Retinal Layer Thickness Detection Using Convolutional Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for measuring retinal layer thickness are manual, time-consuming, and prone to subjective errors, making them inefficient for clinical diagnosis and treatment guidance, especially in large-scale multicenter trials.
Innovation Solution
A retinal layer thickness detection model and system utilizing a convolution neural network to analyze optical coherence tomographic images, featuring image pre-processing, feature selection, training, and confirmation steps, which includes a database of reference images to automate the segmentation and quantification process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual segmentation is used to measure retinal layer thickness, then professional judgment can be applied, but the process becomes time-consuming and prone to subjective errors
Solution Approach 1:
The patent replaces the manual mechanical segmentation process with an automated image processing system that uses optical coherence tomography imaging and computer-based analysis. The system automatically segments retinal layers and measures thickness without human intervention, eliminating the time-consuming nature of manual methods while maintaining measurement accuracy through standardized algorithms.
Solution Approach 2:
The system enables self-service measurement where the imaging device automatically performs segmentation and thickness calculation without requiring professional judgment. The automated algorithm independently processes the images, making the measurement process efficient and removing the bottleneck of manual analysis while ensuring consistency across different cases.
2Reliability
If manual segmentation is used, then detailed professional analysis can be performed, but subjective judgment errors increase and treatment opportunities are lost
Solution Approach 1:
The patent replaces subjective human judgment with an objective automated analysis system. The computer-based segmentation and measurement eliminate variability in professional judgment while providing consistent, reliable results. The system uses standardized algorithms that objectively analyze retinal layer thickness without being influenced by individual clinician experience or fatigue.
Solution Approach 2:
The system provides objective feedback through automated measurement results that can be immediately used for diagnosis and treatment decisions. The consistent, reproducible measurements enable reliable tracking of retinal changes over time and response to treatment, improving diagnostic reliability without requiring complex manual assessment protocols.
3Productivity
If automated segmentation system is implemented, then time efficiency and objectivity are improved, but system complexity increases
Solution Approach 1:
The patent integrates multiple functions into a single automated system that performs image acquisition, preprocessing, segmentation, and thickness measurement. This multi-functional approach consolidates what would otherwise require multiple separate devices or manual steps, improving productivity while managing overall system complexity through integration rather than proliferation of separate components.
Data Source
AI summary
An establishing method of a retinal layer thickness detection model includes following steps. A reference database is obtained, and an image pre-processing step, a feature selecting step, a training step and a confirming step are performed. The reference database includes reference optical coherence tomographic images. In the image pre-processing step, the reference optical coherence tomographic images are duplicated and cell segmentation lines of retinal layers are marked to obtain control optical coherence tomographic images. In the feature selecting step, the reference optical coherence tomographic images are analyzed to obtain reference image features. The training step is to train with the reference image features and obtain the retinal layer thickness detection model. In the confirming step, marked optical coherence tomographic images are output by the retinal layer thickness detection model, and compared with the control optical coherence tomographic images to confirm an accuracy of the retinal layer thickness detection model.