient has normal diastolic function, diastolic dysfunction, or indeterminate.

FIG. 13 is a diagram graphically illustrating the output of classification, annotation and measurement data to an example JSON file.

FIG. 14 is a diagram illustrating a portion of an example report showing highlighting values that are outside the range of International guidelines.

FIG. 15 is a diagram illustrating a portion of an example report of Main Findings that may be printed and/or displayed by the user.

FIG. 16A is a diagram showing a graph A plotting PCWP and HFpEF scores.

FIG. 16B is a diagram showing a graph B plotting CV mortality or hospitalization for HF.

FIG. 17 is a diagram illustrating a federated training platform associated with the automated clinical workflow system.

FIG. 18 is a diagram illustrating an automated clinical workflow system embodiment where the echo workflow engine is configured to provide a mobile ultrasound device with automatic recognition and measurements of 2D and Doppler modality Echocardiographic images.

FIG. 19 is a flow diagram for enhancing an ultrasound device to perform AI recognition of echocardiogram images.

FIG. 20A is a diagram illustrating acquisition of echo images by the mobile ultrasound device.

FIG. 20B is a diagram of a user interface displayed by the mobile ultrasound device.

FIGS. 20C, 20D and 20E are diagrams illustrating example pages of a report showing calculated measurements of features in the echo images.

FIG. 21 is a flow diagram showing processing by the echo workflow engine in a connected configuration comprising the echo workflow client in network communication with the echo workflow engine executing on one or more servers.

FIG. 22A is flow diagram illustrating an AI-based guidance process for an ultrasound device to improve capture of echo image views.

FIG. 22B is a diagram of a user interface displayed by the mobile ultrasound device.

FIGS. 22C-22E are diagrams showing a progression of feedback indications continuing from FIG. 22B.

DETAILED DESCRIPTION

The disclosed embodiments relate to artificial intelligence (AI)-based guidance for an ultrasound device to improve capture of echo image views. The following description is presented to enable one of ordinary skill in the art to make and use the invention and is provided in the context of a patent application and its requirements. Various modifications to the exemplary embodiments and the generic principles and features described herein will be readily apparent. The exemplary embodiments are mainly described in terms of particular methods and systems provided in particular implementations. However, the methods and systems will operate effectively in other implementations. Phrases such as “exemplary embodiment”, “one embodiment” and “another embodiment” may refer to the same or different embodiments. The embodiments will be described with respect to systems and/or devices having certain components. However, the systems and/or devices may include more or less components than those shown, and variations in the arrangement and type of the components may be made without departing from the scope of the invention. The exemplary embodiments will also be described in the context of particular methods having certain steps. However, the method and system operate effectively for other methods having different and/or additional steps and steps in different orders that are not inconsistent with the exemplary embodiments. Thus, the present invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features described herein.

The disclosed embodiments provide method and system for implementing a software-based automatic clinical workflow that diagnoses heart disease based on both cardiac biomarker measurements and AI recognition of 2D and Doppler modality echocardiographic images. In embodiments, the clinical workflow performs diagnosis, prediction and prognosis of heart disease, and can be deployed in workstation or mobile-based point-of-care systems.

FIGS. 1A-1C are diagrams illustrating embodiments of a system for implementing an automated clinical workflow diagnoses heart disease based on both cardiac biomarker measurements and AI recognition of 2D and Doppler modality Echocardiographic images. FIG. 1A shows a basic standalone configuration for the automated clinical workflow system 10A and a connected configuration 10B. The automated clinical workflow 10A is primarily implemented as a software application, referred to as echo workflow engine 12, that executes on a computer 14 operating in a standalone setting, disconnected from other devices on network 26. The computer 14 may be implemented in any form factor including a workstation, desktop, notebook, laptop server or tablet capable of running an operating system, such as Microsoft Windows® (e.g., Windows 7®, Windows 10®), Apple macOS®, Linux®, Apple iOS®, Android®, and the like.

The computer 14 may include typical hardware components (not shown) including a processor, input devices (e.g., keyboard, pointing device, microphone for voice commands, buttons, touchscreen, etc.), output devices (e.g., a display device, speakers, and the like), and wired or wireless network communication interfaces (not shown) for communication. The computer 14 may include internal computer-readable media, such as memory (not shown) containing computer instructions comprising the echo workflow engine 12, which implements the functionality disclosed herein when executed by one or more computer processors.

The computer 14 may further include local internal storage for storing one or more databases 16 and an image file archive 18. In one embodiment, the contents of the image file archive 18 include echocardiogram image files (also referred to herein as echo images), which in some embodiments may be stored in DICOM (Digital Imaging and Communications in Medicine) format.

In one embodiment, the computer 14 is in communication with peripheral devices such a point-of-care (POC) device 25, an ultrasound imaging device 24, or both. The POC device 25 is capable of measuring cardiac biomarkers in POC environments such as an emergency room, intensive care unit, physician's office, an ambulance, a patient setting, and remote emergency sites, as explained below with reference with FIG. 1D.

The ultrasound imaging device 24 captures echocardiogram images of a patient's organ (e.g., a heart), which may then be stored as a patient study using the database 16 and image file archive 18. For example, the computer 14 may be located in a hospital or clinical lab environment where Echocardiography is performed as a diagnostic aid in cardiology for the morphological and functional assessment of the heart. During a typical patient echocardiogram exam (referred to as a study), a sonographer or technician places the ultrasound imaging device 24 against the patient's chest to capture 2D echo images/videos of the heart to help diagnose the particular heart ailment. Measurements of the structure and blood flows are typically made using 2D slices of the heart and the position of the ultrasound imaging device 24 is varied during an echo exam to capture different anatomical sections of the heart from different viewpoints. The technician has the option of adding to these 2D echo images a waveform captured from various possible modalities including: continuous wave Doppler, m-mode, pulsed wave Doppler and pulsed wave tissue Doppler. The 2D images and Doppler waveform images may be saved as DICOM files. Although the type of modality is sometimes indicated in the metadata of the DICOM file, the 2D view is not.

The computer 14 may further include removable storage devices such as an optical disk 20 and/or a flash memory 22 and the like for storage of the echo images. In some embodiments, the removable storage devices may be used as an import source of echo images and related data structures into the internal image file archive 18, rather than or in addition to, the ultrasound imaging device 24. The removable storage devices may also be used as an archive for echocardiogram data stored in the database 16 and/or the image file archive 18.

FIG. 1A also shows an advanced optional embodiment, referred to as connected configuration 10B, where the computer 14 may be connected through the network 26 and a router 28 to other DICOM based devices, such as DICOM servers 30, network file share devices 32, echo workstations 34, and/or cloud storage services 36 hosting DICOM files. It should be understood that the servers 30 are also computers comprising a processor, input and output devices, wired or wireless network communication interfaces for communication, and computer-readable media, such as memory containing computer instructions comprising components of the echo workflow engine 12. In the connected configuration 10B, several possible interactions with the database 16 and the image file archive 18 are possible, as described below.

One possible interaction is to use the cloud storage services 36 as an internal archive. In case of very large archives consisting of large amounts of DICOM files, the computer 14 may not have sufficient storage to host all files and the echo workflow engine 12 may be configured to use external network storage of the cloud storage services 36 for file storage.

Another possible interaction is to use the cloud storage services 36 as an import source by i) selecting a DICOM file set or patient study, which includes the DICOM and Doppler waveforms images and patient data and examination information, including cardiac biomarker measurements. The patient study may also be selected by a reserved DICOMDIR file instance, from which the patient, exams and image files are read.

Yet a further possible interaction is to use the DICOM servers 30, the network file share devices 32, echo workstations 34, and/or DICOM clients (of FIG. 1C) acting as DICOM servers (workstations with modalities CFind, CMove and CStore) in order to retrieve patients, exams and images by performing a CFind operation, followed by a CMove operation, to request the remote device to send the cardiac biomarker measurements and/or images resulting from the CFind operation.

Referring now to FIG. 1B, a handheld configuration 10C for the automated clinical workflow system is shown. In this embodiment, the computer 14 of FIG. 1A is implemented as a handheld device 14′, such as a tablet or a mobile phone, connected to a wired or wireless portable ultrasound scanner probe 24′ that transmits echo images to the handheld device 14′. In one such embodiment, the echo workflow engine 12 may be implemented as an application executed by the handheld device 14′. In another embodiment, the handheld device 14′ may transmit the echo images to another computer/server executing the echo workflow engine 12 for processing.

FIG. 1C illustrates a software as a service (SaaS) configuration 10D for the automated clinical workflow. In this embodiment, the echo workflow engine 12 is run on a server 40 that is in communication over the network 26 with a plurality of client devices 42. In this embodiment, the server 40 and the echo workflow engine 12 may be part of a third-party service that provides automated measurements and the diagnosis, prediction and prognosis of heart disease to client devices (e.g., hospital, clinic, or doctor computers) over the Internet. It should be understood that although the server 40 is shown as a single computer, it should be understood that the functions of server 40 may be distributed over more than one server. In an alternative embodiment, the server 40 and the echo workflow engine 12 of FIG. 1C may be implemented as a virtual entity whose functions are distributed over multiple client devices 42. Likewise, it should be understood that although the echo workflow engine 12 is shown as a single component in each embodiment, the functionality of the echo workflow engine 12 may be separated into a greater number of modules/components.

Conventionally, after a patient examination where echo images are captured stored, a clinician/technician goes through the DICOM files, manually annotates heart chambers and structures and takes measurements, which are presented in a report. In a follow up examination, a doctor will review the DICOM images and measurements, compare them to memorized guideline values and make a diagnosis. Such a process is reliant on the clinicians' training to recognize the view and make the appropriate measurements so that a proper diagnosis can be made. Such a process is error-prone and time consuming.

According to the disclosed embodiments, the echo workflow engine 12 mimics the standard clinical practice of diagnosing heart disease of a patient by combining cardiac biomarker measurements and processing DICOM files of the patient using a combination of machine learning, image processing, and DICOM workflow techniques to derive clinical measurements, diagnose specific diseases, and prognosticate patient outcomes, as described below. While an automated solution to echo image interpretation using machine learning has been previously proposed, the solution fails to take cardiac biomarker measurements into account and only analyzes 2D echo images and not Doppler modality waveform images. The solution also mentions disease prediction, but only attempts to handle two diseases (hypertrophic cardiomyopathy and cardiac amyloidosis) and the control only compares normal patients to diseased patients.

The echo workflow engine 12 of the disclosed embodiments, however, improves on the automated solution by optionally combining cardiac biomarker measurements with machine learning that automatically recognizes and analyzes not only 2D echo images but also Doppler modality waveform images in order to diagnose heart disease. The echo workflow engine 12 is also capable of comparing patients having similar-looking heart diseases (rather than comparing normal patients to diseased patients), and automatically identifies additional diseases, including both heart failure with reduced ejection fraction (HFrEF) and heart failure with preserved ejection fraction (HFpEF). HFrEF is known as heart failure due to left ventricular systolic dysfunction or systolic heart failure and occurs when the ejection fraction is less than 40%. HFpEF is a form of congestive heart failure where in the amount of blood pumped from the heart's left ventricle with each beat (ejection fraction) is greater than 50%. Finally, unlike the proposed automated solution, the echo workflow engine 12 automatically takes into account cardiac biomarker measurements.

Cardiac biomarkers are substances that are released into the blood when the heart is damaged or stressed. Measurements of these biomarkers are used to help diagnose acute coronary syndrome (ACS) and cardiac ischemia, conditions associated with insufficient blood flow to the heart. Tests for cardiac biomarkers can also be used to help determine a person's risk of having these conditions. Increases in one or more cardiac biomarkers in the blood can identify people with ACS or cardiac ischemia, allowing rapid and accurate diagnosis and appropriate treatment of their condition.

Example types of cardiac biomarkers include B-type natriuretic peptide (BNP) and N-terminal pro-brain natriuretic peptide (NT-proBNP), High-sensitivity C-reactive Protein (hs-CRP), Cardiac Troponin, Creatine Kinase (CK), Creatine kinase-MB (CK-MB), and Myoglobin. Cardiac biomarker tests are typically available to a health practitioner 24 hours a day, 7 days a week with a rapid turn-around-time. Some of the tests are performed at the point of care (POC), e.g., in the emergency department or at a patient's bedside.

