# US 10,483,006 B2 · Learning based methods for personalized assessment, long-term prediction and management of atherosclerosis

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 10,483,006 B2
- Document ID: US10483006B2
- Status: Granted
- Kind: B2
- Application number: 15/599,505
- Filed: 2017-05-19
- Published: 2018-11-22
- Granted: 2019-11-19
- Earliest claimed: 2017-05-19
- Assignees: Siemens Healthcare GmbH
- Applicants: Siemens Healthcare GmbH
- Inventors: Lucian Mihai Itu; Tiziano Passerini; Saikiran Rapaka; Puneet Sharma; Dorin Comaniciu
- Web page: https://patentgrep.com/patent/US10483006B2

## CPC

- G16B 25/10: Gene or protein expression profiling; Expression-ratio estimation or normalisation
- G06F 18/24: Classification techniques
- G06V 10/764: using classification, e.g. of video objects
- G16B 25/00: ICT specially adapted for hybridisation; ICT specially adapted for gene or protein expression
- G16H 50/70: for mining of medical data, e.g. analysing previous cases of other patients
- G01N 2800/323: Arteriosclerosis, Stenosis
- G06V 2201/03: Recognition of patterns in medical or anatomical images
- G16B: BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G01N: INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES (measuring or testing apparatus or processes other than immunoassay, involving enzymes or microorganisms C12M, C12Q)

## Abstract

A computer-implemented method for providing a personalized evaluation of assessment of atherosclerotic plaques for a patient acquiring patient data comprising non-invasive patient data, medical images of the patient, and blood biomarkers. Features of interest are extracted from the patient data and one or more machine learning models are applied to the features of interest to predict one or more measures of interest related to atherosclerotic plaque.

## Claims

### Claim 1

A computer-implemented method for providing a personalized evaluation of assessment of atherosclerotic plaques for a patient, the method comprising:
acquiring patient data comprising non-invasive patient data, medical images of the patient, and blood biomarkers;
extracting features of interest from the patient data;
training one or more machine learning models using a database of synthetic data comprising one or more of in silico anatomical models and in vitro anatomical models; and
applying the one or more machine learning models to the features of interest to predict a plurality of measures of interest related to atherosclerotic plaque, wherein the plurality of measures of interest related to atherosclerotic plaque include a risk of cardiovascular event, plaque composition, plaque evolution, effect of a drug treatment, in-stent restenosis, lesions requiring sealing, indication of a future screening data, and effect of a device for therapy, wherein the risk of cardiovascular event includes coronary circulation, cerebral circulation, and peripheral circulation,
wherein the one or more machine learning models are trained using a process comprising:
performing fluid solid growth (FSG) computations for the in silico anatomical models or flow experiments for the in vitro anatomical models to yield output data;
extracting measures of interest from the output data;
extracting geometric features and plaque-related features from the database of synthetic data; and
training the one or more machine learning models to predict measures of interest related to atherosclerotic plaque using the measures of interest from the output data, the geometric features, and the plaque-related features.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
generating a report in a structured format describing the measures of interest related to atherosclerotic plaque; and
storing the report in a patient-specific medical record.

### Claim 3 (depends on 1)

The method of claim 1, wherein the in silico anatomical models are generated by:
initializing a new in silico anatomical model skeleton of a coronary arterial tree anatomical model by prescribing a number of vessels at each generation of the coronary arterial tree;
defining healthy geometric information for each generation of the coronary arterial tree, wherein the healthy geometric information comprises one or more of vessel radius, a degree of tapering, and a branch length;
establishing one or more stenoses in the coronary arterial tree which modify the healthy geometric information; and
establishing plaque composition for each stenosis in the coronary arterial tree;
updating the new in silico anatomical model based on the modified healthy geometric information and the plaque composition for each stenosis in the coronary arterial tree; and storing the new in silico anatomical model in the database of synthetic data.

