Coronary OCT Tissue Classification Using Dilated AI Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing OCT imaging techniques for coronary arteries require significant time and expertise from qualified physicians to analyze atherosclerotic plaque tissues, and there is a need for automated systems that can efficiently extract and classify pathological tissues in OCT images.
Innovation Solution
A system using a trained fully convolutional engine with dilated convolutional layers and a sparse auto-encoder classification engine to automatically extract and classify atherosclerotic pathological tissues in coronary artery OCT images, allowing for real-time analysis without pre-processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple OCT images are acquired to improve diagnostic accuracy, then measurement precision is improved, but loss of time increases due to significant analysis time required
Solution Approach 1:
The patent replaces the manual mechanical analysis process performed by physicians with an automated computer-implemented engine using artificial intelligence. The system automatically extracts pathological tissues and classifies them into atherosclerotic types, substituting human expert analysis with an automated computational system that processes multiple OCT images rapidly without sacrificing diagnostic accuracy.
Solution Approach 2:
The system enables self-service by allowing the OCT images to be automatically processed and analyzed without requiring significant human intervention. The trained engine autonomously performs tissue extraction and classification, making the system serve itself in terms of image analysis while minimizing the need for highly qualified physicians to spend extensive time examining images.
2Manufacturing precision
If complex pre-processing is applied to improve tissue extraction accuracy, then manufacturing precision is improved, but device complexity increases
Solution Approach 1:
The patent changes the parameters of the convolutional engine by using dilated convolutional layers with specific dilation rates different than unity. This parameter modification allows the engine to extract pathological tissues effectively from raw OCT images without requiring complex pre-processing steps, achieving accurate tissue extraction while maintaining relatively simple device architecture.
Solution Approach 2:
The system performs preliminary action by training the fully convolutional engine and auto-encoder classification engine in advance using extensive datasets. This pre-training allows the system to achieve high tissue extraction accuracy when deployed, eliminating the need for complex pre-processing during actual operation since the trained engine can directly process raw images effectively.
3Reliability
If traditional classification methods are used to ensure reliable tissue type determination, then reliability is improved, but productivity decreases due to time-consuming analysis
Solution Approach 1:
The patent segments the diagnostic process into two distinct functional components: a fully convolutional engine for pathological tissue extraction and an auto-encoder classification engine for tissue type determination. This segmentation allows each component to be optimized independently, with the classification engine trained specifically for reliable tissue type determination while the extraction engine handles image processing, together achieving both reliability and improved productivity through parallelized automated processing.
Solution Approach 2:
The auto-encoder classification engine serves as an intermediary between the raw OCT images and the final diagnostic classification. It receives extracted pathological tissues from the fully convolutional engine and transforms them into classified tissue types, acting as a mediator that ensures reliable determination while enabling rapid automated processing that increases diagnostic throughput compared to traditional direct analysis methods.
Data Source
AI summary
There is described a system for determining an atherosclerotic pathological tissue type of a coronary artery, the system comprising: an optical coherence tomography (OCT) imaging system being configured for acquiring an OCT image of tissue within said coronary artery; and a controller configured for: using a trained fully convolutional engine stored on said memory and having a plurality of convolutional layers with respective dilation rates different than unity, extracting pathological tissues regardless their type in at least a region of interest of said OCT image; using a trained auto-encoder classification engine stored on said memory and having a layer characterized with a sparsity regularization parameter, determining an atherosclerotic pathological tissue type associated to said region of interest of said OCT image based on said extracted pathological tissues; and outputting said atherosclerotic pathological tissue type of said coronary artery.