Retinal Fundus to OCT Projection Image Reconstruction Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting age-related macular degeneration (AMD) rely on expensive optical coherence tomography (OCT) imaging, which is not widely available in remote and rural areas, limiting early detection and prediction capabilities.
Innovation Solution
A method to generate estimated OCT projection images from retinal fundus images using feature extraction and registration, allowing for the creation of a model that predicts AMD without the need for expensive OCT imaging equipment, leveraging deep convolutional neural networks and auto-encoders to correlate changes in retinal fundus images with OCT projection images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If optical coherence tomography (OCT) imaging is used for AMD prediction, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent creates a computational model that copies or replicates OCT projection images from retinal fundus images. The model learns the relationship between these two image types through training on paired data, allowing it to generate synthetic OCT-like images from inexpensive fundus photography. This copying approach enables high-precision AMD prediction without requiring actual OCT equipment.
Solution Approach 2:
The patent replaces expensive, complex OCT imaging equipment with inexpensive retinal fundus photography. The computational model acts as a disposable software layer that translates cheap fundus images into useful OCT-like projections, eliminating the need for costly hardware while maintaining prediction accuracy.
2Measurement precision
If optical coherence tomography (OCT) imaging is used for AMD prediction, then measurement precision is improved, but cost increases
Solution Approach 1:
The system copies OCT projection images from retinal fundus images using a trained computational model. This allows generation of high-quality OCT-like images from inexpensive fundus photography, significantly reducing the cost barrier for AMD screening and prediction while maintaining diagnostic accuracy.
Solution Approach 2:
The patent uses inexpensive retinal fundus photography as a substitute for expensive OCT equipment. The computational model serves as a cost-effective intermediary that transforms cheap input images into useful output images, making high-precision AMD prediction accessible to low-resource settings.
3Device complexity
If OCT projection images are generated from retinal fundus images, then device complexity is reduced, but manufacturing precision may worsen
Solution Approach 1:
The computational model is trained using feedback from paired retinal fundus images and corresponding OCT projection images. During training, the model continuously refines its ability to accurately transform fundus images into OCT-like projections by comparing generated outputs with ground truth OCT images, ensuring high manufacturing precision is achieved despite reduced device complexity.
Solution Approach 2:
The patent changes the fundamental parameters of image acquisition by transitioning from optical-based OCT imaging to computational image generation. The model learns to preserve critical structural and pathological features by transforming image parameters through a neural network, maintaining manufacturing precision while eliminating complex imaging hardware.
Data Source
AI summary
An AMD prediction model utilizes an OCT image estimation model. The OCT image estimation module is created by segmenting an OCT image to generate an OCT projection image for each of multiple biological layers; extracting from each of the generated OCT projection images a first set of features; extracting a second set of features from an input retinal fundus image; for each respective biological layer, registering the input retinal fundus image to the respective OCT projection image by matching at least some of the second set of features with corresponding ones of the first set of features; repeating the above with changes to the input retinal fundus image; and modelling how the changes to the input retinal fundus image are manifest at the correspondingly registered projection images. Estimated OCT projection images can then be generated for the multiple biological layers from a given retinal fundus image.