Angiography Side Branch Segmentation for Accurate FFR Image Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for segmenting side branch vessels in angiography images are limited, leading to inaccurate and time-consuming results, which affects the accuracy of fractional flow reserve (FFR) calculation and the fusion of endovascular and coronary angiography images.
Innovation Solution
A training method for angiography image processing that utilizes a neural network to perform local segmentation of side branch vessels by training with specific training data, including original angiography image data and local segmentation result data, focusing on a defined range from the head end to a segmentation terminal of the side branch vessels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional fusion technology of endovascular images and coronary angiography is used, then image matching can be achieved, but the operation process must be changed and higher operation level is required
Solution Approach 1:
The system performs automatic side branch vessel segmentation using neural networks and algorithms, enabling the system to serve itself without requiring clinician intervention for manual segmentation, thus maintaining high matching accuracy while simplifying operation
Solution Approach 2:
Manual mechanical segmentation operations by clinicians are replaced with automated computational algorithms including neural networks, level set algorithms, and graph cut algorithms, substituting human manual work with automated systems
2Manufacturing precision
If manual segmentation of side branch vessels is performed, then segmentation can be achieved, but it is time-consuming and labor-intensive
Solution Approach 1:
The system performs preliminary automatic segmentation of side branch vessels using neural networks and algorithms before any manual review, pre-processing the images to identify vessel structures, which reduces the time required for subsequent analysis while maintaining accuracy
Solution Approach 2:
Manual segmentation operations are replaced with automated neural network-based segmentation systems that process images computationally, eliminating the need for manual tracing and measurement while maintaining or improving segmentation precision
3Loss of information
If full vessel segmentation is performed, then complete vessel tree is obtained, but it increases processing complexity and time
Solution Approach 1:
The system extracts only the specific side branch vessel information needed for FFR calculation and image matching, rather than processing the entire vessel tree, by isolating and segmenting only the relevant side branch vessels from the coronary artery tree
Solution Approach 2:
The processing is segmented into focused tasks: detecting bifurcation points, identifying side branch vessels, and segmenting only those specific vessels, rather than processing the complete vessel tree uniformly, which reduces overall processing complexity
4Measurement precision
If traditional FFR measurement using pressure wire is used, then accurate FFR can be obtained, but it is time-consuming, high cost and increases vascular injury risk
Solution Approach 1:
The invasive mechanical pressure wire measurement is replaced with a non-invasive computational method that uses angiography images and neural network-based vessel segmentation to calculate FFR, eliminating the need for physical wire insertion while maintaining measurement accuracy
Solution Approach 2:
The system creates a computational model and virtual representation of the vessel geometry from angiography images, using this digital copy to perform FFR calculations instead of requiring physical measurement with a pressure wire, thus avoiding vascular injury
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A training method and apparatus for angiography image processing and a method and apparatus for automatically processing a vessel image. The training method for angiography image processing includes: obtaining training data (S101); wherein the training data includes original angiography image data and local segmentation result data of a side branch vessel, and the local segmentation result data of the side branch vessel are local segmentation image data of the side branch vessel on a main branch vessel determined from an original angiography image. A neural network is trained according to the obtained training data to make the neural network perform local segmentation on the side branch vessel on the determined main branch vessel in the original angiography image (S102). The training method for angiography image processing of the present invention can obtain the neural network for performing local segmentation on the side branch vessel, thereby realizing improvement of segmentation accuracy while improving segmentation efficiency, and avoiding missing segmentation and wrong segmentation of the side branch vessel.