Method and system for machine learning based assessment of fractional flow reserve

Patent metadata

Field	Value
Publication number	US-10258244-B2
Application number	US-201815958483-A
Country	US
Kind code	B2
Filing date	Apr 20, 2018
Priority date	Oct 17, 2013
Publication date	Apr 16, 2019
Grant date	Apr 16, 2019

Abstract

A method and system for determining fractional flow reserve (FFR) for a coronary artery stenosis of a patient is disclosed. In one embodiment, medical image data of the patient including the stenosis is received, a set of features for the stenosis is extracted from the medical image data of the patient, and an FFR value for the stenosis is determined based on the extracted set of features using a trained machine-learning based mapping. In another embodiment, a medical image of the patient including the stenosis of interest is received, image patches corresponding to the stenosis of interest and a coronary tree of the patient are detected, an FFR value for the stenosis of interest is determined using a trained deep neural network regressor applied directly to the detected image patches.

First claim

The invention claimed is:

1. A method for determining a rest state hemodynamic index for a stenosis of interest for a patient, comprising:
receiving medical image data of the patient including the stenosis of interest;
extracting a set of features for the stenosis of interest from the medical image data of the patient; and
determining a value of a rest state hemodynamic index for the stenosis of interest based on the extracted set of features using a trained machine-learning based mapping, wherein the trained machine-learning based mapping is trained based on geometric features extracted from synthetically generated stenosis geometries and rest state a rest state hemodynamic index values corresponding to the synthetically generated stenosis geometries computed using computational fluid dynamics (CFD) simulations performed on the synthetically generated stenosis geometries.

2. The method of claim 1 , wherein the trained machine-learning based mapping is an empirical learned model that combines features from the set of features with respective learned weights.

3. The method of claim 1 , wherein the trained machine-learning based mapping is a regression function trained using image-based boosting ridge regression.

4. The method of claim 1 , wherein extracting a set of features for the stenosis of interest from the medical image data of the patient comprises:
extracting a plurality of features characterizing geometry of the stenosis of interest.

5. The method of claim 4 , wherein the features characterizing the geometry of the stenosis of interest include proximal and distal reference diameters, minimal lumen diameter, lesion length.

6. The method of claim 5 , wherein the features characterizing the geometry of the stenosis of interest further include entrance angle, entrance length, exit angle, exit length, percentage of diameter blocked by the stenosis, and percentage of the area blocked by the stenosis.

7. The method of claim 4 , wherein extracting a set of features for the stenosis of interest from the medical image data of the patient further comprises:
extracting one or more features characterizing morphology of the stenosis.

8. The method of claim 7 , wherein the features characterizing morphology of the stenosis comprise binary parameters indicating presence or absence of one or more of calcification, plaque, thrombus, diffuse disease, total or sub-total occlusion, or myocardial bridging.

9. The method of claim 4 , wherein extracting a set of features for the stenosis of interest from the medical image data of the patient further comprises:
extracting one or more features characterizing geometry of a coronary artery branch in which the stenosis of interest is located.

10. The method of claim 4 , wherein extracting a set of features for the stenosis of interest from the medical image data of the patient further comprises:
extracting one or more features characterizing geometry of an entire coronary artery tree of the patient.

11. The method of claim 4 , wherein extracting a set of features for the stenosis of interest from the medical image data of the patient further comprises:
extracting one or more features characterizing coronary anatomy and function.

12. The method of claim 1 , further comprising:
receiving functional measurements of the patient;
extracting one or more features from the functional measurements of the patient; and
including the one or more features extracted from the functional measurements of the patient in the set of features used for determining the value of the rest state hemodynamic index.

13. The method of claim 1 , further comprising:
receiving demographic information of the patient;
extracting one or more features from the demographic information of the patient; and
including the one or more feature extracted from the demographic information of the patient in the set of features used for determining the value of the rest state hemodynamic index.

14. The method of claim 13 , wherein the one or more features extracted from the demographic information comprise binary parameters indicating whether or not the patient has one or more of past history of heart disease, past history of valve dysfunction, or past history of valve repair or replacement.

15. The method of claim 4 , wherein extracting a set of features for the stenosis of interest from the medical image data of the patient further comprises:
extracting one or more features characterizing in-vitro blood tests.

16. The method of claim 1 , further comprising:
receiving a user input modifying one or more features in the set of features, resulting in a modified set of features reflecting a treatment scenario; and
determining the value of the rest state hemodynamic index for the stenosis of interest based on the modified set of features using the trained machine-learning based mapping.

17. The method of claim 1 , wherein the trained machine-learning based mapping is further trained based on rest state hemodynamic index values computed using a mechanistic model to simulate blood flow in a set of training data.

18. The method of claim 1 , wherein the rest state hemodynamic index comprises one or more of pressure-drop, instantaneous wave-free ratio (IFR), or basal stenosis resistance (BSR).

19. An apparatus for determining a rest state hemodynamic index for a stenosis of interest for a patient, comprising:
a processor; and
a memory storing computer executable instructions, which when executed by the processor cause the processor to perform operations comprising:
receiving medical image data of the patient including the stenosis of interest;
extracting a set of features for the stenosis of interest from the medical image data of the patient; and
determining a value of a rest state hemodynamic index for the stenosis of interest based on the extracted set of features using a trained machine-learning based mapping, wherein the trained machine-learning based mapping is trained based on geometric features extracted from synthetically generated stenosis geometries and rest state hemodynamic index values corresponding to the synthetically generated stenosis geometries computed using computational fluid dynamics (CFD) simulations performed on the synthetically generated stenosis geometries.

20. The apparatus of claim 19 , wherein the trained machine-learning based mapping is a regression function trained using image-based boosting ridge regression.

21. The apparatus of claim 19 , wherein extracting a set of features for the stenosis of interest from the medical image data of the patient comprises:
extracting a plurality of features characterizing geometry of the stenosis of interest.

22. The apparatus of claim 21 , wherein extracting a set of features for the stenosis of interest from the medical image data of the patient further comprises:
extracting one or more features characterizing morphology of the stenosis.

23. The apparatus of claim 21 , wherein extracting a set of features for the stenosis of interest from the medical image data of the patient further comprises:
extracting one or more features characterizing geometry of a coronary artery branch in which the stenosis of interest is located; and
extracting one or more features characterizing geometry of an entire coronary artery tree of the patient.

24. The apparatus of claim 19 , wherein the rest state hemodynamic index comprises one or more of pressure-drop, instantaneou

Assignees

• Siemens Healthcare Gmbh

Inventors

• Sharma Puneet
• Kamen Ali
• Georgescu Bogdan
• Sauer Frank
• Comaniciu Dorin
• Zheng Yefeng
• Nguyen Hien
• Singh Vivek Kumar

Classifications

• G06V10/774 Primary

  Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• A61B5/026 Primary

  Measuring blood flow {(A61B3/1233, A61B3/1241 take precedence)} · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• A61B6/5217

  extracting a diagnostic or physiological parameter from medical diagnostic data · CPC title

• G16H50/30

  for calculating health indices; for individual health risk assessment · CPC title