Coronary Stenosis Assessment Using Myocardial Perfusion Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for assessing the functional significance of coronary artery stenosis, such as FFR, FFRCT, and CCTA, are invasive, computationally complex, or rely heavily on anatomical vessel geometry segmentation, which is challenging due to imaging artifacts and lack consideration of myocardial microvasculature and collateral flow, leading to potential overestimation or underestimation of stenosis severity.
Innovation Solution
A machine learning-based approach using features extracted from a single CCTA dataset to classify patients with functionally significant stenosis by segmenting myocardium and utilizing texture and morphologic features, without relying on detailed coronary morphology, and incorporating secondary information like coronary tree anatomy and myocardial perfusion characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If FFR (Fractional Flow Reserve) is used to assess functional severity of coronary stenosis, then measurement precision is improved, but device complexity and invasiveness increase due to requiring pressure wire insertion and hyperemia induction
Solution Approach 1:
The patent creates a virtual copy of the FFR measurement process by training a machine learning model on CCTA images and corresponding FFR values. The model learns to predict FFR from anatomical images alone, eliminating the need for invasive pressure wire insertion while replicating the functional assessment capability
Solution Approach 2:
The patent replaces the mechanical invasive measurement system (pressure wire and hyperemia induction) with a computational imaging system. The machine learning model processes standard CCTA images to infer functional severity, substituting physical intervention with algorithmic analysis
2Ease of operation
If FFRCT (computational FFR from CCTA) is used to non-invasively assess stenosis, then ease of operation is improved, but device complexity increases due to computationally intensive CFD simulations requiring detailed vessel segmentation
Solution Approach 1:
The patent uses a pre-trained machine learning model that can be rapidly deployed without requiring complex computational resources during actual use. The model was trained once on a large dataset and can now provide FFR predictions from standard CCTA images without intensive CFD simulations
Solution Approach 2:
The patent changes the input parameters from requiring detailed 3D vessel geometry and boundary conditions for CFD to using standard 2D CCTA image features. The machine learning model extracts relevant features directly from the images, eliminating the need for complex vessel segmentation and geometric reconstruction
3Measurement precision
If detailed coronary artery segmentation is performed to assess stenosis, then measurement precision is improved, but difficulty of detecting and measuring increases due to imaging artifacts and complexity of vessel geometry
Solution Approach 1:
The patent extracts only the essential information needed for FFR prediction directly from CCTA images using machine learning, without performing full vessel segmentation. The model learns to identify relevant features associated with functional severity while ignoring artifacts and irrelevant anatomical details
Solution Approach 2:
The patent introduces a machine learning model as an intermediary between the CCTA images and the FFR assessment. The model acts as a mediator that translates image features into functional severity predictions without requiring explicit vessel segmentation or geometric reconstruction
4Ease of operation
If CCTA is used for anatomical assessment of coronary arteries, then ease of operation is improved, but measurement precision worsens because anatomical imaging alone cannot assess physiological significance of lesions
Solution Approach 1:
The patent merges anatomical information from CCTA with functional assessment capabilities by training the machine learning model on pairs of CCTA images and FFR values. The model learns to integrate structural and functional relationships, providing both anatomical visualization and functional severity prediction from a single non-invasive exam
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods and systems are provided for assessing the presence of functionally significant stenosis in one or more coronary arteries, further known as a severity of vessel obstruction. The methods and systems can implement a prediction phase that comprises segmenting at least a portion of a contrast enhanced volume image data set into data segments corresponding to wall regions of the target organ, and analysing the data segments to extract features that are indicative of an amount of perfusion experiences by wall regions of the target organ. The methods and systems can obtain a feature-perfusion classification (FPC) model derived from a training set of perfused organs, classify the data segments based on the features extracted and based on the FPC model, and provide, as an output, a prediction indicative of a severity of vessel obstruction based on the classification of the features.