A key reason for under-diagnosis of HF is the non-specificity of presenting symptoms and signs, necessitating objective diagnostic tests. The measurement of plasma natriuretic peptide (NP) concentration is recommended by international guidelines for the initial diagnosis of HF, particularly in situations where echocardiography is not readily available such as non-hospital settings and primary care. For instance, an N-terminal pro-brain natriuretic peptide (NT-proBNP) concentration below 125 pg/mL has high negative predictive value and is recommended for ruling-out HF in non-acute settings. However, several cardiovascular and non-cardiovascular causes of elevated NPs weaken their positive predictive value in HF. This is especially the case in HFpEF, where atrial fibrillation, advanced age, renal failure and obesity are common comorbidities and importantly impede the interpretation of NP measurements. In such cases, the demonstration of objective cardiac dysfunction by echocardiography is mandated for the diagnosis of HF.

Echocardiography is needed to distinguish among the types of HF (HFpEF or HF with reduced ejection fraction [HFrEF])—a distinction that cannot be made by raised NP levels alone and is critical for the selection of appropriate therapies. Traditional echocardiography is highly manual, time consuming, error-prone, limited to specialists, and involves long waiting times (e.g. up to 9 months in some areas of NHS Scotland). However, the Artificial Intelligence (AI) approached described herein allows fully automated, fast and reproducible echocardiographic image analysis; turning a manual process of 30 minutes, 250 clicks, with 21% variability, into an AI-automated process taking 2 minutes, 1 click, with 0% variability. Such AI-enabled echocardiographic interpretation therefore not only increases efficiency and accuracy, but also opens the door to decision support for non-specialists.

According the disclosed embodiments, a combination of circulating cardiac and echocardiographic biomarkers represents an ideal diagnostic panel for HF. Such combined interpretation of multimodal data was not possible in the past since blood-based and imaging-based labs largely functioned independent of each other. In the current era of linked electronic health records and picture archiving and communication system (PACS) in many hospitals, the development of true “companion diagnostics” with combined interpretation of both blood and imaging biomarkers is possible. Moreover, advancements in medical AI enable deep learning models to be developed for greater diagnostic/predictive precision than ever achieved before. Automation of these algorithms, built into decision support tools for clinical application, has the potential to transform the diagnosis of HF.

FIG. 1D illustrates a block diagram of an example point-of-care (POC) device for measuring cardiac biomarkers. As stated above, the POC device 42A is capable of measuring cardiac biomarkers in POC environments such as an emergency room, intensive care unit, physician's office, an ambulance, a patient setting, and remote emergency sites. In embodiments, a patient blood sample may be delivered to the POC device 42A either through a strip reader 50 that receives an insert strip (not shown) containing the sample, or through a sample reader 52 that receives the sample from a syringe or a patient's finger. The POC device 42A analyzes the blood sample for one or more cardiac biomarkers and displays the results on a display 54 within minutes. The POC device 42A can store the results as well as wirelessly transmit the results to other systems/devices in the POC system, such as the echo workflow engine 12, and/or the results can be saved in the patient record or study.

In one embodiment, the POC device 42A measures at least B-type natriuretic peptide (BNP) and/or N-terminal pro-brain natriuretic peptide (NT-proBNP), as shown. In one embodiment, the POC device 42A may measure only NT-proBNP. In another embodiment, the POC device 42A may also measure other cardiac biomarkers including High-sensitivity C-reactive Protein (hs-CRP), Cardiac Troponin, Creatine Kinase (CK), Creatine kinase-MB (CK-MB), and Myoglobin. In one specific embodiment, an example of the POC device 42A is the commercially available COBAS 232 POC System™ by ROCHE.

The availability of point-of-care (POC) testing for both NT-proBNP and echocardiography (e.g., using mobile echo probes connected to handheld smart devices as in FIG. 1B) enables the use of AI-enabled tools within the primary care or community setting. Indeed, the current COVID-19 pandemic has highlighted the urgent need for such point-of-care community-based testing in Recovery Plans to respond to COVID-19.

FIG. 2 illustrates architectural layers of the echo workflow engine 12. In one embodiment, the echo workflow engine 12 may include a number of software components such as software servers or services that are packaged together in one software application. For example, the echo workflow engine 12 may include a machine learning layer 200, a presentation layer 202, and a database layer 204.

The machine learning layer 200 comprises several neural networks to process incoming echo images and corresponding metadata. The neural networks used in the machine learning layer may comprise a mixture of different classes or model types. In one embodiment, machine learning layer 200 utilizes a first neural network to classify 2D images by view type, and uses a second set of neural networks 200B to both extract features from Doppler modality images and to use the extracted features to classify the Doppler modality images by region (the neural networks used to extract features may be different than the neural network used to classify the images). The first neural network 200A and the second set of neural networks 200B may be implemented using convolutional neural network (CNN) and may be referred to as classification neural networks or CNNs.

Additionally, a third set of neural networks 2000, including adversarial networks, are employed for each classified 2D view type in order to segment the cardiac chambers in the 2D images and produce segmented 2D images. A fourth set of neural networks 200D are used for each classified Doppler modality region in order to segment the Doppler modality images to generate waveform traces. In additional embodiments, the machine learning layer 200 may further include a set of one or more prediction CNNs for disease prediction and optionally a set of one or more prognosis CNNs for disease prognosis (not shown). The third and fourth sets of

neural networks

2000 and 200D may be implemented using CNNs and may be referred to as segmentation neural networks or CNNs.

In machine learning, a CNN is a class of deep, feed-forward artificial neural network typically use for analyzing visual imagery. Each CNN comprises an input and an output layer, as well as multiple hidden layers. In neural networks, each node or neuron receives an input from some number of locations in the previous layer. Each neuron computes an output value by applying some function to the input values coming from the previous layer. The function that is applied to the input values is specified by a vector of weights and a bias (typically real numbers). Learning in a neural network progresses by making incremental adjustments to the biases and weights. The vector of weights and the bias are called a filter and represents some feature of the input (e.g., a particular shape).

The machine learning layer 200 operates in a training mode to train each of the CNNs 200A-200D prior to the echo workflow engine 12 being placed in an analysis mode to automatically recognize and analyze echo images in patient studies. In one embodiment, the CNNs 200A-200D may be trained to recognize and segment the various echo image views using thousands of echo images from an online public or private echocardiogram DICOM database.

The presentation layer 202 is used to format and present information to a user. In one embodiment, the presentation layer is written in HTML 5, Angular 4 and/or JavaScript. The presentation layer 202 may include a Windows Presentation Foundation (WPF) graphical subsystem 202A for implementing a light weight browser-based user interface that displays reports and allows a user (e.g., doctor/technician) to edit the reports. The presentation layer 202 may also include an image viewer 202B (e.g., a DICOM viewer) for viewing echo images, and a python server 202C for running the CNN algorithms and generating a file of the results in JavaScript Object Notation (JSON) format, for example.

The database layer 204 in one embodiment comprises a SQL database 204A and other external services that the system may use. The SQL database 204A stores patient study information for individual patient studies, including cardiac biomarker measurements input to the system. In some embodiments, the database layer 204 may also include the image file archive 18 of FIG. 1A.

FIG. 3 is a flow diagram illustrating one embodiment of a process for performed by the echo workflow engine 12 to diagnose heart disease based on both cardiac biomarker measurements and AI recognition of both 2D and Doppler modality echo images to perform automated measurements. The process occurs once the echo workflow engine 12 is trained and placed in analysis mode.

The process may begin by the echo workflow engine 12 receiving from a memory one or more patient studies comprising i) one or more cardiac biomarker measurements derived from a patient sample, and ii) a plurality of echocardiogram images taken by an ultrasound device of a patient organ, such as a heart (block 300). In embodiments, the cardiac biomarker measurements may be obtained directly from a local or remote source, including from a handheld point-of-care (POC) device, such as the POC device 42A shown in FIG. 1D. In another embodiment, the cardiac biomarker measurements may be obtained through traditional lab test. The cardiac biomarker measurements may be stored in a patient study and/or an archive, such as in an electronic medical record system (EMR) record and the like. In one embodiment, the patient study may include 70-90 images and videos.

A first module of the echo workflow engine 12 may be used to operate as a filter to separate the plurality of echocardiogram images according to 2D images and Doppler modality images based on analyzing image metadata (block 302). The first module analyzes the DICOM tags, or metadata, incorporated in the image, and runs an algorithm based upon the tag information to distinguish between 2D and modality images, and then separate the modality images into either pulse wave, continuous wave, PWTDI or m-mode groupings. A second module of the echo workflow engine 12 may perform color flow analysis on extracted pixel data using a combination of analyzing both DICOM tags/metadata and color content within the images, to separate views that contain color from those that do not. A third module then anonymizes the data by removing metatags that contain personal information and cropping the images to exclude any identifying information. A fourth module then extracts the pixel data from the images and converts the pixel data to numpy arrays for further processing.

Because sonographers do not label the view types in the echo images, one or more of neural networks are used classify the echo images by view type. In one embodiment, a first neural network is used by the echo workflow engine 12 to classify the 2D images by view type (block 304); and a second set of neural networks is used by the echo workflow engine 12 to extract the features from Doppler modality images and to use the extracted features to classify the Doppler modality images by region (block 306). As shown, the processing of 2D images is separate from the processing of Doppler modality images. In one embodiment, the first neural network and the second set of neural networks may implemented using the set of classification convolutional neural networks (CNNs) 200A. In one specific embodiment, a five class CNN may be used to classify the 2D images by view type and an 11 class CNN may be used to classify the Doppler modality images by region. In one embodiment, a plurality of each of type of neural network can be implemented and configured to use a majority voting scheme to determine the optimal answer. For example a video can be divided into still image frames, and each frame may be given a classification label, i.e., of a vote, and the classification label receiving the most votes is applied to classify the video.

In one embodiment, the echo workflow engine 12 is trained to classify many different view types. For example, the echo workflow engine 12 may classify at least 11 different view types including parasternal long axis (PLAX), apical 2-, 3-, and 4-chamber (A2C, A3C, and A4C), A4C plus pulse wave of the mitral valve, A4C plus pulse wave tissue Doppler on the septal side, A4C plus pulse wave tissue Doppler on the lateral side, A4C plus pulse wave tissue Doppler on the tricuspid side, A5C plus continuous wave of the aortic valve, A4C+Mmode (TrV), A5C+PW (LVOT).

Based on the classified images, a third set of neural networks is used by the echo workflow engine 12 to segment regions of interest (e.g., cardiac chambers) in the 2D images to produce annotated or segmented 2D images (block 308). A fourth set of neural networks is used by the echo workflow engine 12 for each classified Doppler modality region to generate waveform traces and to generate annotated or segmented Doppler modality images (block 309). The process of segmentation includes determining locations where each of the cardiac chambers begin and end to generate outlines of structures of the heart (e.g., cardiac chambers) depicted in each image and/or video. Segmentation can also be used to trace the outline of the waveform depicting the velocity of blood flow in a Doppler modality. In one embodiment, the third and fourth sets of neural networks maybe referred to as segmentation neural networks and my comprise the set of

segmentation CNNs

200B and 2000. The choice of segmentation CNN used is determined by the view type of the image, which makes the prior correct classification of view type a crucial step. In a further embodiment, once regions of interest are segmented, a separate neural network can be used to smooth outlines of the segmentations.

As will be explained further below, the segmentation CNNs may be trained from hand-labeled real images or artificial images generated by general adversarial networks (GANs).

Using both the segmented 2D images and the segmented Doppler modality images, the echo workflow engine 12 calculates for the patient study, measurements of cardiac features for both left and right sides of the heart (block 310).

The echo workflow engine 12 then generates conclusions by comparing the one or more cardiac biomarker measurements and calculated measurements of cardiac features with International cardiac guidelines (block 312). The echo workflow engine 12 further outputs at least one report to a user showing ones of the one or more cardiac biomarker measurements and the calculated measurements that fall within or outside of the International cardiac guidelines (block 314). In one embodiment, two reports are generated and output: the first report is a list of the cardiac biomarker measurements and the calculated values for each measurement with the highest confidence as determined by a rules based engine, highlighting values among the measurements that fall outside of the International guidelines; and the second report is a comprehensive list of all cardiac biomarker measurements and echo image measurements calculated on every image frame of every video, in every view, generating large volumes of data. All report data and extracted pixel data may be stored in a structured database to enable machine learning and predictive analytics on images that previously lacked the quantification and labelling necessary for such analysis. The structured database may be exported to a cloud based server or may remain on premises (e.g., of the lab owning the images) and can be connected to remotely. By connecting these data sources into a single network, disease prediction algorithms can be progressively trained across multiple network nodes, and validated in distinct patient cohorts. In one embodiment, the reports may be electronically displayed to a doctor and/or a patient on a display of an electronic device and/or as a paper report. In some embodiments, the electronic reports may be editable by the user per rule or role based permissions, e.g., a cardiologist may be allowed to modify the report, but a patient may have only view privileges.

