DIAGNOSIS OF SKIN DISEASES USING CONVOLUTIONAL NEURAL NETWORK WITH DEEP LEARNING

Abstract

This paper proposes a CNN-based system for automated skin disease detection from images, addressing the shortage of dermatologists. Skin diseases are traditionally diagnosed by dermatologists, yet the shortage of specialists highlights the need for reliable automated solutions. This study introduces a robust image-based detection system using Convolutional Neural Networks (CNNs), which are highly effective in processing visual data. Our model achieves a validation accuracy exceeding 80% in identifying specific skin conditions, underscoring its potential despite challenges inherent to image-based diagnosis. A key limitation lies in the difficulty of compiling large, diverse datasets of skin disease images, which are essential for generalization across different skin types and conditions. Prediction errors may also arise from poor image quality, lighting variations, and visually similar diseases. By analysing and mitigating these factors, the system improves reliability and moves closer to bridging the gap between automated detection and expert human diagnosis. Overall, this research advances medical image analysis and provides a promising tool to support dermatologists and enhance patient outcomes.