OCT Image Normalization via Unified GAN Representation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Medical professionals face difficulties in comparing medical images captured by different image capture devices due to variations in resolution and noise types, making it challenging to analyze and monitor patient conditions consistently across different medical facilities.
Innovation Solution
A computer-implemented method using machine learning models, specifically Generative Adversarial Networks (GANs) and Convolutional Neural Networks (CNNs), normalizes OCT image data to generate device-specific scans with consistent resolution and noise types, enabling comparison of images from various devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If OCT images are captured by different image capture devices, then the quantity of images increases and more data becomes available, but the images have different resolutions and noise types making them difficult to compare
Solution Approach 1:
The patent applies parameter changes by transforming images from different devices into a unified representation space with standardized parameters (resolution, noise type). The GAN-based system modifies image parameters through learned transformations, converting variable-resolution and variable-noise images into consistent normalized images while preserving diagnostic information.
Solution Approach 2:
The patent introduces an intermediary unified representation space that mediates between images from different devices. This intermediate space acts as a common language that all device-specific images are transformed into, enabling consistent comparison without requiring direct comparison between incompatible device formats.
2Adaptability or versatility
If medical facilities use different image capture device technologies, then device versatility increases, but comparing images across facilities becomes difficult
Solution Approach 1:
The patent implements universality by creating a unified representation space that can accommodate images from multiple different device types. The system maintains the ability to process and normalize images from various OCT devices while producing a single standardized output format that can be universally compared across all facilities.
Solution Approach 2:
The system changes image parameters through GAN-based transformation, converting device-specific parameters (resolution, noise characteristics) into a unified set of parameters. This allows images from diverse devices to be transformed into a common format that maintains diagnostic quality while enabling easy comparison.
3Measurement precision
If images are normalized to a unified representation, then image consistency improves, but the process requires complex machine learning models
Solution Approach 1:
The patent uses copying by creating a unified representation that replicates the essential diagnostic features of original images from different devices. Instead of modifying the original complex device-specific images, the system creates simplified copies in a unified space that retain the necessary diagnostic information while being consistent across devices.
Solution Approach 2:
The unified representation space serves as an intermediary that simplifies the comparison process. Rather than developing complex methods to directly compare incompatible images, the system introduces this intermediate representation layer that automatically handles the complexity of cross-device normalization.
Data Source
AI summary
In an aspect for generating device-specific OCT image, one or more processors may be configured for receiving, at a unified domain generator, first image data corresponding to OCT image scans captured by one or more OCT devices; processing, by the unified domain generator, the first image data to generate second image data corresponding to a unified representation of the OCT image scans; determining by a unified discriminator, third image data corresponding to a quality subset of the unified representation of the OCT image scans having a base resolution satisfying a first condition and a base noise type satisfying a second condition; and processing, using a conditional generator, the third image data to generate fourth image data corresponding to device-specific OCT image scans having a device-specific resolution satisfying a third condition and a device-specific noise type satisfying a fourth condition.