FIG. 4A is a flow diagram illustrating further details of the process for automatically recognizing and analyze both 2D and Doppler modality echo images to perform automated measurements and the diagnosis, prediction and prognosis of heart disease according to one embodiment.

The process may begin with receiving one or more patient studies (FIG. 3 block 300), which comprises blocks 400-4010. In one embodiment, echo images from each of the patient studies are automatically downloaded into the image file archive 18 (block 400). The cardiac biomarker measurements and the echo images may be received from a local or remote storage source of the computer 14. The local storage sources may include internal/external storage of the computer 14 including removable storage devices. The remote storage sources may include the ultrasound imaging device 24, the POS device 25, the DICOM servers 30, the network file share devices 32, the echo workstations 34, and/or the cloud storage services 36 (see FIG. 1 ). In one embodiment, the echo workflow engine 12 includes functions for operating as a picture archiving and communication server (PACS), which is capable of handling images from multiple modalities (source machine types, one of which is the ultrasound imaging device 24). The echo workflow engine 12 uses PACS to download and store the echo images into the image file archive 18 and provides the echo workflow engine 12 with access to the echo images during the automated workflow. The format for PACS image storage and transfer is DICOM (Digital Imaging and Communications in Medicine).

Patient information, including any cardiac biomarker measurements, from each of the patient studies is extracted and stored in the database 16 (block 402). Non-image patient data may include metadata embedded within the DICOM images and/or scanned documents, which may be incorporated using consumer industry standard formats such as PDF (Portable Document Format), once encapsulated in DICOM. In one embodiment, received patient studies are placed in a processing queue for future processing, and during the processing of each patient study, the echo workflow engine 12 queues and checks for unprocessed echo images (block 404). The echo workflow engine 12 monitors the status of patient studies, and keeps track of them in a queue to determine which have been processed and which are still pending. In one embodiment, prioritization of the patient studies in the queue may be configured by a user. For example, the patient studies may be prioritized in the queue for processing according to the date of the echo exam, the time of receipt of the patient study or by estimated severity of the patient's heart disease.

Any unprocessed echo images are then filtered for having a valid DICOM image format and non DICOM files in an echo study are discarded (block 406). In one embodiment, the echo images are filtered for having a particular type of format, for example, a valid DICOM file format, and any other file formats may be ignored. Filtering the echo images for having a valid image file format enhances the reliability of the echo workflow engine 12 by rejecting invalid DICOM images for processing.

Any unprocessed valid echo images are then opened and processed in the memory of the computer 14 (block 408). Opening of the echo images for the patient study in memory of the computer 14 is done to enhance processing speed by echo workflow engine 12. This is in contrast to an approach of opening the echo files as sub-processes, saving the echo files to disk, and then reopening each echo image during processing, which could significantly slow processing speed.

The echo workflow engine 12 then extracts and stores the metadata from the echo images and then anonymizes the images by blacking out the images and overwriting the metadata in order to protect patient data privacy by covering personal information written on the image (block 410). As an example, DICOM formatted image files include metadata referred to as DICOM tags that may be used to store a wide variety of information such as patient information, Doctor information, ultrasound manufacture information, study information, and so on. In one embodiment, the extracted metadata may be stored in the database 16 and the metadata in image files is over written for privacy.

After receipt and processing of the patient studies, the echo workflow engine 12 separates 2D images from Doppler modality images so the two different image types can be processed by different pipeline flows, described below. In one embodiment, the separating of the images (FIG. 3 block 302) may comprise blocks 412-414. First, the 2D images are separated from the Doppler modality images by analyzing the metadata (block 412).

FIGS. 5A and 5B are diagrams illustrating an example 2D echo image 500 and an example Doppler modality image 502 including a waveform, respectively. The echo workflow engine 12 may determine the image type by examining metadata/DICOM tags. In one embodiment, information within the DICOM tags may be extracted in order to group the images into one of the following four classes: 2D only, pulsed-wave (PW), continuous wave (CW), and m-mode. Similarly, the transducer frequency of the ultrasound imaging device 24 in the metadata may be used to further separate some of the PW images into a fifth class: pulsed-wave tissue doppler imaging (PWTDI).

Referring again to FIG. 4A, the echo workflow engine 12 may also filter out images with a zoomed view, which may also be determined by analyzing the metadata (block 414). Any of the echo images that has been zoomed during image capture are not processed through the pipeline because when zooming, useful information is necessarily left out of the image, meaning the original image would have to be referenced for the missing data, which is a duplication of effort that slows processing speed. Accordingly, rather than potentially slowing the process in such a manner, the echo workflow engine 12 filters out or discards the zoomed images to save processing time. In an alternative embodiment, filtering out zoomed images in block 414 may be performed prior to separating the images in block 412.

After separating the 2D images from the Doppler modality images, the echo workflow engine 12 extracts and converts the image data from each echo image into numerical arrays 416 (block 416). For the echo images that are 2D only, the pixel data comprises a series of image frames played in sequence to create a video. Because the image frames are unlabeled, the view angle needs to be determined. For the Doppler modality images that include waveform modalities, there are two images in the DICOM file that may be used for subsequent view identification, a waveform image and an echo image of the heart. The pixel data is extracted from the DICOM file and tags in the DICOM file determine the coordinates to crop the images. The cropped pixel data is stored in numerical arrays for further processing. In one embodiment, blocks 412, 414 and 416 may correspond to the separating images block 302 of FIG. 3 .

After separating images, the echo workflow engine 12 attempts to classify each of the echo images by view type. In one embodiment, view classification (FIG. 3

blocks

304 and 306) correspond to blocks 418-422.

According to the disclosed embodiments, the echo workflow engine 12 attempts to classify each of the echo images by view type by utilizing parallel pipeline flows. The parallel pipeline includes a 2D image pipeline and a Doppler modality image pipeline. The 2D pipeline flow begins by classifying, by a first CNN, the 2D images by view type (block 418), corresponding to block 304 from FIG. 3 . The Doppler modality image pipeline flow begins by classifying, by a second CNN, the Doppler modality images by view type (block 420), corresponding to block 306 from FIG. 3 .

FIGS. 6A-6K are diagrams illustrating some example view types automatically classified by the echo workflow engine 12. As stated previously, example view types may include parasternal long axis (PLAX), apical 2-, 3-, and 4-chamber (A2C, A3C, and A4C), A4C plus pulse wave of the mitral valve, A4C plus pulse wave tissue Doppler on the septal side, A4C plus pulse wave tissue Doppler on the lateral side, A4C plus pulse wave tissue Doppler on the tricuspid side, A5C plus continuous wave of the aortic valve, A4C+Mmode (TrV), A5C+PW (LVOT).

Referring again to FIG. 4A, in one embodiment, 2D image classification is performed as follows. If the DICOM file contains video frames from a 2D view, only a small subset of the video frames are analyzed to determine 2D view classification for more efficient processing. In one embodiment, the subset of the video frames may range approximately 8-12 video frames, but preferably 10 frames are input into one of the CNNs 200A trained for 2D to determine the actual view. In an alternative embodiment, subset a video frames may be randomly selected from the video file. In one embodiment, the CNNs 200A classify each of the analyzed video frames as one of: A2C, A3C, A4C, A5C, PLAX Modified, PLAX, PSAX AoV level, PSAX Mid-level, Subcostal Ao, Subcostal Hep vein, Subcostal IVC, Subcostal LAX, Subcostal SAX, Suprasternal and Other.

Doppler modality images comprise two images, an echocardiogram image of the heart and a corresponding waveform, both of which are extracted from the echo file for image processing. In one embodiment, Doppler modality image classification of continuous wave (CW), pulsed-wave (PW), and M-mode images is performed as follows. If the DICOM file contains a waveform modality (CW, PW, PWTDI, M-mode), the two extracted images are input to one of the CNNs 200A trained for CW, PW, PWTDI and M-mode view classification to further classify the echo images as one of: CW (AoV), CW (TrV), CW Other, PW (LVOT), PW (MV), PW Other, PWTDI (lateral), PWTDI (septal), PWTDI (tricuspid), M-mode (TrV) and M-mode Other.

There are many more potential classifications available for modalities, but the present embodiments strategically select the classes above, while grouping the remaining potential classes into “Other”, in order to maximize processing efficiency, while identifying the most clinically important images for further processing and quantification. Customization of the CNNs 200A occurs in the desired number of layers used and the quantity of filters within each layer. During the training phase, the correct size of the CNNs may be determined through repeated training and adjustments until optimal performance levels are reached.

During view classification, the echo workflow engine 12 maintains classification confidence scores that indicate a confidence level that the view classifications are correct. The echo workflow engine 12 filters out the echo images having classification confidence scores that fail to meet a threshold, i.e., low classification confidence scores (block 422). Multiple algorithms may be employed to derive classification confidence scores depending upon the view in question. Anomalies detected in cardiac structure annotations, image quality, cardiac cycles detected and the presence of image artifacts may all serve to decrease the classification confidence score and discard an echo image out of the automated echo workflow.

With respect to the confidence scores, the echo workflow engine 12 generates and analyzes several different types of confidence scores at different stages of processing, including classification, annotation, and measurements (e.g., blocks 422, 434 and 442). For example, poor quality annotations or classifications, which may be due to substandard image quality, are filtered out by filtering the classification confidence scores. In another example, in a patient study the same view may be acquired more than once, in which case the best measurements are chosen by filtering out low measurement confidence scores as described further below in block 442. Any data having a confidence score that meets a predetermined threshold continues through the workflow. Should there be a duplication of measurements both with high confidence, the most clinically relevant measurement may be chosen.

Next the echo workflow engine 12 performs image segmentation to define regions of interest (ROI). In computer vision, image segmentation is the process of partitioning a digital image into multiple segments (sets of pixels) to locate and boundaries (lines, curves, and the like) of objects. Typically, annotations are a series of boundary lines overlaying overlaid on the image to highlight segment boundaries/edges. In one embodiment, the segmentation to define ROI (FIG. 3 block 308) corresponds to blocks 426-436.

In one embodiment, the 2D image pipeline annotates, by a third CNN, regions of interests, such as cardiac chambers in the 2D images, to produce annotated 2D images (block 426). An annotation post process then erodes the annotations to reduce their dimensions, spline fits outlines of cardiac structures and adjusts locations of the boundary lines closer to the region of interest (ROIs) (block 427). The 2D image pipeline continues with analyzing the ROIs (e.g., cardiac chambers) in the annotated 2D images to estimate volumes and determine key points in the cardiac cycle by finding systolic/diastolic end points (block 430). For 2D only views, measurements are taken at the systolic or diastolic phase of the cardiac cycle, i.e. when the left ventricle reaches the smallest volume (systole) or the largest volume (diastole). From the 2D video images, it must be determined which end points are systolic and which are diastolic based on the size of the estimated volumes of the left ventricle. For example a significantly large left ventricle may indicate a dystonic end point, while a significantly small volume may indicate a systolic end point. Every video frame is annotated and the volume of the left ventricle is calculated throughout the whole cardiac cycle. The frames with minimum and maximum volumes are detected with a peak detection algorithm.

The Doppler modality pipeline analyzes the Doppler modality images and generates, by a fourth CNN, a mask and a waveform trace in the Doppler modality images to produce annotated Doppler modality images (block 431).

FIG. 7 is a diagram illustrating an example 2D image segmented in block 426 to produce annotations 700 indicating cardiac chambers, and an example Doppler modality image segmented in block 431 to produce a mask and waveform trace 702.

