Bruch's Membrane Segmentation Using OCTA-Enhanced OCT Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated retinal layer segmentation tools struggle with accurately segmenting poorly defined layers such as the Bruch's membrane and choroidal-scleral interface, especially in the presence of pathologies and low-quality OCT data, leading to misidentification errors and high computational costs.
Innovation Solution
Enhance structural OCT data using corresponding OCTA data to improve contrast around target retinal layers by attenuating similar regions and subtracting a weighted mixture of OCT and OCTA data, followed by a two-stage segmentation process to refine the results, and utilize OCTA data to identify and correct segmentation errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional automated segmentation algorithms are used, then processing speed is maintained, but segmentation accuracy deteriorates for poorly defined layers such as Bruch's membrane and choroidal-scleral interface
Solution Approach 1:
The patent divides the segmentation process into multiple stages: initial automated segmentation, error identification using OCTA data, and selective correction. This multi-stage approach maintains processing efficiency while improving accuracy for difficult-to-segment layers by focusing computational resources only where needed.
Solution Approach 2:
The patent introduces OCTA (OCT angiography) data as an intermediary to verify and correct segmentation results from structural OCT data. The OCTA data serves as a reference to identify segmentation errors, particularly for layers with poor definition, without requiring complete manual re-segmentation.
2Measurement precision
If manual segmentation is performed to improve accuracy, then segmentation precision improves, but time consumption increases significantly
Solution Approach 1:
Instead of performing complete manual segmentation, the patent applies partial manual verification only to regions where automated segmentation is likely to fail (poorly defined layers). This selective approach achieves high accuracy without the time cost of full manual segmentation.
Solution Approach 2:
The system uses OCTA data to provide feedback on segmentation accuracy, automatically identifying regions where segmentation may be incorrect. This feedback mechanism allows the system to focus verification efforts on problematic areas, reducing overall time consumption while maintaining high accuracy.
3Productivity
If automated segmentation is used for all layers, then processing efficiency is maintained, but reliability deteriorates in the presence of pathologies and low-quality OCT data
Solution Approach 1:
The patent changes the verification parameter by using OCTA data (which has different contrast characteristics) to validate structural OCT segmentation results. This parameter change allows the system to detect segmentation errors that would be invisible when using only structural OCT data, improving reliability without sacrificing efficiency.
Solution Approach 2:
The system combines structural OCT data and OCTA data into a composite verification approach. By using multiple data types with different properties, the system achieves higher reliability in pathological cases where single-modality data may be insufficient or misleading.
Data Source
AI summary
Retinal layer segmentation in optical coherence tomography (OCT) data is improved by using OCT angiography (OCTA) data to enhance a target retinal layer within the OCT data that may lack sufficient definition for segmentation. The OCT data is enhanced based on a mixture of the OCT data and OCTA data, such that contrast in the OCT data is enhanced in areas where OCT and OCTA data are dissimilar, and is reduced in areas where the OCT and OCTA data are similar. The target retinal layer in the OCT data is segmented based on the enhanced data. Two en face images of the OCTA data that include the target retinal layer are used to check for errors in the segmentation of the target retinal layer in the OCT data. Identified errors are replaced with an approximation based on the locations of top and bottom retinal layers of one of the en face images.