Eko.ai Pte. Ltd.

Agency for Science, Technology and Research

National University of Singapore

National University Hospital

Su Ping Carolyn Lam

James Otis Hare, Ii

Heng-phon Too

Yoran Hummel

Mathias Iversen

Lihan Zhou

Ruiyang Zou

Paul James Seekings

Jasper Tromp

Zhubo Jiang

Wouter Ouwerkerk

Wouter Ouwekerk

Andrie Ochtman

Matthew Frost

Sze Chi Lim

Weile Wayne Tee

Cyril Equilbec

Lee Lee Wong

Arthur Mark Richards

A computer-implemented method for automated diagnosis of cardiac amyloidosis (CA) and hypertrophic cardiomyopathy (HCM) performed by an automated workflow engine executed by at least one processor includes separating a plurality of echocardiogram (echo) images a heart according to 2D images and Doppler modality images. The 2D images are classified by view type, including A4C video. The 2D images are segmented to produce segmented A4C images having a segmentation mask over the left ventricle. Phase detection is performed on the segmented A4C images to determine systole and diastole endpoints per cardiac cycle. Disease classification is performed on beat-to-beat A4C images for respective cardiac cycles. The cardiac cycle probability scores generated for all of the cardiac cycles are aggregated for each A4C video, and the aggregated probability scores for all the A4C videos are combined to generate a patient-level conclusion for CA and HCM.

Automatic clinical workflow that recognizes and analyzes 2D and doppler modality echocardiogram images for automated cardiac measurements and diagnosis of cardiac amyloidosis and hypertrophic cardiomyopathy

A computer-implemented method for grading of Aortic Stenosis severity performed by an automated workflow engine executed by at least one processor includes receiving, from a memory, a plurality of echocardiogram images a heart. The plurality of echocardiogram (echo) images according to 2D images and Doppler modality images. The 2D images are classified by view type, and the Doppler modality images are classified by region. The regions of interest in the 2D images are segmented to produce segmented 2D images. The Doppler modality images are segmented to generate waveform traces to produce segmented Doppler modality images. The segmented images are used to calculate measurements of cardiac features of the heart. A conclusion of Aortic Stenosis severity is generated by comparing the calculated measurements to cardiac guidelines. A report is then output showing the calculated measurements of the cardiac features that fall within or outside of the cardiac guidelines.

Automatic clinical workflow that recognizes and analyzes 2D and doppler modality echocardiogram images for automated cardiac measurements and grading of aortic stenosis severity

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.

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

Artificial intelligence (AI) recognition of echocardiogram (echo) images by a mobile ultrasound device comprises 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 automatically classify the echo images by view type. The view type of the echo images is simultaneously displayed in the UI of the ultrasound device along with the echo images. A report is generated showing the calculated measurements of features in the echo images. The report showing the calculated measurements is displayed on a display device.

Artificial intelligence (AI) recognition of echocardiogram images to enhance a mobile ultrasound device

An automated workflow receives a patient study comprising cardiac biomarker measurements and a plurality of echocardiographic images taken by an ultrasound device of a patient heart. A filter separates the plurality of echocardiogram images by 2D images and Doppler modality images based on analyzing image metadata. The 2D images are classified by view type, and the Doppler modality images are classified by view type. The cardiac chambers are segmented in the 2D images, and the Doppler modality images are segmented to generate waveform traces, producing segmented 2D images and segmented Doppler modality images. Using both the sets of images, measurements of cardiac features for both left and right sides of the heart are calculated. The cardiac biomarker measurements and the calculated measurements are compared with international cardiac guidelines to generate conclusions, and a report is output showing the measurements that fall within or outside of the guidelines.

Clinical workflow to diagnose heart disease based on cardiac biomarker measurements and AI recognition of 2D and doppler modality echocardiogram images

A method for training neural networks of an automated workflow performed by a software component executing on a server in communication with remote computers at respective laboratories includes downloading and installing a client and a set of neural networks to a first remote computer of a first laboratory, the client accessing the echocardiogram image files of the first laboratory to train the set of neural networks and to upload a first trained set of neural networks to the server. The process continues until the client and the second trained set of neural networks is downloaded and installed to a last remote computer of a last laboratory, the client accessing the echocardiogram image files of the last laboratory to continue to train the second trained set of neural networks and to upload a final trained set of neural networks to the server.

Training neural networks of an automatic clinical workflow that recognizes and analyzes 2D and doppler modality echocardiogram images

An automated workflow performed by software executing on at least one processor includes receiving a plurality of echocardiogram images taken by an ultrasound device. A filter separates the plurality of echocardiogram images by 2D images and Doppler modality images based on analyzing image metadata. The 2D images are classified by view type, and the Doppler modality images are classified by view type. The cardiac chambers are segmented in the 2D images, and the Doppler modality images are segmented to generate waveform traces, producing segmented 2D images and segmented Doppler modality images. Using both the sets of images, measurements of cardiac features for both left and right sides of the heart are calculated. The calculated measurements are compared with international cardiac guidelines to generate conclusions and a report is output showing the calculated measurements that fall both within and outside of the guidelines.

Automatic clinical workflow that recognizes and analyzes 2D and doppler modality echocardiogram images for automated cardiac measurements and the diagnosis, prediction and prognosis of heart disease

An automated workflow performed by software executing on at least one processor includes receiving a plurality of echocardiogram images taken by an ultrasound device. A first filter separates the plurality of echocardiogram images by 2D images and Doppler modality images based on analyzing image metadata. A first neural network classifies the 2D images by view type, and a second neural network classifies the Doppler modality images by view type. A third neural network segments cardiac chambers in the 2D images and a fourth neural network segments the Doppler modality images to generate waveform traces, producing segmented 2D images and segmented Doppler modality images. Using both the sets of images, measurements of cardiac features for both left and right sides of the heart are calculated. The calculated measurements are compared with international cardiac guidelines to generate conclusions and a report is output highlighting the measurements that fall outside of the guidelines.

Growing community of inventors

Singapore, Singapore

Su Ping Carolyn Lam