In one embodiment, the third and fourth CNNs may correspond to segmentation CNNs 200B. In one embodiment, each of the CNNs 200B used to segment the 2D images and Doppler modality images may be implemented as U-Net CNN, which is convolutional neural network developed for biomedical image segmentation. Multiple U-Nets may be used. For example, for 2D images, a first U-Net CNN can be trained to annotate ventricles and atria of the heart from the A2C, A3C, A4C views. A second U-net CNN can be trained to annotate the chambers in the PLAX views. For M-mode views, a third U-Net CNN can be trained to segment the waveform, remove small pieces of the segments to find likely candidates for the region of interest, and then reconnect the segments to provide a full trace of the movement of the mitral valve. For CW views, a fourth U-net CNN can be trained to annotate and trace blood flow. For PW views, a fifth U-net CNN trained to annotate and trace the blood flow. For PWTDI views, a sixth U-net CNN can be trained to annotate and trace movement of the tissues structures (lateral/septal/tricuspid valve).

Referring again to FIG. 4A, the Doppler modality pipeline continues by processing the annotated Doppler modality images with a sliding window to identify cycles, peaks are measured in the cycles, and key points in the cardiac cycle are determined by finding systolic/diastolic end points (block 432). Typically a Doppler modality video may capture three heart cycles and the sliding window is adjusted in size to block out two of the cycles so that only one selected cycle is analyzed. Within the selected cycle, the sliding window is used to identify cycles, peaks are measured in the cycles, and key points in the cardiac cycle are determined by finding systolic/diastolic end points.

FIGS. 8A and 8B are diagrams illustrating examples of finding systolic/diastolic end points in the cardiac cycle, for both 2D and Doppler modalities, respectively, which are key points in order to take accurate cardiac measurements.

Referring again to FIG. 4A, in one embodiment, the echo workflow engine 12 maintains annotation confidence scores corresponding to the estimated volumes, systolic/diastolic end points, identified cycles and measured peaks. The echo workflow engine 12 filters out annotated images having annotation confidence scores that fail to meet a threshold, i.e., low annotated confidence scores (block 434). Examples of low confidence annotations may include annotated images having one or more of: excessively small areas/volumes, sudden width changes, out of proportion sectors, partial heart cycles, and insufficient heart cycles.

After images having low annotation confidence scores are filtered out, the echo workflow engine 12 defines an imaging window for each image, and filters out annotations that lie outside of the imaging window (block 435).

FIGS. 9A and 9B are diagrams illustrating processing of an imaging window in a 2D echo image 900. In FIG. 9A, a shaded ROI 902 is shown having unshaded regions or holes 904 therein. Open CV morphology transformations are used to fill the holes 904 inside the ROI, as shown in FIG. 9B. Thereafter, a Hough line transformation may be used to find an imaging window border 906. As is well known, a Hough transform is a feature extraction technique used in digital image processing to find imperfect instances of objects within a certain class of shapes. After an imaging window border is are found, a pixel account of annotations beyond the imaging window border is made. Annotations 908 with a significant number of pixels outside the border of the imaging window are then discarded.

Referring again to FIG. 4A, in the handheld configuration for the automated clinical workflow system 10C (See FIG. 1B), the patient studies may not include Doppler modality images. According to a further aspect, the disclosed embodiments accommodate for such handheld configurations by using the 2D images to simulate Doppler modality measurements by using Left Ventricular (LV) and Left Atrial (LA) volume measurements to derive E, e′ and A (e.g., early and late diastolic transmittal flow and early/mean diastolic tissue velocity) measurements (block 436). In one embodiment, simulating the Doppler modality measurements may be optional and may be invoked based on a software setting indicating the presence of a handheld configuration and/or absence of Doppler modality images for the current patient study.

Referring again to FIG. 4A, in one embodiment, once cardiac features are annotated during segmentation, the cardiac features are then measured during a measurement process (block 310 of FIG. 3 ), which in one embodiment may comprises block 438-448. The process of measuring cardiac features may begin by quantifying a plurality of measurements using the annotations. First, the 2D pipeline measures for the 2D images left/right ventricle, left/right atriums, left ventricular outflow (LVOT) and pericardium (block 438). For the Doppler modality images, the Doppler modality image pipeline measures blood flow velocities (block 440).

More specifically, for A2C, A3C, A4C, and A5C image views, volumetric measurements of chamber size are conducted on the systolic and diastolic frames of the video, and image processing techniques mimic a trained clinician at measuring the volume using the method of disks (MOD). For 2D Plax, PSAX (mid level), PSAX (AoV level), Subcostal, Suprasternal and IVC image views, linear measurements of chamber size and inter-chamber distances are conducted on the systolic and diastolic frames of the video using image processing techniques to mimic the trained clinician. For M-mode image views, from the annotated segments of the movement of the Tricuspidvalve, a center line is extracted and smoothed, and then the peaks and valleys are measured in order to determine the minimum and maximum deviations over the cardiac cycle. For PW image views, from the annotations of the blood flow, a mask is created to isolate parts of the waveform. A sliding window is then run across the trace to identify one full heart cycle, in combination with heart rate data from the DICOM tags, to use as a template. This template is then used to identify all other heart cycles in the image. Peak detection is then performed on each cycle and then run through an algorithm to identify which part of the heart cycle each peak represents. For CW image views, from the annotations of the trace of the blood flow, curve fitting is performed on the annotation to then quantify the desired measurements. For PWTDI image views, from the annotations of the movement of the tissue, a mask is created to isolate parts of the waveform. A sliding window is then run across the trace to identify one full heart cycle, in combination with heart rate data from the DICOM tags, to use as a template. This template is then used to identify all other heart cycles in the image. Peak detection is then performed on each cycle and then run through an algorithm to identify which part of the heart cycle each peak represents.

FIGS. 10A and 10B are diagrams graphically illustrating structural measurements automatically generated from annotations of cardiac chambers in a 2D image, and velocity measurements automatically generated from annotations of waveforms in a Doppler modality, respectively.

The measurement table below list the measurements that may be compiled by the echo workflow engine 12 according to one embodiment.

