A Novel Fusion Technique for Early Detection of Alopecia Areata Using ResNet-50 and CRSHOG

The study presents AlopeciaDet, a novel technique for early detection of Alopecia Areata using standard camera images, achieving an impressive accuracy of 99.45%. This method combines features from ResNet-50 and a new extraction technique called CRSHOG, enhanced by Principal Component Analysis for feature selection. The dataset included 811 healthy hair images and 71 Alopecia Areata images, with augmentation to improve generalization. AlopeciaDet surpasses existing methods in precision, recall, F1-score, and accuracy, offering a cost-effective alternative to dermoscopic imaging and potential application to other dermatological conditions.