OCT Registration Using Multiple Sub-Volume Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The registration of corresponding OCT data sets is challenging due to pathological changes, layer segmentation errors, and quality variations in OCT data, leading to difficulties in accurately aligning landmarks and detecting subtle changes over time.
Innovation Solution
The method improves registration by using multiple pairs of 2D maps derived from OCT volumes, including thickness maps and curvature maps, to identify more consistent characteristic features and define a custom transformation model for registering OCT data sets, even in low-quality or deceased tissue images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single en face or vasculature map is used for registration, then the registration process is simple, but the accuracy and reliability of landmark alignment deteriorates due to insufficient or poorly distributed landmarks
Solution Approach 1:
The patent segments the registration task by using multiple separate en face maps (structural en face, vascular en face, thickness map, curvature map) instead of a single map. Each map provides different characteristic features for landmark detection, and all maps are processed independently before being combined for registration, thereby improving landmark distribution and alignment accuracy without excessive complexity increase
Solution Approach 2:
The patent merges multiple en face maps and their corresponding characteristic features into a unified registration process. The system combines landmark matches from structural en face, vascular en face, thickness map, and curvature map to create a comprehensive transformation model, thereby improving registration accuracy by utilizing diverse landmark sources
2Quantity of substance
If multiple en face maps are combined to improve landmark distribution, then the number of identifiable landmarks increases, but the complexity of processing and selecting appropriate maps increases
Solution Approach 1:
The patent segments the processing into distinct stages: first generating multiple en face maps from OCT volume data, then detecting characteristic features in each map independently, and finally combining the landmark matches for registration. This segmentation allows systematic handling of multiple maps without overwhelming processing complexity
Solution Approach 2:
The patent creates a universal registration framework that can handle multiple types of en face maps (structural, vascular, thickness, curvature) using a common processing pipeline. The system universally applies characteristic feature detection and landmark matching across all map types, reducing processing complexity through standardized procedures
3Productivity
If automated segmentation algorithms are used to generate 2D maps, then the process is efficient and automated, but segmentation errors occur due to severe pathology leading to partially incorrect 2D maps
Solution Approach 1:
The patent applies local quality by using multiple different en face maps (structural, vascular, thickness, curvature) that highlight different local characteristics of the retina. Each map provides complementary information that can compensate for segmentation errors in other maps, allowing the system to adapt to severe pathology by utilizing the most reliable local features available
Solution Approach 2:
The patent incorporates feedback by using the detected characteristic features and landmark matches to validate and refine the segmentation results. The system iteratively processes the multiple en face maps and adjusts the registration transformation model based on the quality and consistency of identified landmarks, thereby correcting segmentation errors
4Ease of operation
If traditional single-map registration is used, then the method is well-established and simple, but it cannot adequately handle large lateral motion or subtle changes over time
Solution Approach 1:
The patent introduces another dimension by incorporating multiple en face maps with different characteristics (structural, vascular, thickness, curvature) into the registration process. This multi-dimensional approach provides richer information for detecting both large lateral motion and subtle changes over time, improving measurement precision while maintaining operational simplicity through automated processing
Data Source
AI summary
A system/method/device for registering two OCT data sets defines multiple image pairs of corresponding 2D representations of one or more corresponding sub-volumes in the two OCT data sets. Matching landmarks in the multiple image pairs are identified, as a group, and a set of transformation parameters are defined based on matching landmarks from all of the image pairs. The two OCT data sets may then be registered based on the set of transformation parameters.