Measurement Table


Measurement Name	Measurement Description

LAESV MOD A2C	Left Atrial End Systolic Volume in A2C calculation
	based on Method Of Discs
LAESVi MOD A2C	Left Atrial End Systolic Volume in A2C calculation
	based on Method Of Discs indexed to BSA
LAL A2C	Left Atrial End Sysolic Length measured in A2C
LVEDV MOD A2C	Left Ventricular End Diastolic Volume in A2C
	calculation based on Method Of Discs
LVEDVi MOD A2C	Left Ventricular End Diastolic Volume in A2C
	calculation based on Method Of Discs indexed to BSA
LVEF MOD A2C	Left Ventricular Ejection Fraction in A2C based on
	Method Of Discs
LVESV MOD A2C	Left Ventricular End Systolic Volume in A2C
	calculation based on Method Of Discs
LVESVi MOD A2C	Left Ventricular End Systolic Volume in A2C
	calculation based on Method Of Discs indexed to BSA
LV length A2C	Left Ventricular Length measured in A2C
LAESV MOD A4C	Left Atrial End Systolic Volume in A4C calculation
	based on Method Of Discs
LAESVi MOD A4C	Left Atrial End Systolic Volume in A4C calculation based
	on Method Of Discs indexed to BSA
LAL A4C	Left Atrial End Systolic Length measured in A4C
LAW A4C	Left Atrial End Systolic Width measurement in A4C
LA area A4C	Left Atrial Area measured in A4C
LAESV A-L A4C	Left Atrial End Systolic Volume in A4C calculation
	based on Area-Length method
LVEDV MOD A4C	Left Ventricular End Diastolic Volume in A4C
	calculation based on Method Of Discs
LVEDVi MOD A4C	Left Ventricular End Diastolic Volume in A4C
	calculation based on Method Of Discs indexed to BSA
LVEF MOD A4C	Left Ventricular Ejection Fraction in A4C based on
	Method Of Discs
LVESV MOD A4C	Left Ventricular End Systolic Volume in A4C
	calculation based on Method Of Discs
LVESVi MOD A4C	Left Ventricular End Systolic Volume in A4C
	calculation based on Method Of Discs indexed to BSA
LV length A4C	Left Ventricular Length measured in A4C
LVAd A4C	Left Ventricular Area measured at end diastole in A4C
TAPSE	Tricuspid Annular Plane Systolic Excursion
DecT	Deceleration Time of early diastolic MV transmitral flow
E/A ratio	Ratio of early and late diastolic transmitral flow
MV-A	Late diastolic transmitral flow
MV-Adur	Duration of late diastolic transmitral flow
MV-E	Early diastolic transmitral flow
e′ lateral	Early diastolic tissue velocity taken from the lateral
	region
e′ mean	Mean early diastolic tissue velocity (mean of lateral
	and septal region)
e′ septal	Early diastolic tissue velocity taken from the septal
	region
E/e′ lateral	Ratio of early transmitral flow and early diastolic tissue
	velocity taken form the lateral region
E/e′ septal	Ratio of early transmitral flow and early diastolic tissue
	velocity taken form the septal region
E/e′ mean	Ratio of early transmitral flow and mean diastolic
	tissue velocity
a′ lateral	Late diastolic tissue velocity taken from the lateral
	region
a′ septal	Late diastolic tissue velocity taken from the septal
	region
s′ lateral	Systolic tissue velocity taken from the lateral region
s′ septal	Systolic tissue velocity taken from the septal region
LAESV A-L biplane	Left Atrial End Systolic Volume biplane calculation
	based on Area-Length method
LAESV MOD biplane	Left Atrial End Systolic Volume biplane calculation
	based on Method Of Discs
LAESVi MOD biplane	Left Atrial End Systolic Volume biplane calculation
	based on Method Of Discs indexed to BSA
LAESVi A-L biplane	Left Atrial End Systolic Volume biplane calculation
	based on Area-Length method indexed to BSA
LVCO MOD biplane	Left Ventricular Cardiac Output biplane calculation
	based on Method Of Discs
LVEDV MOD biplane	Left Ventricular End Diastolic Volume biplane
	calculation based on Method Of Discs
LVEDVi MOD biplane	Left Ventricular End Diastolic Volume biplane
	calculation based on Method Of Discs indexed to BSA
LVEF MOD biplane	Left Ventricular Ejection Fraction biplane based on
	Method Of Discs
LVESV MOD biplane	Left Ventricular End Systolic Volume biplane
	calculation based on Method Of Discs
LVESVi MOD biplane	Left Ventricular End Systolic Volume biplane
	calculation based on Method Of Discs indexed to BSA
LVSV MOD biplane	Left Ventricular Stroke Volume biplane calculation
	based on Method of Disks
AoV Vmax	Aortic Valve maximum Velocity
AoV Vmean	Aortic Valve mean Velocity
AoV Pmax	Aortic Valve maximum Pressure gradient
LVOT Vmax	Left Ventricular Outflow Tract maximum Velocity
LVOT Vmean	Left Ventricular Outflow Tract mean Velocity
IVSd	Inter Ventricular Septal thickness measured end
	diastolic
LV mass	Left Ventricular mass
LVIDd	Left Ventricular internal Diameter measured at end
	diastole
LVIDd index	Left Ventricular internal Diameter measured at end
	diastole indexed to BSA
LVIDs	Left Ventricular internal Diameter measured at end
	systole
LVIDs index	Left Ventricular internal Diameter measured at end
	systole indexed to BSA
LVMi	Left Ventricular Mass indexed to BSA
LVOT	Left Ventricular Outflow Tract diameter
LVPWd	Left Ventricular Posterior Wall thickness measured
	end diastolic
RWT	Relative Wall Thickness
LA area A2C	Left Atrial Area measured in A2C
LAESV A-L A2C	Left Atrial End Systolic Volume in A2C calculation
	based on Area-Length method
LVAd A2C	Left Ventricular Area measured at end diastole in A2C
LVAs A2C	Left Ventricular Area measured at end systole in A2C
LVAs A4C	Left Ventricular Area measured at end systole in A4C
PASP	Pulmonary Artery Systolic Pressure
RAL	Right Atrial End Systolic Length
RAW	Right Atrial End Systolic Width
RAESV MOD A4C	Right Atrial end systolic Volume in A4C calculation
	based on Method Of Discs
RAESV A-L A4C	Right Atrial end systolic Volume in A4C calculation
	based on Area-Length method
RAESVi MOD A4C	Right Atrial end systolic Volume in A4C calculation
	based on Method Of Discs indexed to BSA
RA area	Right Atrial area
RVIDd	Right Ventricular End Diastolic Internal Diameter
RV area (d)	Right Ventricular Area (measured at end-diastole)
RV area (s)	Right Ventricular Area (measured at end systole)
LVOT Pmax	Left Ventricular Outflow Tract max pressure gradient
LVOT Pmean	Left Ventricular Outflow Tract mean pressure gradient
LVSV (Doppler)	Left Ventricular Stroke Volume based on Doppler
LVOT VTI	Left Ventricular Outflow Tract Velocity Time Integral
LVCO (Doppler)	Left Ventricular Cardiac Output (based on Doppler)
LVCOi (Doppler)	Left Ventricular Cardiac Output (based on Doppler)
	indexed to Body Surface Area
LVSVi (Doppler)	Left Ventricular Stroke Volume (based on Doppler)
	indexed to Body Surface Area
TR Vmax	Tricuspid Regurgitation maximum velocity
CSA LVOT	Crossectional Area of the LVOT
Sinotub J	Sinotubular junction diameter
Sinus valsalva	Sinus valsalva diameter
Asc. Ao	Ascending Aorta diameter
Asc. Ao index	Ascending Aorta diameter index
Sinus valsalva index	Sinus valsalva diameter indexed to BSA
IVC max	Inferior Vena Cava maximum diameter
IVC min	Inferior Vena Cava minimum diameter
IVC Collaps	Inferior Vena Cava collaps
RVIDd mid	Right Ventricular Internal Diameter at mid level
	(measured end diastole)
RVOT prox	Right Ventricular Outflow Tract proximal diameter
RVOT dist	Right Ventricular Outflow Tract distal diameter
RV FAC	Right Ventricular Fractional Area Change
TEI index
RVAWT	Right Ventricular Anterior Wall Thickness
TrV-E	Tricuspid valve E wave
TrV-A	Tricuspid valve A wave
TrV E/A	Tricuspid valve E/A ratio
TrV DecT	Tricuspid valve deceleration time
MV Vmax	Mitral valve maximum velocity
MV Vmean	Mitral valve mean velocity
MV VTI	Mitral valve velocity time intergal
MV PHT	Mitral valve pressure half time
MVA (by PHT)	Mitral valve area (by pressure half time)
RV e′	Early diastolic tissue velocity taken from the right
	ventricular free wall region
RV a′	Late diastolic tissue velocity taken from the right
	ventricular free wall region
RV s′	Systolic tissue velocity taken from the right ventricular
	free wall region
RVCO	Right Ventricular Cardiac Output
ULS	Unidimensional Longitudinal Strain
Ao-arch	Aortic arch diameter
Descending Ao	Descending Aortic diameter
Ao-arch index	Aortic arch diameter indexed to BSA
Descending Ao index	Descending Aortic diameter indexed to BSA
LA GLS (reservoir) (A4C)	Left Atrial strain during systole measured in A4C
LA GLS (conduit) (A4C)	Left Atrial strain during early diastole measured in A4C
LA GLS (booster) (A4C)	Left Atrial strain during pre atrial contraction measured
	in A4C
LA GLS (reservoir) (A2C)	Left Atrial strain during systole measured in A2C
LA GLS (conduit) (A2C)	Left Atrial strain during early diastole measured in A2C
LA GLS (booster) (A2C)	Left Atrial strain during pre atrial contraction measured
	in A2C
LA GLS (reservoir)	Left Atrial strain during systole
LA GLS (conduit)	Left Atrial strain during early diastole
LA GLS (booster)	Left Atrial strain during pre atrial contraction
LVSr-e (A4C)	Left Ventricular strain rate during early diastole
	measured in A4C
LVSr-a (A4C)	Left Ventricular strain rate during late diastole
	measured in A4C
LVSr-s (A4C)	Left Ventricular strain rate during systole measured in
	A4C
LVSr-e (A2C)	Left Ventricular strain rate during early diastole
	measured in A2C
LVSr-a (A2C)	Left Ventricular strain rate during late diastole
	measured in A2C
LVSr-s (A2C)	Left Ventricular strain rate during systole measured in
	A2C
LVSr-e	Left Ventricular strain rate during early diastole
LVSr-a	Left Ventricular strain rate during late diastole
LVSr-s	Left Ventricular strain rate during systole
LASr-e (A4C)	Left Atrial strain rate during early diastole
LASr-a (A4C)	Left Atrial strain rate during late diastole
LASr-s (A4C)	Left Atrial strain rate during systole
LASr-e (A2C)	Left Atrial strain rate during early diastole
LASr-a (A2C)	Left Atrial strain rate during late diastole
LASr-s (A2C)	Left Atrial strain rate during systole
LASr-e	Left Atrial strain rate during early diastole
LASr-a	Left Atrial strain rate during late diastole
LASr-s	Left Atrial strain rate during systole
AV-S (A4C)	Atrio Ventricular strain measured in A4C
AV-S (A2C)	Atrio Ventricular strain measured in A2C
AV-S	Atrio Ventricular strain
Sr-Sav (A4C)	Atrio Ventricular strain rate during systole measured in
	A4C
Sr-Eav (A4C)	Atrio Ventricular strain rate during early diastole
	measured in A4C
Sr-Aav (A4C)	Atrio Ventricular strain rate during late diastole
	measured in A4C
Sr-Sav (A2C)	Atrio Ventricular strain rate during systole measured in
	A2C
Sr-Eav(A2C)	Atrio Ventricular strain rate during early diastole
	measured in A2C
Sr-Aav (A2C)	Atrio Ventricular strain rate during late diastole
	measured in A2C
Sr-Sav	Atrio Ventricular strain rate during systole
Sr-Eav	Atrio Ventricular strain rate during early diastole
Sr-Aav	Atrio Ventricular strain rate during late diastole
LVVr-e (A4C)	Left Ventricular volume rate during early diastole
	measured in A4C
LVVr-a (A4C)	Left Ventricular volume rate during late diastole
	measured in A4C
LVVr-s (A4C)	Left Ventricular volume rate during systole measured
	in A4C
LVVr-e (A2C)	Left Ventricular volume rate during early diastole
	measured in A2C
LVVr-a (A2C)	Left Ventricular volume rate during late diastole
	measured in A2C
LVVr-s (A2C)	Left Ventricular volume rate during systole measured
	in A2C
LVVr-e	Left Ventricular volume rate during early diastole
LVVr-a	Left Ventricular volume rate during late diastole
LVVr-s	Left Ventricular volume rate during systole
LAVr-e (A4C)	Left Atrial volume rate during early diastole measured
	in A4C
LAVr-a (A4C)	Left Atrial volume rate during late diastole measured in
	A4C
LAVr-s (A4C)	Left Atrial volume rate during systole measured in A4C
LAVr-e (A2C)	Left Atrial volume rate during early diastole measured
	in A2C
LAVr-a (A2C)	Left Atrial volume rate during late diastole measured in
	A2C
LAVr-s (A2C)	Left Atrial volume rate during systole measured in A2C
LAVr-e	Left Atrial volume rate during early diastole
LAVr-a	Left Atrial volume rate during late diastole
LAVr-s	Left Atrial volume rate during systole
TLVd	Total Left heart volume end-diastolic
TLVs	Total Left heart volume end-systolic
TLVd (A4C)	Total Left heart volume end-diastolic measured in A4C
TLVs (A4C)	Total Left heart volume end-systolic measured in A4C
TLVd (A2C)	Total Left heart volume end-diastolic measured in A2C
TLVs (A2C)	Total Left heart volume end-systolic measured in A2C
Ar	Pulmonary vein Atrial reversal flow
Ardur	Pulmonary vein Atrial reversal flow duration
D	Pulmonary vein diastolic flow velocity
S	Pulmonary vein systolic flow velocity
S/D ratio	Ratio of Pulmonary vein systolic- and diastolic flow
	Vel.
RV GLS	Right Ventricular Global Longitudinal Strain (mean)
RV GLS (A4C)	Right Ventricular Global Longitudinal Strain measured
	in A4C
RV GLS (A2C)	Right Ventricular Global Longitudinal Strain measured
	in A2C
RV GLS (A3C)	Right Ventricular Global Longitudinal Strain measured
	in A3C
LA GLS	Left Atrial Global Longitudinal Strain (mean)
LA GLS (A4C)	Left Atrial Global Longitudinal Strain measured in A4C
LA GLS (A2C)	Left Atrial Global Longitudinal Strain measured in A2C
LA GLS (A3C)	Left Atrial Global Longitudinal Strain measured in A3C
RA GLS	Right Atrial Global Longitudinal Strain (mean)
RA GLS (A4C)	Right Atrial Global Longitudinal Strain measured in
	A4C
RA GLS (A2C)	Right Atrial Global Longitudinal Strain measured in
	A2C
RA GLS (A3C)	Right Atrial Global Longitudinal Strain measured in
	A3C
PV Vmax	Pulmonary Valve maximum Velocity
PV Vmean	Pulmonary Valve mean Velocity
PV Pmax	Pulmonary Valve maximum Pressure gradient
PV Pmean	Pulmonary Valve mean Pressure gradient
PV VTI	Pulmonary Valve Velocity Time Integral
MV-Adur - Ardur	Difference between late diastolic transmitral flow and
	pulmonary vein atrial reversal flow duration
APC	Arteria pulmonalis communis
LV eccentricity index	LV eccentricity index
Mean % WT A2C	Mean percentual Wall Thickening of 6 segments in A2C
AA-% WT	Percentile wall thickening of apical anterior segment
AA-WTd	Wall thickness of apical anterior segment in diastole
AA-WTs	Wall thickness of apical anterior segment in systole
AI-% WT	Percentile wall thickening of apical inferior segment
AI-WTd	Wall thickness of apical inferior segment in diastole
AI-WTs	Wall thickness of apical inferior segment in systole
BA-% WT	Percentile wall thickening of basal anterior segment
BA-WTd	Wall thickness of basal anterior segment in diastole
BA-WTs	Wall thickness of basal anterior segment in systole
BI-% WT	Percentile wall thickening of basal interior segment
BI-WTd	Wall thickness of basal interior segment in diastole
BI-WTs	Wall thickness of basal interior segment in systole
MA-% WT	Percentile wall thickening of mid anterior segment
MA-WTd	Wall thickness of mid anterior segment in diastole
MA-WTs	Wall thickness of mid anterior segment in systole
MI-% WT	Percentile wall thickening of mid inferior segment
MI-WTd	Wall thickness of mid inferior segment in diastole
MI-WTs	Wall thickness of mid inferior segment in systole
Pericardial effusion	Pericardial effusion
Mean % WT A3C	Mean percentual Wall Thickening of 6 segments in A3C
AAS-%WT	Percentile wall thickening of apical antero-septal
	segment
AAS-WTd	Wall thickness of apical antero-septal segment in
	diastole
AAS-WTs	Wall thickness of apical antero-septal segment in
	systole
AP-% WT	Percentile wall thickening of apical posterior segment
AP-WTd	Wall thickness of apical posterior segment in diastole
AP-WTs	Wall thickness of apical posterior segment in systole
BAS-% WT	Percentile wall thickening of basal antero-septal
	segment
BAS-WTd	Wall thickness of basal antero-septal segment in
	diastole
BAS-WTs	Wall thickness of basal antero-septal segment in
	systole
BP-% WT	Percentile wall thickening of basal posterior segment
BP-WTd	Wall thickness of basal posterior segment in diastole
BP-WTs	Wall thickness of basal posterior segment in systole
MAS-% WT	Percentile wall thickening of mid antero-septal
	segment
MAS-WTd	Wall thickness of mid antero-septal segment in
	diastole
MAS-WTs	Wall thickness of mid antero-septal segment in systole
MP-% WT	Percentile wall thickening of mid posterior segment
MP-WTd	Wall thickness of mid posterior segment in diastole
MP-WTs	Wall thickness of mid posterior segment in systole
Mean % WT A4C	Mean percentual Wall Thickening of 6 segments in A4C
AL-% WT	Percentile wall thickening of apical lateral segment
AL-WTd	Wall thickness of apical lateral segment in diastole
AL-WTs	Wall thickness of apical lateral segment in systole
AS-% WT	Percentile wall thickening of apical septal segment
AS-WTd	Wall thickness of apical septal segment in diastole
AS-WTs	Wall thickness of apical septal segment in systole
BL-% WT	Percentile wall thickening of basal lateral segment
BL-WTd	Wall thickness of basal lateral segment in diastole
BL-WTs	Wall thickness of basal lateral segment in systole
BS-% WT	Percentile wall thickening of basal septal segment
BS-WTd	Wall thickness of basal septal segment in diastole
BS-WTs	Wall thickness of basal septal segment in systole
ML-% WT	Percentile wall thickening of mid lateral segment
ML-WTd	Wall thickness of mid lateral segment in diastole
ML-WTs	Wall thickness of mid lateral segment in systole
MS-% WT	Percentile wall thickening of mid septal segment
MS-WTd	Wall thickness of mid septal segment in diastole
MS-WTs	Wall thickness of mid septal segment in systole
Global % WT	Global percentual Wall Thickening of the Left Ventricle
AoV Vmean	Aortic Valve mean Velocity
AoV Pmean	Aortic Valve mean Pressure gradient
AOV VTI	Aortic Valve Velocity Time Integral
AVA Vmax	Aortic Valve Area (measured by max Vel.)
AVA VTI	Aortic valve Area (measured by Velocity Time Integral)
AVAi Vmax	Aortic Valve Area (measured by maximum Velocity)
	indexed to Body Surface Area
AVAi VTI	Aortic valve Area (measured by Velocity Time Integral)
	indexed to Body Surface Area
ivrt	IsoVolumic Relaxation Time
LV GLS (A4C)	Left Ventricular Global Longitudinal Strain measured
	in A4C
LV GLS (A2C)	Left Ventricular Global Longitudinal Strain measured
	in A2C
LV GLS (A3C)	Left Ventricular Global Longitudinal Strain measured
	in A3C
LV GLS	Left Ventricular Global Longitudinal Strain (mean)

