Optic Nerve Head Shape Classification Using Adaptive Image Quality Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for classifying optic nerve head shapes in glaucoma diagnosis are subjective and lack precision due to variability in fundus image quality, making automation challenging and inaccurate.
Innovation Solution
An ophthalmic information processing system that utilizes morphological data from eye fundus images and background data, employing machine learning techniques such as neural networks, gradient boosting decision trees, support vector machines, and Bayes classifiers to automate optic nerve head shape classification with improved accuracy and precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fundus images are used for optic nerve head shape classification, then the classification can be performed, but the image quality is greatly affected by photographing conditions making quantification difficult
Solution Approach 1:
The patent introduces an intermediary processing system that includes image quality assessment and adaptive parameter selection. The system evaluates photographing conditions and selects or adjusts classification parameters accordingly, acting as a mediator between the variable image quality and the classification process. This allows reliable classification even when image quality varies due to different photographing conditions.
2Measurement precision
If only fundus images are used for classification, then the processing is simple, but the classification accuracy and precision are insufficient
Solution Approach 1:
The patent segments the classification process into multiple independent modules: image quality assessment module, parameter selection module, classification execution module, and result interpretation module. Each module handles a specific aspect of the classification task, allowing the system to incorporate multiple data sources (fundus images, OCT data, patient history) without creating a monolithic complex system. This modular approach improves accuracy while managing complexity through clear separation of functions.
3Measurement precision
If all available ophthalmic information is considered for classification, then the classification accuracy may improve, but the processing becomes complicated and requires enormous resources
Solution Approach 1:
The patent dynamically changes parameters based on image quality assessment and clinical context. Instead of always processing all available ophthalmic information, the system adjusts the set of parameters to be processed based on the assessed quality of input data and the specific clinical scenario. This selective parameter processing maintains high classification accuracy when needed while improving processing efficiency by avoiding unnecessary computation when certain parameters are not critical or when image quality is insufficient.
Data Source
AI summary
An ophthalmic information processing system according to an embodiment includes a receiver and a processor. The receiver receives morphological data of an eye fundus of a subject. In addition, the receiver receives background data of the subject. The processor executes optic nerve head shape classification based on the morphological data and the background data received by the receiver.