Automated Trichoscopic Analysis for Hair Disorder Classification Using Semantic Segmentation

This study introduces an automated trichoscopic image analysis framework to improve the classification of hair disorders such as Alopecia, Alopecia Areata, and Telogen Effluvium, which are prevalent among young adults. The framework uses a U-Net architecture for semantic segmentation, extracting quantitative features like hair density and shaft diameter from 110 trichoscopic images. Statistical analysis showed significant differences between the disorders, and the Random Forest classifier achieved the highest accuracy of 86.67% with a ROC-AUC of 93.38%. This method offers objective and reproducible scalp disorder classification, paving the way for future real-time and mobile-based diagnostic systems.