Referring again to FIG. 4A, in one embodiment, the echo workflow engine 12 maintains measurement confidence scores corresponding to the measured left/right ventricles, left/right atriums, LVOTs, pericardiums and measured velocities. The echo workflow engine 12 filters out echo images having measurement confidence scores that fail to meet a threshold, i.e., low measurement confidence scores (block 442).

Measurement of cardiac features continues with calculating longitudinal strain graphs using the annotations generated by the CNNs (block 444). Thereafter, a fifth CNN is optionally used to detect pericardial effusion 446.

Referring again to FIG. 4A, for all remaining non-filtered out data, the echo workflow engine 12 selects as best measurement data the measurements associated with cardiac chambers with the largest volumes, and saves with the best measurement data, image location, classification, annotation and other measurement data associated with the best measurements (block 447).

FIG. 12A is a diagram graphically illustrating an example set of best measurement data 1200 based on largest volume cardiac chambers and the saving of the best measurement data 1200 to a repository, such as database 16 of FIG. 1 .

Referring again to FIG. 4A, the echo workflow engine 12 then generates conclusions by inputting the cardiac biomarker measurements and the best measurement data 1200 to a set of rules based on international measurement guidelines to generate conclusions for medical decisions support (block 448). The following is an example rule set based on International cardiac guidelines in which the HF diagnosis is based on a point system:

1) If any of the following measurement values are true:

- septal e′<7 cm/s, or
- lateral e′<10 cm/s, or
- Average E/e′>=15, or
- TR velocity>2.8 m/s and PASP>35 mmHg,
- then add 2 points.

2) If any of the following measurement values are true:

- LAVI>34 ml/m², or
- LVMI>=149/122 g/m2 (m/w) and RWT>0.42,
- then add 2 points.

3) If the patient is in sinus rhythm any of the following measurement values are true:

- NT-proBNP>220 pg/ml, or
- BNP>80 pg/ml,
- then add 2 points.

4) If the patient has atrial fibrillation and any of the following measurement values are true:

- NT-proBNP>660 pg/ml, or
- BNP>240 pg/ml,
- then add 2 points.

5) If any of the following are true:

- Average E/e′=9-14 or
- GLS<16%,
- then add 1 point.

6) If any of the following are true:

- LAVI 29/34 ml/m², or
- LVMI>115/95 g/m²(m/w), or
- RWT>0.42, or
- LV wall thickness>=12 mm
- then add 1 point.

7) If the patient is in sinus rhythm and any of the following are true:

- NT-proBNP 125-220 pg/ml, or
- BNP 35-80 pg/ml,
- then add 1 point.

8) If the patient has atrial fibrillation and any of the following are true:

- NT-proBNP 365-660 pg/ml, or
- BNP 105/240 pg/ml,
- then add 1 point.

9) If total point score equals 2, 3 or 4, then determine that a Diastolic Stress Test or Invasive Haemodynamic Measurements are required. If total points equals 5 or more, then diagnosis a high probability that the patient has HFpEF.

FIG. 12B is a diagram graphically illustrating the input of normal LV EF measurements into a set of rules to determine a conclusion that the patient has normal diastolic function, diastolic dysfunction, or indeterminate.

Referring again to FIG. 4A, after the conclusions are generated, a report is generated and output (FIG. 3 block 314), which may comprise blocks 450-456. Report generation may begin by the echo workflow engine 12 outputting the cardiac biomarker measurements and the best measurement data 1200 to a JSON file for flexibility of export to other applications (block 450).

Referring again to FIG. 4A, a lightweight browser-based user interface (UI) displayed showing a report that visualizes the cardiac biomarker measurements and the best measurement data 1200 from the JSON file and that is editable by a user (e.g., doctor/technician) for human verification (block 452). As is well known, a lightweight web browser is a web browser that is optimized to reduce consumption of system resources, particularly to minimize memory footprint, and by sacrificing some of the features of a mainstream web browser. In one embodiment, any edits made to the data are stored in the database 16 and displayed in the UI.

In order to make clinically relevant suggestion to the user, the cardiac biomarker measurements and the best measurement data 1200 are automatically compared to current International guideline values and any out of range values are highlighted for the user (block 454).

Referring again to FIG. 4A, the user is provided with an option of generating a printable report showing an automated summary of Main Findings (i.e., a conclusion reached after examination) and underlining measurements of the patient's health (block 456).

In one embodiment, the automated workflow of the echo workflow engine 12 may end at block 456. However, in further aspects of the disclosed embodiments, the process may continue with advance functions, as described below.

FIG. 4B is a flow diagram illustrating advanced functions of the echo workflow engine. According to this embodiment, the echo workflow engine 12, takes as inputs values of the cardiac biomarkers and specific measurements that were automatically derived using machine learning (see block 310), and analyzes the input measurements to determine disease diagnosis/prediction/prognosis versus both disease and matched controls and normal patient controls (block 460). In one embodiment, the echo workflow engine 12 may use the disease predictions to perform diagnosis of any combination of: cardiac amyloidosis, hypertrophic cardiomyopathy, restrictive pericarditis, cardiotoxity, early diastolic dysfunction and Doppler free diastolic dysfunction assessment (block 462). A prognosis in the form of an automated score may then be generated to predict mortality and hospitalizations in the future (block 464).

Echocardiography is key for the diagnosis of heart failure with preserved ejection fraction (HFpEF). However, existing guidelines are mixed in their recommendations for echocardiogram criteria and none of the available guidelines have been validated against gold-standard invasive hemodynamic measurements in HFpEF.

According to one embodiment, the echo workflow engine 12 further generates a diagnostic score for understanding predictions (block 466). Using machine learning, the echo workflow engine 12 validates the diagnostic score against invasively measured pulmonary capillary wedge pressure (PCWP), and determines the prognostic utility of the score in a large HFpEF cohort.

In one embodiment, the echo workflow engine 12, takes as the inputs values, including the measurements that were automatically derived using machine learning workflow, and analyzes the input values using an HFpEF algorithm to compute the HFpEF diagnostic score.

Recognizing that hypertensive heart disease is the most common precursor to HFpEF and has overlapping echocardiogram characteristics with HFpEF, echocardiogram features of 233 patients with HFpEF (LVEF>50%) was compared to 273 hypertensive controls with normal ejection fraction but no heart failure. An agnostic model was developed using penalized logistic regression model and Classification and Regression Tree (CART) analysis. The association of the derived echocardiogram score with invasively measured PCWP was investigated in a separate cohort of 96 patients. The association of the score with the combined clinical outcomes of cardiovascular mortality of HF hospitalization was investigated in 653 patients with HFpEF from the Americas echocardiogram sub study of the TOPCAT trial.

According to one embodiment, left ventricular ejection fraction (LVEF<60%), peak TR velocity (>2.3 m/s), relative wall thickness (RWT>0.39 mm), interventricular septal thickness (>12.2 mm) and E wave (>1 m/s) are selected as the most parsimonious combination of variables to identify HFpEF from hypertensive controls. A weighted score (range 0-9) based on these 5 echocardiogram variables had a combined area under the curve of 0.9 for identifying HFpEF from hypertensive controls.

FIGS. 16A and 16B are diagrams showing a graph A plotting PCWP and HFpEF scores, and a graph B plotting CV mortality or hospitalization for HF, respectively. Graph A shows that in the independent cohort, the HFpEF score was significantly associated with PCWP in patients with HFpEF (R²=0.22, P=0.034). Graph B shows that a one-point increase was associated with a 12% increase in risk (hazard ratio [HR] 1.12; 95% CI 1.02-1.23, P=0.015) for the combined outcome after multivariable correction.

According to the disclosed embodiments, the echocardiographic score can distinguish HFpEF from hypertensive controls and is associated with objective measurements of severity and outcomes in HFpEF.

Neural Network Training

According to a further aspect of the disclosed embodiments, the echo workflow engine 12 incorporates a federated training platform for effectively training the assortment of the neural networks employed by the machine learning layer 200 (FIG. 2 ), as shown in FIG. 17 .

FIG. 17 is a diagram illustrating a federated training platform associated with the automated clinical workflow system. According to one embodiment, the federated training platform 1700 comprises the echo workflow engine 1701 executing on one or more servers 1702 in the cloud. As described above, the echo workflow engine 1701 comprises multiple neural networks (NNs) 1708, which may include some combination of GANs and CNNs, for example. The servers 1702 and echo workflow engine 1701 are in network communication with

remote computers

1703 a, 1703 b, 1703 n (collectively computers 1703) located on premises at respective laboratories 1704 (e.g., lab 1, lab2 . . . , lab N). Each of the laboratories 1704 maintains cardiac biomarker and echo (image) biomarker file archives referred to herein as cardiac and echo

biomarker files

1706 a, 1706 b . . . , 1706N (collectively echo image files 1706) of patient cohorts. For example, the laboratories 1704 may comprise a hospital or clinical lab environment where Echocardiography is performed as a diagnostic aid in cardiology for the morphological and functional assessment of the heart. When performing echocardiography and taking manual measurements, doctors typically select a small subset of the available videos from the echo image files, and may only measure about two of the frames in those videos, which typically may have 70-100 image frames each. In addition, it is believed cardiac biomarkers have yet to be analyzed by machine learning.

To increase the accuracy of the neural networks comprising the echo workflow engine, it would be desirable to make use of the each lab's cardiac and echo biomarker files 1706 as training data for machine learning. However, some or all of the labs 1704 may treat the image file archives as proprietary (graphically illustrated by the firewall), and thus do not allow their cardiac and echo biomarker files 1706 to leave the premises, which means the cardiac and echo biomarker files 1706 are unavailable as a source of training data.

According to another aspect of the disclosed embodiments, the federated training platform 1700 unlocks the proprietary cardiac and echo biomarker files 1706 of the separate laboratories 1704. This is done by downloading and installing lightweight clients and a set of

NNs

1708 a, 1708 b, 1708 c on

computers

1703 a, 1703 b, 1703 c (collectively computers 1703) local to the respective labs 1704. More specifically, lightweight client executing on computer 1703 a of a first lab (Lab 1) accesses the first lab's cardiac and echo biomarker files 1706 a and uses those cardiac and echo biomarker files 1706 a to train the NNs 1708 a and upload a first trained set of NNs back to the server 1702 after training. The first set of trained NNs 1708 are then trained at a second lab (e.g., lab 2) by downloading the lightweight clients and NNs 1708 b the computer 1703 b located at the second lab 2. The lightweight client executing on the computer 1703 b of the second lab can then access the second lab's cardiac and echo biomarker files 1706 b and use those cardiac and echo biomarker files 1706 b to continue to continue to train the NNs 1708 b and to upload a second trained of NNs set back to the server 1702. This process may continue until the NN's complete training at the last lab N by the lightweight client executing on the computer 14N of the last lab N to access lab N's cardiac and echo biomarker files 1706N to train the NNs and to upload a final trained set of neural networks to the server 1702. Once uploaded to the server 1702 the final train set of neural networks are then used in analysis mode to automatically recognize and analyze the cardiac and echo biomarkers in the patient studies of the respective labs 1704.

