Skin Disease Classification Using CNN

Abstract: Skin diseases represent a significant global healthcare challenge, with certain malignant lesions such as melanoma requiring early diagnosis to improve patient survival rates. Traditional dermatological diagnosis primarily depends on visual examination by specialists, which may be subjective, time-consuming, and limited by the availability of experienced dermatologists. This paper presents a deep learning–based multi-class skin disease classification system using Convolutional Neural Networks (CNNs) for automated analysis of dermoscopic skin images. The proposed framework classifies skin lesions into nine distinct disease categories using a publicly available dermoscopic image dataset. Image preprocessing techniques including resizing, normalization, and augmentation are applied to improve model generalization and classification performance. To enhance interpretability, Explainable Artificial Intelligence (XAI) techniques based on Gradient-weighted Class Activation Mapping (Grad-CAM) are incorporated to visualize important lesion regions influencing prediction outcomes. The trained CNN model is integrated into a Streamlit-based web application that enables real-time image upload and disease prediction along with confidence visualization. Experimental evaluation demonstrates that the proposed system achieves reliable multi-class classification performance with a validation accuracy of approximately 87%. The proposed framework serves as an intelligent decision-support tool for educational and preliminary diagnostic assistance and highlights the potential of explainable deep learning techniques in medical image analysis.This paper discusses the system architecture, methodology, implementation, and performance evaluation of the proposed solution.

Keywords: Skin Disease Classification, Deep Learning, Convolutional Neural Network, Explainable AI, Grad-CAM, Dermoscopic Image Analysis, Streamlit