FAF Deep Learning Prediction of Geographic Atrophy Lesion Growth
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for evaluating geographic atrophy (GA) progression in age-related macular degeneration rely heavily on human graders, which are time-consuming, prone to errors, and produce variable results, and do not provide a visual depiction of future GA growth.
Innovation Solution
A deep learning system using fundus autofluorescence (FAF) image data to predict GA lesion growth over time, employing a trained long-short term memory convolutional neural network to generate growth images and masks indicating future GA lesion regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human graders are used to evaluate GA progression, then measurement precision can be achieved, but productivity decreases and time consumption increases
Solution Approach 1:
The patent uses deep learning models to create digital copies of human grader expertise. The trained neural networks replicate the measurement capabilities of human graders but execute them automatically and rapidly, eliminating the time-consuming manual review process while maintaining measurement accuracy through learned patterns from training data.
Solution Approach 2:
The patent replaces the mechanical human grading process with an automated deep learning system. The neural network algorithms substitute for manual visual inspection and measurement, enabling rapid processing of fundus autofluorescence images without human intervention while maintaining precision through computational analysis.
2Measurement precision
If human refiners manually adjust software outlines, then measurement precision improves, but productivity decreases and time consumption increases
Solution Approach 1:
The deep learning system performs self-service by automatically generating and refining GA lesion outlines without requiring human intervention. The neural network independently analyzes fundus autofluorescence images, identifies lesion boundaries, and produces measurements autonomously, eliminating the need for time-consuming manual refinement while maintaining high precision through its trained detection capabilities.
3Measurement precision
If human graders evaluate GA progression, then measurement precision is achieved, but reliability decreases due to variability in grader expertise
Solution Approach 1:
The patent implements homogeneity by replacing variable human expertise with a standardized deep learning model. The neural network provides consistent measurement results across different evaluations because it applies uniform computational criteria learned from training data, eliminating the variability inherent in human grader expertise while maintaining measurement precision through reproducible algorithmic analysis.
4Measurement precision
If current evaluation methods are used, then GA progression can be measured, but device complexity increases due to multi-step processes
Solution Approach 1:
The patent merges multiple evaluation steps into a single integrated deep learning process. The neural network combines image analysis, lesion identification, boundary detection, and measurement calculation into one unified operation, eliminating the need for separate manual outlining and measurement steps while maintaining comprehensive measurement precision through integrated computational analysis.
Data Source
AI summary
A method, implemented by one or more computer devices, includes receiving fundus autofluorescence (FAF) image data for a retina of a subject. The FAF image data includes a first FAF image associated with a first point in time. An image input for a deep learning system is generated using the FAF image data. A predicted growth output for a geographic atrophy (GA) lesion in the retina is generated via the deep learning system using the image input. The predicted growth output is associated with at least one future point in time after the first point in time.