Automated Diagnosis and Classification of Eczema and Scabies Using Deep Learning Technique of Ai

Abstract

Scabies and eczema are two distinct yet often visually similar dermatological conditions that pose significant diagnostic challenges for dermatologists. Scabies is a contagious skin infestation caused by the mite Sarcoptesscabiei, a parasite that burrows into the skin, leading to intense itching, small blisters, and patchy lesions. The disease is highly transmissible through direct contact and can affect various body regions, including the elbows, betweenthefingers,genitalarea,armpits,waist,and, in severe cases, even the face in infants. Eczema, on the other hand, isan inflammatory,non-contagious skin disorder resulting froma complex interplayofgenetic, environmental,andimmune factors. While not classified asan autoimmune disease, eczema is linked to immune regulations that increases susceptibility to other immune-related conditions. Clinically, eczema is characterized by scaly, red patches and blisters, commonly appearing on flexural regions such as the knees, wrists, hands, and neck. Due to overlapping visual manifestations such as erythema, scaling, and blistering, differentiating between eczema and scabies through traditional clinical inspection can be challengingandmay leadto misdiagnosis.To address this diagnostic complexity,this research proposes an automated classification system for eczema and scabies using a Deep Convolutional Neural Network (DCNN).The DCNN architecture is designed to effectively extract and learndiscriminativefeaturesfromdermatologicalimagesthrough multiple layers of convolution, pooling, and fully connected operations. The dataset used in this study comprises labeleddermatoscopicandclinicalimagesofeczemaandscabiescollected from verified dermatological sources. Pre-processing techniques such as image augmentation, normalization, resizing, and contrast enhancement were applied to improve the robustness and generalization ability of the model. The DCNN was trained on these processed images to capture spatial hierarchies of patterns associated with each disease, enabling high-level feature representation for accurate classification. The superior accuracy achieved by the DCNN can be attributed to its deep feature extraction capability, which efficiently captures texture irregularities, lesion boundaries, and structural variations unique to each skin condition. The model’s performance was further validated using evaluation metrics such as precision, recall, F1-score, and confusion matrix analysis, all indicating high reliability and robustness. The study emphasizes that the integration of DCNN- based diagnostic systems can significantly assist dermatologistsin early and accurate identification of skin diseases, reducing diagnosticerrorsandenab lingtimely treatment. Ultimately, the proposed system paves the way toward intelligent, data- driven healthcare solutions for improved dermatological disease management and patient outcomes.