3D OCT Volumetric Processing for Choroidal Vessel Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional OCT imaging techniques face challenges in accurately segmenting choroidal vasculature due to noise, shadow artifacts, and limited depth resolution, leading to inaccurate quantifications and lengthy computation times, which hinder practical clinical applications for disease diagnosis and monitoring.
Innovation Solution
A medical diagnostic apparatus and method that processes 3D volumetric OCT data using a combination of pre-processing techniques, including deep-learning-based noise reduction and multiple segmentation methods, to generate enhanced 3D data sets for accurate choroidal vessel analysis, enabling visualization and quantification of choroidal vasculature.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional thresholding is applied to OCT images for vessel segmentation, then choroidal vasculature can be separated from stroma, but noise and shadow artifacts cause segmentation errors
Solution Approach 1:
The patent applies preliminary noise reduction filtering to the OCT data before performing thresholding-based vessel segmentation. By pre-processing the data to remove noise and minimize shadow artifacts, the subsequent segmentation step can proceed with higher accuracy and reliability, avoiding the errors that would otherwise occur when thresholding is applied directly to noisy data
2Measurement precision
If local binarization method is used to determine luminal and stromal areas, then choroidal structure can be segmented, but enhanced depth imaging protocol and averaged line scans are required to achieve sufficient quality
Solution Approach 1:
The patent changes the imaging parameters by using a single OCT scan with standard imaging protocol instead of requiring enhanced depth imaging protocol and averaged line scans. By optimizing the segmentation algorithm and applying appropriate noise reduction filtering, the system achieves sufficient segmentation quality from a single scan, thereby reducing the scanning time while maintaining measurement precision
3Productivity
If 2D projection images are used for choroidal vessel density measurement, then analysis can be performed, but depth resolution is lost and shadow artifacts affect results
Solution Approach 1:
The patent transitions from 2D projection images to 3D volumetric OCT data for vessel density measurement. By utilizing the full three-dimensional data set and applying noise reduction filtering, the system maintains analysis speed while recovering depth resolution and eliminating shadow artifact effects that plague 2D projections, thereby achieving both productivity and measurement precision
4Extent of automation
If automated detection of vessel boundaries is performed in 2D B-scans, then analysis can be automated, but shadow artifacts and limited vessel size range affect accuracy
Solution Approach 1:
The patent extends automated vessel boundary detection from 2D B-scans to 3D volumetric data. By performing automated detection in three dimensions, the system maintains automation while overcoming the limitations of shadow artifacts and restricted vessel size ranges that affect 2D methods, as the 3D context provides additional information for accurate boundary identification across all vessel scales
5Ease of manufacture
If segmentation is performed for each B-scan separately, then processing can be simplified, but vessel continuity is poor and piecing together segmented volume is required
Solution Approach 1:
The patent merges the segmentation process by performing it on the entire 3D volumetric data set as a unified whole rather than processing each B-scan separately. This combining approach maintains computational simplicity while ensuring continuous and consistent vessel segmentation across the entire volume, eliminating the need to piece together fragmented segmented B-scans and preserving vessel continuity
6Measurement precision
If conventional methods are used for 3D data processing, then analysis can be performed, but computation time is too long to limit the data that can be analyzed
Solution Approach 1:
The patent applies preliminary noise reduction filtering to the 3D volumetric data before performing detailed segmentation and analysis. By pre-processing the data to remove noise, the subsequent computational steps require less processing power and time, enabling faster analysis of large 3D data sets while maintaining the measurement precision needed for accurate clinical evaluation
Data Source
AI summary
A medical diagnostic apparatus includes a receiver circuit that receives three-dimensional volumetric data of a subject's eye, and a processor configured to separate portions of the three-dimensional volumetric data into separate segments, perform processing differently on each of the separate segments, and combine the separately processed segments to produce an enhanced three-dimensional volumetric data set. The processor is further configured to generate at least one diagnostic metric from the enhanced three-dimensional volumetric data set, and the processor is further configured to evaluate a pathological condition based on the at least one diagnostic metric. Related methods and computer readable media are also disclosed.