3D OCTA Feature Space Classification for Blood Flow Noise
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current OCTA methods face challenges in accurately distinguishing blood flow from static areas due to noise interference, particularly in low signal-noise ratio (SNR) regions, leading to high classification errors and poor contrast in blood flow images.
Innovation Solution
A 3D angiography method using a multivariate time series model to establish a 2D feature space combining SNR and decorrelation coefficients, allowing for SNR-adaptive classification of dynamic blood flow signals and static tissues, thereby correcting noise artifacts and improving image contrast.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If decorrelation coefficient is used to classify blood flow signals, then classification reliability is improved, but noise interference causes false positive rates to increase in low SNR regions
Solution Approach 1:
The patent transitions from using a single decorrelation coefficient metric to a two-dimensional feature space that combines both intensity and decorrelation coefficient. This dimensional expansion allows the system to distinguish between true blood flow signals and noise artifacts by evaluating both signal strength and temporal variability simultaneously, thereby reducing false positives in low SNR regions while maintaining classification reliability.
Solution Approach 2:
The patent changes the classification parameters from a single decorrelation coefficient to a composite feature vector including intensity and decorrelation coefficient. By adjusting the feature space parameters and using adaptive thresholding based on local SNR estimation, the system optimizes the balance between detecting true blood flow and rejecting noise, resolving the contradiction between reliability and noise sensitivity.
2Ease of manufacture
If simple intensity masking is applied to remove low SNR signals, then noise removal is simplified, but classification error rate increases due to complex dependency relationship
Solution Approach 1:
The patent introduces an intermediate SNR estimation step that acts as a mediator between the raw OCT signals and the final classification. By first estimating local SNR using the relationship between intensity and decorrelation coefficient, then using this estimation to guide adaptive thresholding in the two-dimensional feature space, the system achieves more accurate noise removal while maintaining classification precision, avoiding the pitfalls of both simple and complex approaches.
3Reliability
If repeated sampling is performed to improve blood flow detection, then signal reliability is improved, but acquisition time increases
Solution Approach 1:
The patent extracts the essential blood flow information by performing repeated sampling only at selected spatial positions rather than uniformly across the entire imaging volume. By identifying and focusing computational resources on regions with potential blood flow based on preliminary analysis, the system maintains signal reliability through repeated measurements where needed while reducing overall acquisition time by minimizing redundant sampling in static tissue regions.
Data Source
AI summary
A three-dimensional (3D) optical coherence tomography angiography (OCTA) method and system based on feature space is provided. OCT signals of scattering samples in a 3D space are acquired by a collector; a theoretical classifier, local signal-noise ratios (SNRs) and decorrelation coefficients of the OCT scattering signals are combined to establish a two-dimensional (2D) feature space to realize classification of dynamic blood flow signals and static tissues. Specifically, computational analysis of first-order and zero-order autocovariance is used to obtain two features of SNR and decorrelation of each OCT scattering signal; a 2D inverse SNR (iSNR)-decorrelation feature (ID) space is established; and based on the multivariate time series theory, a linear classifier is established in the ID space to remove a static surrounding tissue background. The 3D OCTA method and system can improve the contrast of blood flow images and-improve the accuracy of blood flow quantification.