Retinal Layer Auto-Segmentation Model Using U-Net
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for retinal layer segmentation in optical coherence tomography images are manual, time-consuming, and prone to errors, limiting their clinical utility and accuracy in diagnosing ophthalmic and neurodegenerative diseases.
Innovation Solution
A retinal layer auto-segmentation model is developed using a U-net convolution neural network, which involves preprocessing, feature selection, data enhancement, training, and validation to automatically segment and quantify retinal layers, reducing human error and increasing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual segmentation is used for retinal layer analysis, then professional judgment can be applied, but the process becomes time-consuming and prone to human error
Solution Approach 1:
The patent replaces the manual mechanical segmentation process with an automated deep learning system using U-net convolutional neural network. The system processes optical coherence tomography images through automated algorithms that identify and segment retinal layers without human intervention, eliminating time consumption and human error while maintaining high accuracy through trained neural networks
2Reliability
If manual segmentation by professional judges is used, then accurate clinical judgment can be made, but the complexity and cost of the system increases
Solution Approach 1:
The patent implements a self-service automated segmentation system where the U-net neural network independently processes retinal images without requiring professional judges. The system performs preprocessing, segmentation, and measurement automatically, reducing system complexity by eliminating the need for human expert intervention while maintaining diagnostic reliability through algorithmic consistency
Solution Approach 2:
The patent introduces an automated image processing system as an intermediary between image acquisition and clinical diagnosis. This intermediary layer uses deep learning algorithms to bridge the gap, providing consistent and reliable segmentation results without requiring direct human judgment, thereby simplifying the overall system workflow
3Measurement precision
If manual segmentation is performed, then detailed professional assessment can be conducted, but productivity and efficiency are reduced
Solution Approach 1:
The patent replaces manual measurement processes with automated computational algorithms that rapidly process retinal images. The U-net system simultaneously performs multiple measurements including layer thickness, area, and volume calculations, achieving high precision measurements at speeds impossible for manual analysis, thereby dramatically increasing clinical throughput and productivity
Data Source
AI summary
A retinal layer quantification system includes an image capturing unit and a processor electrically connected to the image capturing unit. The image capturing unit is configured to capture a target optical coherence tomographic image of a subject. The processor stores a program including a target image pre-processing module, a retinal layer auto-segmentation model, a target image enhancement module, a layer thickness detection module and a layer area detection module. The program detects a retinal layer thickness and a retinal layer area of the subject when the program is executed by the processor.