Retinal Microaneurysm Classification Using FA Leakage Dynamics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current clinical methods for diagnosing diabetic retinopathy, such as fluorescein angiography (FA) and optical coherence tomography (OCT), are inadequate in accurately detecting and classifying retinal microaneurysms, particularly in distinguishing between leaky and non-leaky microaneurysms, which are crucial for early-stage diagnosis and treatment planning.
Innovation Solution
A method and system for detecting and classifying retinal microaneurysms using time sequences of fluorescein angiography images, involving preprocessing, binarization, vessel removal, feature extraction, and classification based on morphological grading metrics, including rigid and non-rigid registration, and analysis of fluorescence intensity changes to determine leaky or non-leaky microaneurysms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fluorescein angiography is used to detect microaneurysms, then detection capability is improved, but the ability to classify leaky vs non-leaky microaneurysms remains insufficient
Solution Approach 1:
The patent segments the FA image analysis into multiple phases (early, intermediate, late) and divides the classification task into separate modules: detection module for locating microaneurysms, and classification module for determining leakage status. This segmentation allows independent optimization of detection accuracy and leakage classification, resolving the contradiction by addressing each function separately rather than relying on a single undifferentiated detection approach.
Solution Approach 2:
The patent performs preliminary actions by pre-processing the FA images to enhance contrast and segment blood vessels before the actual detection and classification. It also pre-establishes morphological grading metrics and leakage criteria. These preliminary preparations enable more accurate subsequent detection and classification, improving both detection accuracy and leakage status identification without requiring additional complex processing steps during the main analysis phase.
2Reliability
If manual clinical diagnosis methods are used, then diagnostic capability is maintained, but productivity and detection accuracy are insufficient
Solution Approach 1:
The patent implements self-service by enabling the system to automatically perform both detection and classification of microaneurysms without requiring manual intervention. The automated algorithm processes FA images independently, applying morphological grading metrics and leakage detection criteria to classify each microaneurysm. This self-service capability maintains diagnostic reliability by using standardized, objective criteria while dramatically improving productivity through automated high-speed processing.
Solution Approach 2:
The patent replaces the mechanical manual diagnostic process with an automated computational system. Instead of human clinicians manually reviewing images and making subjective assessments, the system uses image processing algorithms, morphological operations, and automated classification criteria. This substitution maintains diagnostic reliability through objective, reproducible criteria while achieving superior productivity through rapid automated analysis of multiple images.
3Ease of operation
If simple detection methods are used, then ease of operation is improved, but measurement precision and classification capability deteriorate
Solution Approach 1:
The patent achieves multi-functionality by integrating both detection and classification capabilities into a single unified system. The same processed FA images and morphological metrics serve dual purposes: first for detecting the presence and location of microaneurysms, then for classifying their leakage status. This universal approach maintains ease of operation through a single integrated workflow while improving measurement precision by applying comprehensive analysis rather than simple single-purpose detection.
Solution Approach 2:
The patent applies parameter changes by systematically varying morphological grading metrics (size, shape, intensity) and temporal parameters (fluorescence intensity over time) to enable accurate classification. The system adjusts these parameters dynamically during analysis, using different weightings and thresholds for detection versus classification phases. This parameter optimization maintains operational simplicity through automated parameter adjustment while achieving high measurement precision in microaneurysm classification.
Data Source
AI summary
Systems and methods for detecting and classifying retinal microaneurysms. The method including: receiving a time sequence of fluorescein angiography input images; generating a binary map of hyperfluorescent elements in the input images; determining which hyperfluorescent elements in the binary map are microaneurysms, by grading each against a combination of morphological metrics; classifying each of the detected microaneurysms as leaky or not leaky, the classification having: identifying an outer ring mask surrounding the detected microaneurysm in the binary map; identifying parenchyma in the outer ring mask using a fluorescence intensity determination; determining a rate of change of fluorescence intensity of the identified parenchyma over time; and classifying the detected microaneurysm as leaky where the rate of change is positive and not leaky where the rate of change is negative or zero; and outputting the classifications of the detected microaneurysms.