MG-FACS: MULTI-STAGE GRAPH-BASED AND FUZZY ACTIVE CONTOUR SEGMENTATION FOR ROBUST TOENAIL DISEASE DETECTION

Abstract

Pre-disease detection using non-invasive imaging is emerging as a critical component of early healthcare diagnostics, particularly in dermatology and podiatry. This research introduces MG-FACS (Multi-Stage Graph-Based and Fuzzy Active Contour Segmentation), an advanced segmentation framework tailored for detecting and delineating diseased toenail regions. MG-FACS integrates graph-based modeling, superpixel-based region simplification, and adaptive fuzzy logic for enhanced contour detection. The system addresses common challenges in toenail image segmentation, including background noise, lighting inconsistency, and anatomical variability. Performance comparisons across ten images using Otsu thresholding, Watershed, K-Means, and MG-FACS reveal significant improvements in accuracy, sensitivity, specificity, Dice coefficient, and Jaccard index. Results consistently show MG-FACS achieving over 95% accuracy and more than 90% Dice similarity, outperforming traditional approaches. The proposed method demonstrates potential for real-world diagnostic support in mobile and clinical environments.