Echocardiographic image analysis

Patent metadata

Field	Value
Publication number	US-11129591-B2
Application number	US-201716095601-A
Country	US
Kind code	B2
Filing date	Apr 21, 2017
Priority date	Apr 21, 2016
Publication date	Sep 28, 2021
Grant date	Sep 28, 2021

Abstract

A computer-implemented system for facilitating echocardiographic image analysis is disclosed. The system includes at least one processor configured to receive signals representing a first at least one echocardiographic image, associate the image with a first view category of a plurality of predetermined view categories, determine, based on the first at least one echocardiographic image and the first view category, a first quality assessment value representing a view category specific quality assessment of the first at least one echocardiographic image, and produce signals representing the first quality assessment value for causing the first quality assessment value to be associated with the first at least one echocardiographic image. The at least one processor may also be configured to do the above steps for a second at least one echocardiographic and a second view category that is different from the first view category image. Other systems, methods, and computer-readable media are also disclosed.

First claim

The invention claimed is:

1. A computer-implemented system for facilitating echocardiographic image analysis, the system comprising at least one processor configured to:
receive signals representing a first at least one echocardiographic image;
associate the first at least one echocardiographic image with a first view category of a plurality of predetermined echocardiographic image view categories;
determine, based on the first at least one echocardiographic image and the first view category, a first quality assessment value representing a view category specific quality assessment of the first at least one echocardiographic image;
produce signals representing the first quality assessment value for causing the first quality assessment value to be associated with the first at least one echocardiographic image;
receive signals representing a second at least one echocardiographic image;
associate the second at least one echocardiographic image with a second view category of the plurality of predetermined echocardiographic image view categories, said second view category being different from the first view category;
determine, based on the second at least one echocardiographic image and the second view category, a second quality assessment value representing a view category specific quality assessment of the second at least one echocardiographic image; and
produce signals representing the second quality assessment value for causing the second quality assessment value to be associated with the second at least one echocardiographic image;
wherein each of the plurality of predetermined echocardiographic image view categories is associated with a respective set of assessment parameters, each of the sets of assessment parameters being a set of neural network parameters that define a neural network having a plurality of layers including an input layer configured to receive one or more echocardiographic images and an output layer configured to output one or more quality assessment values, and wherein the at least one processor is configured to determine the first quality assessment value by:
determining that a first set of assessment parameters of the sets of assessment parameters is associated with the first view category; and
in response to determining that the first set of assessment parameters is associated with the first view category, inputting the first at least one echocardiographic image into the neural network defined by the first set of assessment parameters; and

wherein the at least one processor is configured to determine the second quality assessment value by:
determining that a second set of assessment parameters of the sets of assessment parameters is associated with the second view category; and
in response to determining that the second set of assessment parameters is associated with the second view category, inputting the second at least one echocardiographic image into the neural network defined by the second set of assessment parameters.

2. The system of claim 1 wherein the first quality assessment value represents an assessment of suitability of the first at least one echocardiographic image for quantified clinical measurement of anatomical features and wherein the second quality assessment value represents an assessment of suitability of the second at least one echocardiographic image for quantified measurement of anatomical features.

3. The system of claim 1 wherein the at least one processor is configured to:
produce signals for causing a representation of the first quality assessment value to be transmitted to at least one display for causing the at least one display to display the first quality assessment value in association with the first at least one echocardiographic image, to assist one or more operators of an echocardiographic device in capturing at least one subsequent echocardiographic image; and
produce signals for causing a representation of the second quality assessment value to be transmitted to the at least one display for causing the at least one display to display the second quality assessment value in association with the second at least one echocardiographic image, to assist the one or more operators in capturing at least one subsequent echocardiographic image.

4. The system of claim 1 wherein the at least one processor is configured to:
apply one or more view categorization functions to the first at least one echocardiographic image to determine that the first at least one echocardiographic image falls within the first view category; and
apply one or more view categorization functions to the second at least one echocardiographic image to determine that the second at least one echocardiographic image falls within the second view category.

5. The system of claim 1 wherein the first at least one echocardiographic image comprises a plurality of echocardiographic images and wherein the at least one processor is configured to determine the first quality assessment value by determining a single quality assessment value representing a view category specific assessment of the plurality of echocardiographic images.

6. The system of claim 1 wherein each of the sets of assessment parameters includes:
a set of common assessment parameters, which are common to each of the sets of assessment parameters; and
a set of view category specific assessment parameters, which are unique to the set of assessment parameters.

7. The system of claim 1 wherein the at least one processor is configured to train the neural networks by:
receiving signals representing a plurality of echocardiographic training images, each of the plurality of echocardiographic training images associated with one of the plurality of predetermined echocardiographic image view categories;
receiving signals representing respective expert quality assessment values representing view category specific quality assessments of the plurality of echocardiographic training images, each of the expert quality assessment values provided by an expert echocardiographer and associated with one of the plurality of echocardiographic training images; and
training the neural networks using the plurality of echocardiographic training images as inputs and the associated expert quality assessment values as desired outputs to determine the sets of neural network parameters defining the neural networks.

8. The system of claim 7 wherein each of the expert quality assessment values represents an assessment of suitability of the associated echocardiographic image for quantified clinical measurement of anatomical features.

9. The system of claim 7 wherein the at least one processor is configured to derive each of the expert quality assessment values at least in part from a clinical plane assessment value representing an expert opinion whether the associated echocardiographic training image was taken in an anatomical plane suitable for quantified clinical measurement of anatomical features.

10. The system of claim 7 wherein each of the sets of neural network parameters includes:
a set of common neural network parameters, which are common to each of the sets of neural network parameters; and
a set of view category specific neural network parameters, which are unique to the set of neural network parameters; and

wherein the at least one processor is configured to, for each echocardiographic training image:
select one of the sets of view category specific neural network parameters based on the predetermined echocardiographic image view category associated with the echocardiographic training image; and
using the echocardiographic training image as an input and the associated expert quality assessment values as a desired output,

Assignees

• Univ British Columbia

Inventors

• Abolmaesumi Purang
• Rohling Robert
• Abdi Amir H
• Tsang Teresa S M

Classifications

• G06N3/08

  Learning methods · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06T2207/20081

  Training; Learning · CPC title