A Deep Learning Approach for Classification of Vitiligo and Scar Images

Skin diseases such as vitiligo present significant challenges in accurate diagnosis due to their similarities with other conditions, often requiring time-consuming assessments by medical professionals. Leveraging advancements in deep learning and medical image processing, this research proposes a novel approach to efficiently and accurately classify these conditions. The proposed model comprises multiple convolutional layers for extracting intricate features from skin images, followed by max-pooling layers to downsample spatial dimensions while retaining essential information. Additionally, fully connected layers enable complex feature learning and classification. This architecture, with six convolutional layers progressively increasing in depth and spatial reduction through max-pooling, culminates in three fully connected layers for precise classification. This comprehensive design ensures effective classification between vitiligo and scar images. By utilising this architecture, our model achieves a high training accuracy of 93.53% and precision of 96.50%, demonstrating its effectiveness in discriminating between vitiligo and scar images. This approach holds promise for enhancing diagnostic accuracy and efficiency in clinical settings, contributing to improved patient outcomes and healthcare delivery.