Pixel-Level Corneal Biomechanics Identification via Ensemble Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for measuring corneal biomechanics, such as ORA and Corvis ST, lack the accuracy to detect local subtle mechanical changes in the cornea, as they primarily provide overall mechanical information.
Innovation Solution
A subtle cornea deformation identification method and device based on pixel-level corneal biomechanical parameters, which involves sampling and analyzing dynamic corneal stress deformation videos to calculate pixel-level data and using an ensemble classifier to detect local changes in corneal biomechanics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If commercial devices (ORA or Corvis ST) are used to measure corneal biomechanics, then overall mechanical information can be obtained, but local subtle mechanical changes cannot be detected due to insufficient measurement accuracy
Solution Approach 1:
The patent divides the cornea into multiple regions of interest (ROIs) and performs measurements at different locations including the apex and peripheral areas. This segmentation approach enables detection of local subtle mechanical changes that would be missed by overall corneal measurements alone, directly resolving the contradiction between measurement accuracy and device complexity.
Solution Approach 2:
The patent transitions from traditional overall corneal biomechanics measurement to pixel-level or regional biomechanics analysis by introducing spatial dimensionality. Through coordinate transformation and region-of-interest extraction, the system achieves high-precision local measurement without requiring completely new complex hardware, thus resolving the measurement accuracy versus device complexity contradiction.
2Measurement precision
If pixel-level corneal biomechanical parameters are calculated through sampling and ensemble classification, then local subtle deformation can be identified with high accuracy, but computational complexity increases
Solution Approach 1:
The patent performs preliminary actions by pre-defining regions of interest, pre-calculating coordinate transformation matrices, and pre-establishing ensemble classification models. These preliminary preparations reduce the computational burden during actual measurement and enable high-precision local deformation detection without excessive real-time computational complexity.
Solution Approach 2:
The patent uses coordinate transformation to create a mapped copy of the corneal surface from the imaging coordinate system to the anatomical coordinate system. This copying approach allows pixel-level parameter calculation in the transformed space, achieving high detection accuracy while simplifying the computational process through mathematical mapping rather than direct complex image analysis.
3Loss of information
If 14 pixel-level corneal biomechanical parameters are calculated for each region, then comprehensive local mechanical information is obtained, but data processing time increases
Solution Approach 1:
The patent extracts only the most relevant 14 pixel-level biomechanical parameters from the full image data, such as deformation depth, curvature changes, and strain values at key locations. This selective extraction approach maintains comprehensive local mechanical information while significantly reducing data processing time compared to analyzing all possible parameters.
Solution Approach 2:
The patent applies local quality by calculating parameters specifically at defined regions of interest rather than uniformly across the entire corneal surface. This approach concentrates computational resources on areas most likely to show subtle changes, maintaining information completeness in critical areas while reducing overall processing time.
Data Source
AI summary
The present disclosure relates to a subtle cornea deformation identification method and device based on a pixel-level corneal biomechanical parameter, including the following steps: step 1, sampling and analyzing a dynamic video of corneal stress deformation in a historical database, and calculating pixel-level data; and step 2, configuring an ensemble classifier based on a sampling result and detecting a local change in corneal biomechanics, thus identifying a subtle cornea deformation. The present disclosure has high measurement accuracy and is capable of detecting a local subtle biomechanical change.