### Claim 4 (depends on 3)

The method of claim 3, wherein the plaque composition for each stenosis is established by randomly selecting a particular plaque composition from a plurality of predefined plaque composition types.

### Claim 5 (depends on 4)

The method of claim 4, wherein the plaque composition for each stenosis comprises a plaque material established by:
randomly selecting a center of each volume of plaque material;
randomly selecting a size and a shape for the plaque material; and
randomly selecting material proprieties for the plaque composition; and
establishing the plaque material using the center, size, shape, and material properties.

### Claim 6 (depends on 4)

The method of claim 4, wherein the plaque composition for each stenosis mimics a predefined high risk plaque composition.

### Claim 7 (depends on 1)

The method of claim 1, wherein the one or more machine learning models comprise:
a first machine model trained to predict plaque formation,
a second machine model trained to predict plaque development, and
a third machine model trained to predict plaque rupture.

### Claim 8 (depends on 1)

The method of claim 1, wherein the machine learning models are applied in parallel to the features of interest.

### Claim 9 (depends on 1)

The method of claim 1, wherein the one or more machine learning models comprise:
a first machine learning model trained to predict an ischemic weight of each branch;
a second machine learning model trained to predict ischemic contribution scores,
a third machine learning model trained to predict hemodynamic measures of interest,
a fourth machine learning model trained to predict plaque related measures of interest, and
a fifth machine learning model trained to predict a risk of future cardiovascular event.

### Claim 10 (depends on 9)

The method of claim 9, wherein the machine learning models are applied in a cascaded workflow that sequentially applies the machine learning models using outputs of each machine learning model as inputs for a subsequent machine learning model in the cascaded workflow.

### Claim 11 (depends on 1)

The method of claim 1, further comprising:
generating a visualization of the measures of interest related to atherosclerotic plaque.

### Claim 12 (depends on 11)

The method of claim 11, wherein the visualization comprises a coronary artery image and the method further comprises:
receiving a user selection of a location within the coronary artery image; and
in response to the user selection, presenting a measure of interest corresponding to the location in the visualization.

### Claim 13

A computer-implemented method for training a machine learning model to provide a personalized evaluation of assessment of atherosclerotic plaques for a patient, the method comprising:
generating a database of synthetic data comprising one or more of in silico anatomical models;
performing fluid solid growth (FSG) computations for the in silico anatomical models to yield output data;
extracting measures of interest from the output data, wherein the measures of interest include a risk of cardiovascular event, plaque composition, plaque evolution, effect of a drug treatment, in-stent restenosis, and lesions requiring sealing;
extracting geometric features of the in silico anatomical models and plaque-related features from the database of synthetic data, wherein the plaque-related features include features describing likelihood of plaque development at a particular location, features describing growth speed of the plaque, wherein the plaque-related features are defined based on features describing likelihood of a rupture of the plaque and features describing likelihood of thrombus formation of a plaque surface; and
training one or more machine learning models to generate predicted measurements related to atherosclerotic plaque based on the geometric features of the in silico anatomical models trees and the plaque-related features.

### Claim 14 (depends on 13)

The method of claim 13, further comprising:
applying the one or more machine learning models to the features of interest to generate predicted measurements related to atherosclerotic plaque.

### Claim 15 (depends on 14)

The method of claim 14, further comprising:
receiving patient-specific measurements related to atherosclerotic plaque;
comparing the patient-specific measurements related to atherosclerotic plaque to the predicted measurements related to atherosclerotic plaque;
if the patient-specific measurements do not match the predicted measurements, retaining the one or more machine learning models using the patient-specific measurements.

### Claim 16 (depends on 13)

The method of claim 13, further comprising:
receiving a specification of input uncertainties associated with each of the one or more of in silico anatomical models;
determining a confidence interval for each of the measures of interest;
using the confidence interval and the in silico anatomical models to train the one or more machine learning models to predict confidence intervals of the predicted measurements related to atherosclerotic plaque based on geometric features of anatomical models.