Recognition of Scalp Psoriasis Dermoscopic Patterns Signs Based on Histogram Equalization and Convolutional Neural Networks

Scalp psoriasis is a common chronic autoimmune inflammatory skin disease that affects a large number of people worldwide. Although scalp psoriasis is not physically a serious life-threatening condition for patients, the itchiness, inflammation, and social distress it causes have a significant negative impact on their quality of life. Currently, the urgent need for accurate diagnosis of scalp psoriasis has driven researchers to explore efficient diagnostic techniques utilizing deep learning techniques, among which dermoscopy is widely used as a simple and easy-to-use non-invasive testing device. In this paper, a convolutional neural network (HE-CNN) based on histogram equalization is proposed to solve the problem of automatic recognition of different dermoscopic patterns of scalp psoriasis. In this, histogram equalization is used to perform global brightness enhancement of dermoscopic images to ensure that important lesion features in the image are not obscured by hair shadows leading to neglect. CBAM is added to the CNN and the enhanced images are used to train a classification model for automated recognition of scalp psoriasis for clinical detection. We conducted experiments on a scalp psoriasis dataset provided by 72nd Group army hospital of PLA. In order to verify the effectiveness of the method, we use the unenhanced image as a control group and compare it with the image that has undergone enhancement, and the results after five-fold cross-validation show that the recognition effect of HE-CNN outperforms that of the control group in terms of accuracy, specificity, sensitivity and youden index. It is verified that HE-CNN has better effect in assisting doctors in the diagnosis of scalp psoriasis.