Granted Patent B2

US 12,614,616 · App. 17/492,163 · Granted Apr 28, 2026

Inventors: Colin Compas (San Jose, CA); Yaniv Gur (San Jose, CA); Mehdi Moradi (San Jose, CA); Mohammadreza Negahdar (San Jose, CA); Tanveer Syeda-Mahmood (San Jose, CA)

Assignee: International Business Machines Corporation

G16H10/60G06F18/24143G06Q10/0633G06V10/82G06V30/19173G16H30/20G16H50/20G06F2218/10G06N3/045G06N3/08G06N5/01G06N20/10G06N20/20G06V20/635G06V30/10G06V2201/03

Abstract

Claims (48)

1 . A method comprising:

reading a plurality of frames of an ultrasound video, wherein each frame of the ultrasound video comprises a Doppler pattern;

providing at least one of the plurality of frames to a trained feature generator comprising a convolutional neural network, the convolutional neural network trained using images from an imaging domain other than Doppler echocardiography, the convolutional neural network comprising a plurality of layers, the plurality of layers comprising a first fully-connected layer and an output layer comprising a plurality of output nodes;

obtaining from the trained feature generator at least one feature vector corresponding to the at least one of the plurality of frames, wherein the at least one feature vector is obtained from the first fully-connected layer of the convolutional neural network, the at least one feature vector representing features applicable in Doppler echocardiography and the imaging domain other than Doppler echocardiography;

determining an ultrasound mode label indicative of an ultrasound mode for each of the plurality of frames;

selecting one of a plurality of valve detection stages according to the ultrasound mode label, the selected valve detection stage comprising a trained classifier;

providing the at least one feature vector to the trained classifier;

obtaining from the trained classifier a valve label indicative of a valve represented in the at least one of the plurality of frames;

based on the obtained valve label, determining whether aortic stenosis is present; and

based on the determination, treating the aortic stenosis with one or more transcatheter valve replacements.

2 . The method of claim 1 , wherein the at least one feature vector comprises generic features applicable to multiple imaging domains including Doppler echocardiography images.

3 . The method of claim 1 , wherein the plurality of layers further comprises five convolutional layers, two fully-connected layers, and a Softmax layer.

4 . The method of claim 1 , wherein the at least one feature vector has a size of 4096.

5 . The method of claim 1 , wherein providing the at least one of the plurality of frames to the trained feature generator comprises providing only a region of each frame depicting Doppler patterns.

6 . The method of claim 1 , further comprising extracting one or more measurement derived from the Doppler pattern from the at least one of the plurality of frames, wherein determining whether a disease condition is present is further based on the one or more measurement.

7 . The method of claim 6 , wherein the one or more measurement comprises maximum jet velocity or mean pressure gradient.

8 . The method of claim 1 , wherein the trained classifier is a support vector machine.

9 . The method of claim 1 , wherein the valve label corresponds to the aortic valve, tricuspid valve, pulmonary valve, or mitral valve.

10 . The method of claim 1 , wherein the mode label corresponds to B-Mode, M-Mode, CW-Doppler, PW-Doppler, Text-Panel, Color-Doppler, or Color-Doppler M-Mode.

11 . The method of claim 1 , wherein determining the mode label comprises:

providing the plurality of frames to a second trained feature generator;

obtaining from the second trained feature generator a plurality of feature vectors corresponding to the plurality of frames;

providing the plurality of feature vectors to a second trained classifier; and

obtaining from the second trained classifier the mode label.

12 . The method of claim 11 , wherein the second trained feature generator comprises a convolutional neural network.

13 . The method of claim 11 , wherein the second trained classifier is a support vector machine.

14 . A computer program product for generating feature vector representations of Doppler echocardiography frames, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform a method comprising:

reading a plurality of frames of a medical video, wherein each frame of the medical video comprises a Doppler pattern;

providing at least one of the plurality of frames to a trained feature generator comprising a convolutional neural network, the convolutional neural network trained using images from an imaging domain other than Doppler echocardiography, the convolutional neural network comprising a plurality of layers, the plurality of layers comprising a first fully-connected layer and a output layer comprising a plurality of output nodes;

obtaining from the trained feature generator at least one feature vector corresponding to the at least one of the plurality of frames, wherein the at least one feature vector is obtained from the first fully-connected layer of the convolutional neural network, the at least one feature vector representing features applicable in Doppler echocardiography and the imaging domain other than Doppler echocardiography;

determining an ultrasound mode label indicative of an ultrasound mode for each of the plurality of frames;

selecting one of a plurality of valve detection stages according to the ultrasound mode label, the selected valve detection stage comprising a trained classifier;

providing the at least one feature vector to the trained classifier;

obtaining from the trained classifier a valve label indicative of a valve represented in the at least one of the plurality of frames;

based on the obtained valve label, determining whether aortic stenosis is present; and

based on the determination, treating the aortic stenosis with one or more transcatheter aortic valve replacements.

15 . A system comprising:

a medical imaging records database;

a computing node comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform a method comprising:

reading a plurality of frames of a medical video from the medical imaging records database, wherein each frame of the medical video comprises a Doppler pattern;

providing at least one of the plurality of frames to a trained feature generator comprising a convolutional neural network, the convolutional neural network trained using images from an imaging domain other than Doppler echocardiography, the convolutional neural network comprising a plurality of layers, the plurality of layers comprising a first fully-connected layer and a output layer comprising a plurality of output nodes;

obtaining from the trained feature generator at least one feature vector corresponding to the at least one of the plurality of frames, wherein the at least one feature vector is obtained from the first fully-connected layer of the convolutional neural network, the at least one feature vector representing features applicable in Doppler echocardiography and the imaging domain other than Doppler echocardiography;

determining an ultrasound mode label indicative of an ultrasound mode for each of the plurality of frames;

selecting one of a plurality of valve detection stages according to the ultrasound mode label, the selected valve detection stage comprising a trained classifier;

providing the at least one feature vector to the trained classifier;

obtaining from the trained classifier a valve label indicative of a valve represented in the at least one of the plurality of frames;

based on the obtained valve label, determining whether aortic stenosis is present; and

based on the determination, treating the aortic stenosis with one or more transcatheter aortic valve replacements.

Assignments (1)

Continuity (2)

References Cited (57)