Automated Retinal Image Analysis for AMD Severity Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting and classifying the severity of age-related macular degeneration (AMD) are inefficient, particularly in identifying the intermediate stage, which often goes unnoticed due to lack of symptoms, and require manual expertise, making it challenging to access timely diagnosis for the large at-risk population.
Innovation Solution
An automated system processes retinal images using a 'visual words' approach, identifying key image features and comparing them to reference data to determine the likelihood of AMD presence or progression, enabling early intervention and reducing false positives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual grading of fundus images by trained health care providers is used, then detection accuracy of intermediate stage AMD is improved, but accessibility and timeliness of diagnosis deteriorate due to limited provider availability
Solution Approach 1:
The patent creates a digital copy of the expert grader's knowledge through machine learning models trained on manually graded fundus images. The automated system replicates the detection capabilities of trained health care providers, enabling widespread deployment without requiring actual experts at every location.
Solution Approach 2:
The patent replaces the mechanical system of manual image review by human experts with an automated computational system. The machine learning algorithm processes fundus images automatically, substituting the human visual inspection process while maintaining detection accuracy.
2Productivity
If automated detection systems are implemented, then accessibility and screening capacity are improved, but measurement precision and reliability deteriorate due to lack of expert oversight
Solution Approach 1:
The patent performs preliminary action by training the machine learning model extensively on manually graded images before deployment. This pre-training phase incorporates expert knowledge into the automated system, ensuring it learns accurate detection patterns before being used for actual screening.
Solution Approach 2:
The patent implements feedback mechanisms where the automated system's detections can be reviewed and corrected, with this feedback used to further train and improve the model. This creates a continuous improvement loop that maintains and enhances detection accuracy over time.
3Reliability
If comprehensive manual evaluation of all at-risk individuals is conducted, then detection completeness is improved, but time consumption and resource requirements worsen
Solution Approach 1:
The patent enables self-service by allowing the automated system to independently evaluate fundus images without requiring human intervention for each case. The system autonomously processes images, identifies AMD indicators, and generates results, dramatically reducing the time and resources needed compared to manual evaluation.
Data Source
AI summary
A method of detecting, and classifying severity of, a retinal disease using retinal images includes at least one of receiving, retrieving or generating reference data that includes information concerning occurrences of key image features for each of a plurality of retinal disease and disease severity conditions; receiving a retinal image of an individual; processing the retinal image of the individual to identify occurrences of each of a plurality of distinguishable image features throughout at least a region of interest of the retinal image; identifying which ones of the identified occurrences of the plurality of distinguishable image features of the retinal image of the individual correspond to the key image features of the reference data; calculating, based on the identifying, a number of occurrences of each of the key image features in the retinal image of the individual; and determining at least one of a likelihood of a presence of a retinal disease or a likelihood of developing a retinal disease based on a comparison of the number of occurrences of each of the key image features in the retinal image of the individual to the reference data.