Mahatvak Chikitsa: Automated Skin Oncology Diagnosis with Clinical Decision Support

Abstract

Skin cancer is emerging as a significant global health issue, which makes timely detection increasingly crucial. Identifying skin-related conditions at an early stage can greatly enhance treatment success, yet differentiating between various types of skin lesions through visual inspection alone remains challenging. The project Mahatvak Chikitsa presents an intelligent web-based solution designed to assist in the detection of skin cancer by leveraging deep learning techniques on dermoscopic images. For system development, dermoscopic images from the ISIC dataset are utilized and processed using preprocessing techniques along with data augmentation to enhance image clarity and expand dataset variability. Multiple convolutional neural network architectures such as MobileNet, ResNet50, EfficientNet, VGG16, and VGG19 are implemented and compared to determine the most effective model in terms of classification accuracy and performance. A Flask-powered web interface is developed, allowing users to upload images, obtain predicted skin lesion categories, and access basic treatment-related guidance through a simple and intuitive design. This system is designed to act as an assisting tool for medical professionals rather than replacing expert diagnosis. By integrating artificial intelligence with a user-friendly web application, the project seeks to promote early-stage screening, increase awareness about skin health, and support clinicians in decision-making. Ultimately, this approach can help minimize diagnostic delays and improve patient outcomes by enabling quicker and more consistent evaluation of skin lesions.