The federated training platform 1700 results in a highly trained set of NNs 1708 that produce measurements and predictions with a higher degree of accuracy. Another benefit is that federated training platform 1700 unlocks and extracts value from the existing stores of cardiac and echo biomarker data. The cardiac and echo biomarker files 1706 from the laboratories 1704 previously represented vast numbers of cardiac and echo biomarker data from past patient studies and clinical trials that sat unused and unavailable for machine learning purposes because the images are unstructured, views are un-labelled, and most of the images were ignored. Through the federated training platform 1700, these unused and unavailable echo images are now processed by the lightweight client of the echo workflow engine 1701 to create labelled echo images that are stored in structured

image databases

1710 a, 1710 b . . . , 170N, at each of the labs 1704, which is a necessary prerequisite for any machine learning training or machine learning experiments performed on the images. In one embodiment, the structured image databases remain located on premises at the individual labs 1704 to comply with the security requirements (of the labs 1704 and/or the echo workflow engine 1701).

Accordingly, the federated training platform 1700 provides access to structured cardiac and echo biomarker databases 1710 in multiple lab locations to allow distributed neural network training and validation of disease prediction across multiple patient cohorts without either the original cardiac biomarker data in files 1706 or the labelled echo image files in the structured database 1710 ever having to leave the premise of the labs 1704.

AI-Based Workflow Engine to Enhance Mobile Ultrasound Devices

According to yet another aspect of the disclosed embodiments, the echo workflow engine is configured to enhance functioning of a mobile ultrasound device 24 (FIG. 1 ). Commercially available mobile or hand-held ultrasound devices 24 are typically not available with artificial intelligence capabilities or analysis, and simply capture echocardiogram images. In this embodiment of the clinical workflow system, the echo workflow engine may be implemented with or without cardiac biomarker measurements and/or federated training.

FIG. 18 is a diagram illustrating an automated clinical workflow system embodiment where the echo workflow engine is configured to provide a mobile ultrasound device 24 with automatic recognition and measurements of 2D and Doppler modality Echocardiographic images. In this embodiment, the echo workflow engine may run in the standalone configuration in which the echo workflow engine executes within a computer 1814. In the embodiment shown, the computer 1814 is implemented as a tablet computer, but the computer 1814 may comprise any computer form factor. Alternatively, the echo work engine may run in a connected configuration comprising the echo workflow engine 1812A executing on one or more servers 1834 in communication over network (e.g., the Internet) 1826 with an echo workflow client 1812B executing on computer 1814. As used herein, the standalone embodiment for the echo workflow engine and the connected configuration will be referred to collectively as echo workflow engine 8112.

In one embodiment, the computer 1814 is in electronic communication (wirelessly or wired) with peripheral devices such as an ultrasound imaging device, or simply ultrasound device. In one embodiment, the ultrasound device may comprise a mobile ultrasound device 1824, but any form factor may be used. The mobile ultrasound device 1824 includes a display device, a user interface (UI), and a transducer or probe that captures echocardiogram images of a patient's organ (e.g., a heart). As the transducer captures echocardiogram images, the images are displayed in the UI on the display device in real time. In one embodiment, the mobile ultrasound device 1824 may comprise a tablet computer, a laptop or a smartphone. The computer 14 may be located in a hospital or clinical lab environment where Echocardiography is performed as a diagnostic aid in cardiology for the morphological and functional assessment of the heart.

According to one aspect of the disclosed embodiments, the echo workflow engine 1812 enhances the mobile ultrasound device 1824 by providing simultaneous acquisition and AI recognition of the echocardiogram (echo) images. In embodiments, software is either added to the mobile ultrasound device 1824 or existing software of the mobile ultrasound device 1824 is modified to operate with the echo workflow engine 1812. That is the mobile ultrasound device 1824 may be modified to transmit echo images to the echo workflow engine, and receive and display the view classifications and the report, as described in FIG. 19 . In yet another embodiment, the echo workflow engine 1812 may be incorporated into, and executed within, the mobile ultrasound device 1824 itself.

FIG. 19 is a flow diagram for enhancing an ultrasound device to perform AI recognition of echocardiogram images. The process may begin by the echo workflow engine receiving the echo images (which may include videos) that are captured by the mobile ultrasound device 1824 and displayed in a user interface (UI) of the ultrasound device (block 1900).

FIG. 20A is a diagram illustrating acquisition of echo images by the mobile ultrasound device 1824. As described above, during a typical patient echocardiogram exam (referred to as a study), a user (e.g., a sonographer, a technician or even an untrained user), places one or more probes of the ultrasound imaging device 24 against a patient's chest to capture 2D echo images of the heart to help diagnose the particular heart ailment. Measurements of the structure and blood flows are typically made using 2D slices of the heart and the position of the mobile ultrasound device 24 is varied during an echo exam to capture different anatomical sections of the heart from different viewpoints. The technician has the option of adding to these 2D echo images a waveform captured from various possible modalities including: continuous wave Doppler, m-mode, pulsed wave Doppler and pulsed wave tissue Doppler.

In the standalone configuration embodiment, the echo images may be stored as a patient study in database 16 and image file archive 18 (FIG. 1 ) by the computer 1814, the mobile ultrasound device 1824 or both. In the connected configuration embodiment, the echo workflow client 1812B may store the echo images or otherwise transmit the echo images over the network 1826 to the echo workflow engine 1812 on the servers 1843 for storage as the patient study.

Referring again to FIG. 19 , once the echo images are received, the echo workflow engine 1812 processes the echo images using one or more neural networks to automatically classify by view type, i.e., which angle of the heart is captured (block 1902).

The echo workflow engine 1812 then provides even an untrained user real-time feedback by simultaneously displaying the view type of the echo images in the UI of the mobile ultrasound device 1824 along with display of the echo images (block 1904). See for example, view classification blocks 304 and 306 in FIG. 4A.

FIG. 20B is a diagram of a user interface displayed by the mobile ultrasound device 1824. The UI 2000 displays an echo video (e.g., a series of echo images) as normal, but according to one embodiment, the UI 2000 is further configured to display real-time feedback to the user alongside the echo images (shown), or overlaid on the echo images. The echo workflow engine 1812 uses AI to automatically perform echo image analysis to detect and display the current view type 2002, which in the example shown is “PLAX”. After all the required views are captures, the UI may also display an option 2004 for the user to view a report of echo image measurements generated by the echo workflow engine 1812.

Referring again to FIG. 19 , the process continues with the echo workflow engine 1812 generating a report showing calculated measurements of features in the echo images (block 1906). The echo workflow engine 1812 then displays the report showing the calculated measurements on a display device (block 1908). In one embodiment, the workflow client 1812 b may transmit the report showing the calculated measurements to the ultrasound device 824 for display in the UI.

As described above, the calculated measurements may be generated by the neural networks segmenting regions of interest in the 2D images to produce segmented 2D images, segmenting the Doppler modality images to generate waveform traces to produce segmented Doppler modality images, and using both the segmented 2D images and the segmented Doppler modality images to calculate measurements of cardiac features for both left and right sides of the heart.

FIGS. 20C, 20D and 20E are diagrams illustrating example pages of the report showing calculated measurements of features in the echo images. The report may be displayed in the UI 2006 of the mobile ultrasound device 1824. FIG. 20C shows that the report may comprise multiple pages of various types of information. In an embodiment, page 1 of the report may include an ID of the patient, processing and visitation dates, a Main findings section showing of features and conclusions (e.g., normal, abnormal and the like), and a running video of an echo image. In one embodiment, the UI 206 may display the report in a scrollable window that allows the user to scroll down to view various pages of the report.

FIG. 20

D

show page

2 of the report, which may display feature measurements for the left ventricle. In one embodiment, the report may display a UI control next to one or more of the features for the user to select to display additional details. For example, in response to the user clicking on a UI control next to “LVEF MOD biplane”, the report may display information shown in FIG. 20E.

FIG. 20E shows that responsive to the user selecting “LVEF MOD biplane”, a window expands to display “ASE and EACI 2015 Guidelines—Recommendations for Cardiac Chamber Quantification” in addition to a corresponding echo image from the patient. The user may click on “Male” or “Female” to see different value ranges. If the user selects the feature “LVPWd”, the report may display a full screen video, rather than an expandable window.

FIG. 21 is a flow diagram showing processing by the echo workflow engine in the connected configuration comprising the echo workflow client 1812B in network communication with the echo workflow engine 1812A executing on one or more servers 1834. The process may begin by the echo workflow client 1812B receiving a user command to connect to the mobile ultrasound device 1824 (block 2100). In one embodiment, the echo workflow client 1812B may display a list of available devices via Bluetooth for user selection. After a wireless connection is established with the mobile ultrasound device 1824, the echo workflow client 1812B receives a command from the user to record echo images and transmits a record command to the mobile ultrasound device 1824 (block 2102).

Responsive to receiving the record command, the mobile ultrasound device 1824 initiates echo image recording and transmits captured echo images to the echo workflow client 1812B for screen sharing (block 2104). The echo workflow client 1812B receives the echo images from the mobile ultrasound device 1824 and transmits the echo images to the echo workflow engine 1812A (block 2106). In one embodiment, the echo images may be stored and associated with a patient study by either the echo workflow client 1812B or the echo workflow engine 1812A.

In one embodiment, the echo workflow engine 1812A may be distributed across a prediction server and a job server (not shown). The echo workflow client 1812B may be configured to send single echo images to the prediction server and send videos to the job server for view classification.

The echo workflow engine 1812A automatically classifies the echo images by view type and transmits the view types back to the echo workflow client 1812B (block 2108). The echo workflow client 1812B receives the view types and forwards the view types to the mobile ultrasound device 1824 (block 2110). The mobile ultrasound device 1824 receives and displays the view types in a UI along with the captured echo images (block 2112).

The echo workflow client 1812B determines if there are more view types to be captured (block 2014). In one embodiment, the echo workflow client 1812B may utilize a predetermined workflow or a list of view types to capture. Responsive to determining there are more view types to be captured, the echo workflow client 1812B may transmit a prompt to the mobile ultrasound device 1824 for the user to capture the next view type (block 2116). The mobile ultrasound device 1824 receives and displays the prompt in the UI alongside the captured echo images (block 2118).

Responsive to determining there are no more view types to capture (block 2114), the echo workflow client 1812B, transmits a request to the echo workflow engine 1812A for a report showing calculated measurements of features in the echo images (block 2120). The echo workflow engine 1812A performs AI workflow processing of the recorded images (block 2122), and generates and returns the report (block 2124). Responsive to receiving the report, the echo workflow client 1812B or the mobile ultrasound device 1824 may display the report showing the calculated measurements on a display device (block 2126).

In one embodiment, the echo workflow client 1812B displays the report on the display device of the computer 1814. For example the echo workflow client 1812B may display a UI component, such as a “report” button or other control for user selection that causes display of the report. In another embodiment, the echo workflow client 1812B may, instead of or in addition to, transmit the report to the mobile ultrasound device for display in the UI of the mobile ultrasound device 1824.

Accordingly, the advantages of the present embodiment provide a mobile tool with automated, AI based decision support and analysis of both 2D and Doppler echocardiogram images, thereby providing caregivers with a versatile, point-of care solution for triage, diagnosis, prognosis and management of cardiac care.

AI-Based Guidance for an Ultrasound Device to Improve Capture of Echo Image Views

Proper transducer placement and manipulation are required to optimize ultrasound images. The placement and manipulation of the transducer will differ with each patient depending on the patient's physical build and the position of the heart in the chest. A subtle change in probe position and manipulation can significantly impact the quality of the image and automatic recognition of the view type. As described above with respect to FIG. 20A, the UI 2000 of the mobile ultrasound device conventionally displays echo images, but is further configured to display along with the echo images real-time feedback to the user.

