OCT Image Classification Without Retinal Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Optical Coherence Tomography (OCT) image analysis methods face challenges in accurately classifying retinal images, especially for conditions like Age-related Macular Degeneration (AMD), due to difficulties in segmenting tissue layers and varying image quality, which affects diagnosis and treatment efficacy.
Innovation Solution
A system and method that preprocess OCT images using filters like the Frangi filter, extract vertical transects, generate profiles from these transects, and classify images using algorithms such as linear discriminant analysis (LDA), k-nearest neighbor, or support vector machines, without the need for initial segmentation, allowing for robust classification across different image qualities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional OCT image analysis methods are used to classify retinal images, then segmentation of tissue layers is attempted, but classification accuracy deteriorates due to varying image quality and difficulty in segmenting tissue layers
Solution Approach 1:
The patent extracts only the necessary information (intensity values along vertical transects) from the OCT images without performing full segmentation of tissue layers. By extracting transects at specific locations (including the foveal region) and analyzing intensity profiles directly, the method avoids the reliability issues associated with segmentation while maintaining classification accuracy.
Solution Approach 2:
The patent changes the analysis parameter from segmented layer structures to intensity value profiles along vertical transects. By transforming the image data into intensity profiles and analyzing statistical properties (mean, standard deviation, skewness, kurtosis) of these profiles, the method achieves robust classification that is insensitive to image quality variations.
2Loss of information
If detailed segmentation of retinal layers is performed, then layer identification is achieved, but processing complexity and time increase
Solution Approach 1:
The patent extracts only the essential intensity information along vertical transects rather than performing complete segmentation of all retinal layers. This extraction approach maintains the ability to identify layer characteristics (through intensity profile analysis) while significantly reducing processing time and computational complexity.
Solution Approach 2:
The patent performs partial segmentation by analyzing only specific vertical transects (particularly through the foveal region) rather than segmenting the entire retinal image. This partial action approach maintains diagnostic information while reducing processing requirements.
3Measurement precision
If multiple image processing steps are applied to improve classification, then diagnostic accuracy improves, but system complexity increases
Solution Approach 1:
The patent merges multiple analysis steps into a unified workflow: extracting vertical transects, computing intensity profiles, calculating statistical parameters, and applying classification algorithms. By combining these steps into an integrated system that processes transect data rather than requiring separate segmentation and analysis modules, the patent reduces system complexity while maintaining diagnostic accuracy.
Data Source
AI summary
The present disclosure describes a system and method to classify optical coherence tomography (OCT) images. The present system can classify OCT images without first segmenting the retina tissue. The system can generate one or more profiles from vertical transects through the OCT images. The system can identify image statistics based on the one or more profiles. The system's classifier can then classify the OCT images based on the identified image statistics.