FRCNN based Deep Learning for Identification and Classification of Alopecia Areata

Most patients suffer from scalp hair issues like dermatitis, baldness, and so on as an outcome of unhealthy lifestyles, hormonal imbalance, and so on. The most common type of alopecia is alopecia areata (AA), which is typically detected and diagnosed using medical image processing models. However, they are unreliable for striking or intersecting hairs and are directly influenced by design variables. Deep Learning (DL) in imaging data was thus used to detect and diagnose AA. Similarly, various DL models recognized and diagnosed various scalp hair conditions. Likewise, various scalp hair conditions were recognized and diagnosed by the various DL models. Hence, this paper proposes a Faster Residual Convolutional Neural Network (FRCNN) model to recognize AA and scalp conditions together for many individuals with different kinds of baldness. The main goal of the FRCNN is to use a Region-Of-Interest (ROI) projection layer to enhance the recognition accuracy of AA and scalp hair symptoms. In this FRCNN model, the given scalp and AA images are initially fed to the ROI projection layer and deep convolutional layer to extract feature maps, which are aggregated by the ROI pooling to get the final feature vector representation. Then, the obtained feature vector is passed to the Fully Connected (FC) layer accompanied by the softmax classifier to recognize the various conditions of AA. Finally, the test results show that the FRCNN on hair and scalp image databases achieves an accuracy of 84.3% compared to the other models.