According to the present embodiment, the real-time feedback further includes AI-based guidance for the ultrasound device to improve capture of echo image view types. In one embodiment, the AI-based guidance may include continuously attempting AI recognition of the echo images as they are being captured combined with displaying in the UI of the mobile ultrasound device 1824 feedback indications to the user of which directions to move the probe 24′ so the probe 24′ can be placed in a correct position to capture and successfully recognize the echo image.

In embodiments, software is either added to the mobile ultrasound device 1824 or existing software of the mobile ultrasound device 1824 is modified to operate with the echo workflow engine 1812. That is the mobile ultrasound device 1824 may be modified to transmit echo images to the echo workflow engine, and receive and display the view type classifications. In yet another embodiment the echo workflow engine 1812 may be incorporated into, and executed within, the mobile ultrasound device 1824 itself.

FIG. 22A illustrates a flow diagram of an AI-based guidance process for an ultrasound device to improve capture of echo image views. The process may begin by the echo workflow engine 1812 executing on a processor and receiving the echo images that are displayed in the UI of the mobile ultrasound device 1824 (block 2202). The echo workflow engine 1812 processes the echo images using one or more neural networks to continuously attempt to automatically classify the echo images by view type and generates corresponding classification confidence scores (block 2204).

The echo workflow engine 1812 simultaneously displays the view type of the echo images in the UI of the ultrasound device along with the echo images (block 2206). In one embodiment, the confidence scores may also be displayed.

According to the disclose embodiments, the echo workflow engine 1812 displays in the UI of the mobile ultrasound device feedback indications to the user, including which directions to move a probe of the mobile ultrasound device so the probe can be placed in a correct position to capture and successfully classify the echo image (block 2208).

FIG. 22B is a diagram of a user interface displayed by the mobile ultrasound device 1824 and is similar to the diagram shown in FIG. 20B where the UI 2130 displays a series of echo images (e.g., a video) 2132 in the UI 2130 and simultaneously displays real-time feedback indications 2134 in the UI 2130 alongside the echo images, or overlaid on the echo images 2132. According to the present embodiment, the echo workflow engine 1812 uses AI to not only automatically perform echo image analysis to detect and display the current view type, e.g. “PLAX”, but also to detect and display feedback indications 2134 of which directions to move the probe 24′ of the mobile ultrasound device 1824.

In embodiments, the feedback indications 2134 can be displayed as any combination of alphanumeric characters and graphical objects, audio, and any combination thereof. For example, the feedback indications 2134 may comprise text, symbols, icons, emoji's, pre-recorded digital voice instructions, or any combination thereof.

The feedback indications may further be displayed as both informational guidance 2134A, directional guidance instructions 2134B, and a combination thereof as shown. Informational guidance 2134A relates to display of information and may include setup instructions, a list of view types to be captured, the view type currently being captured, confidence scores of the view types, remaining view types to be captured, all view types captured, and any combination thereof. FIG. 22B shows an example where the informational guidance 2134A comprises the current view being acquired is “PLAX, a percent completion of the acquisition, and a list of view types to be acquired.

Directional guidance instructions

2134B relate to instructions that guide the user in moving the probe and may include i) macro-guidance instructions that instruct the user to adjust alignment or position of the probe to a known view type, and ii) micro-guidance instructions that instruct the user to optimize a view type of an echo image in progress, or a combination of both. Both the macro guidance instructions and the micro-guidance instructions may include probe movement instructions 2134B such as: up/down or side-side, diagonal movements (up left/up right, down left/down right) rotation clockwise or counter-clockwise, tilt up/down or side-side, hold, or a combination thereof. FIG. 22B shows an example where the directional guidance instructions 2134B comprises an instruction to “Hold Still, Acquiring Image”. One or more of the directional guidance instructions 2134B may be associated with a distance value, e.g., “Move the probe to the left 1 inch”, or a time value, e.g., “Hold the probe still for 6 seconds”.

FIGS. 22C-22E are diagrams showing a progression of feedback indications 2134 continuing from FIG. 22B. FIG. 22C shows the informational guidance 2134A has been updated to include indications that the current view type being captured is “A4C”, the percent completion of the acquisition, and the list of view types to be acquired, where a checkmark is displayed next to the captured view type PLAX to indicate that view type has been successfully captured and recognized. The directional guidance instructions 2134B have been updated to display to “Please Adjust Probe” along with a probe movement instruction comprising a graphical clockwise arrow.

FIG. 22D shows the informational guidance 2134A has been updated to include indications that the current view type being captured is “A2C”, the percent completion of the acquisition, and the list of view types to be acquired, where a checkmark is displayed next to the captured view type PLAX and A4C to indicate that these view types have been successfully captured and recognized. The directional guidance instructions 2134B have been updated to display “Please Adjust Probe” along with a probe movement instruction comprising a graphical twisting left arrow.

FIG. 22E shows the informational guidance 2134A has been updated to include indications that the all “All View Types Acquired!”, a blank percent completion of the acquisition, and the list of view types acquired with checkmarks. The directional guidance instructions 2134B are no longer needed and not displayed since the view type captures have completed.

Referring again to FIG. 22A further details are shown of the process for displaying the feedback indications to the user in the UI of the mobile ultrasound device (block 2208). This process may begin by the echo workflow engine 1812 determining if the classification confidence scores meet a predetermined threshold (block 2208A). For example in one embodiment, the predetermined threshold may be greater than 60-75% and may be configurable. Responsive to the confidence scores meeting the predetermined threshold, the echo workflow engine 1812 displays the view type and a directional guidance instruction to the user to optimize the view type until successful capture (block 2208B).

Responsive to the confidence scores not meeting the predetermined threshold, the echo workflow engine 1812 displays an estimated view type and a feedback instruction in a form of directional guidance instructions to adjust alignment or position of the probe to a known view type (block 2108C), and this process continues until the classification score meets the threshold.

The echo workflow engine 1812 then determines if there is another view type to capture, e.g., such as defined in a workflow (block 2208D). If so, the process continues with processing of the echo images (block 2204). Otherwise, the echo workflow engine 1812 may optionally display informational feedback that all views have been captured and the process ends (block 2208E).

Although not shown, the process may further include prior to echo image capture, optionally displaying feedback instructions in a form of setup instructions in the UI of the mobile ultrasound device 1824. This may be of aid to a novice user in terms of receiving help how to properly setup the mobile ultrasound device 1824 and initially place the probe 24′.

Accordingly, the present embodiment provides an ultrasound device user with immediate feedback on whether images acquired in a mobile setting are captured in the correct angle and are of sufficient quality to provide suitable measurements and diagnosis of the patient's cardiac condition.

A method and system for implementing a software-based automatic clinical workflow that diagnoses heart disease and AI recognition of both 2D and Doppler modality Echocardiographic images. The present invention has been described in accordance with the embodiments shown, and there could be variations to the embodiments, and any variations would be within the spirit and scope of the present invention. Accordingly, many modifications may be made by one of ordinary skill in the art without departing from the spirit and scope of the appended claims.

Claims

We claim:

1. A computer-implemented method for artificial intelligence (AI)-based guidance for an ultrasound device to improve capture of echo image views, the method comprising:

receiving, by at least one processor, a plurality of the echo images captured by a probe of the ultrasound device, the ultrasound device including a display and a user interface (UI) that displays the echo images to a user, the echo images comprising 2D images and Doppler modality images of a heart;

processing, by one or more neural networks, the echo images to continuously attempt to automatically classify the echo images by view type and generate corresponding classification confidence scores;

simultaneously displaying the view type of the echo images in the UI of the ultrasound device along with the echo images; and

displaying in the UI, feedback indications to the user including which directions to move the probe of the ultrasound device so the probe can be placed in a correct position to capture and successfully classify the echo images, the feedback indications comprising both informational guidance and directional guidance instructions,

the informational guidance comprising setup instructions, a list of view types to be captured, the view type currently being captured, confidence scores of the view types, remaining view types to be captured, all view types captured, or any combination thereof, and

the directional guidance instructions comprising i) macro-guidance instructions that instruct the user to adjust alignment or position of the probe to a known view type, or ii) micro-guidance instructions that instruct the user to optimize the view type of an echo image in progress, wherein the feedback indications comprise any combination of alphanumeric characters and graphical objects, audio, and any combination thereof.

2. The method of claim 1, wherein the macro-guidance instructions and the micro-guidance instructions comprise probe movement instructions including: up/down or side-side, diagonal movements, rotation clockwise or counter-clockwise, tilt up/down or side-side, hold, or a combination thereof.

3. The method of claim 1, further comprising associating the directional guidance instructions with a distance value or a time value.

4. The method of claim 1, wherein displaying in the UI of the ultrasound device feedback indications to the user further comprises responsive to determining that the classification confidence scores meet a predetermined threshold, displaying the view type and a directional guidance instruction to the user to optimize the view type.

5. The method of claim 4, further comprising responsive to determining that the classification confidence scores do not meet the predetermined threshold, displaying an estimated view type and a feedback instruction in a form of directional guidance instructions to adjust alignment or position of the probe to a known view type.

6. The method of claim 1, further comprising displaying the feedback indications in a form of informational setup instructions in the UI of the ultrasound device prior to receiving the echo images.

7. The method of claim 1, further comprising executing by the at least one processor an echo workflow engine, the at least one processor in electronic communication with the ultrasound device.

8. The method of claim 1, further comprising executing by one or more servers an echo workflow engine in communication over a network with a client component executing on a computer, the computer in electronic communication with the ultrasound device.

9. The method of claim 1, further comprising adding software as an echo workflow engine to the ultrasound device.

10. A system, comprising:

a memory storing an artificial intelligence (AI) workflow instructions; and

one or more processors coupled to the memory, wherein when the artificial intelligence workflow instructions are executed cause the one or more processors to:

receive a plurality of echo images captured by a probe of an ultrasound device, the ultrasound device including a display and a user interface (UI) that displays the echo images to a user, the echo images comprising 2D images and Doppler modality images of a heart;

process, by one or more neural networks, the echo images to continuously attempt to automatically classify the echo images by view type and generate corresponding classification confidence scores;

simultaneously display the view type of the echo images in the UI of the ultrasound device along with the echo images; and

display in the UI, feedback indications to the user including which directions to move the probe of the ultrasound device so the probe can be placed in a correct position to capture and successfully classify the echo images, the feedback indications comprising both informational guidance and directional guidance instructions,

directional guidance instructions comprising i) macro-guidance instructions that instruct the user to adjust alignment or position of the probe to a known view type, or ii) micro-guidance instructions that instruct the user to optimize the view type of an echo image in progress; wherein the feedback indications comprise any combination of alphanumeric characters and graphical objects, audio, and any combination thereof.

11. The system of claim 10, wherein the macro-guidance instructions and the micro-guidance instructions comprise probe movement instructions including: up/down or side-side, diagonal movements, rotation clockwise or counter-clockwise, tilt up/down or side-side, hold, or a combination thereof.

12. The system of claim 10, further comprising associating the directional guidance instructions with a distance value or a time value.

13. The system of claim 10, wherein displaying in the UI of the ultrasound device feedback indications to the user further comprises responsive to determining that the classification confidence scores meet a predetermined threshold, displaying the view type and a directional guidance instruction to the user to optimize the view type.

14. The system of claim 13, further comprising responsive to determining that the classification confidence scores do not meet the predetermined threshold, displaying an estimated view type and a feedback instruction in a form of directional guidance instructions to adjust alignment or position of the probe to a known view type.

15. The system of claim 10, further comprising displaying the feedback indications in a form of informational setup instructions in the UI of the ultrasound device prior to receiving the echo images.

16. The system of claim 10, further wherein the AI workflow instructions execute on a computer in electronic communication with the ultrasound device.

17. The system of claim 10, wherein the artificial intelligence workflow instructions execute on one or more servers in communication over a network with a client component executing on a computer, the computer in electronic communication with the ultrasound device.

18. The system of claim 10, wherein the artificial intelligence workflow instructions are added to the ultrasound device.

19. A non-transitory computer-readable medium containing program instructions for implementing artificial intelligence (AI)-based guidance for an ultrasound device to improve capture of echo image views, which when executed by a processor system configure the processor system for:

receiving a plurality of the echo images captured by a probe of the ultrasound device, the ultrasound device including a display and a user interface (UI) that displays the echo images to a user, the echo images comprising 2D images and Doppler modality images of a heart;