Topological Feature Extraction from Multi-color Channels for Pattern Recognition: An Application to Fundus Image...

Area of Science:

• Medical Imaging Analysis
• Computer Vision
• Artificial Intelligence in Healthcare

Background:

• Automated medical image analysis is vital for early disease detection.
• Deep learning models have shown significant promise in analyzing complex medical images.
• Understanding image topology and color variations can reveal crucial diagnostic information.

Purpose of the Study:

• To explore the efficacy of color-based topological features for pattern recognition in medical images.
• To integrate topological, Local Binary Pattern (LBP), and Gabor features with deep learning for disease classification.
• To evaluate the performance of these models on diverse public fundus image datasets.

Main Methods:

• Extraction of topological features across different color channels to capture shape and connectivity.
• Integration of topological, LBP, and Gabor features.
• Application of machine learning and deep learning models for classification tasks.
• Validation using three large-scale fundus image datasets (APTOS 2019, ORIGA, ICHALLENGE-AMD).

Main Results:

• Color-based topological features provide significant information for disease diagnosis.
• The combined feature approach demonstrated strong performance in classifying retinal diseases.
• The study highlights the importance of color space in topological feature extraction for medical imaging.

Conclusions:

• Color-based topological features are valuable for automated medical image analysis.
• Deep learning models augmented with these features enhance disease detection capabilities.
• This approach offers a promising direction for improving diagnostic accuracy in ophthalmology.