Bridging the Diagnostic Gap: A Web-Based Teledermatology Framework for Accessible Early Detection of Skin Lesions using Deep Learning

Abstract

Early identification of dermatological conditions is critical for preventing long-term complications; however, timely access to qualified dermatologists remains limited, especially in rural and underserved regions. Although Artificial Intelligence (AI) has shown strong performance in skin-image interpretation, most existing systems require specialized hardware, native applications, or high-bandwidth environments. This paper introduces DermaScan, a fully web-based, device-agnostic teledermatology framework designed to deliver rapid, preliminary assessment of skin lesions using a lightweight deep-learning model. The system integrates the MERN stack for frontend–backend communication and employs an optimized MobileNetV2 network for real-time inference through a Python microservice. The proposed framework emphasizes accessibility, fairness across diverse skin tones, low computational overhead, and clinically meaningful triage performance. Experimental results demonstrate high sensitivity, low latency, and stable performance under varied lighting and device conditions. DermaScan highlights the potential of web-native AI tools to reduce diagnostic delay, support early awareness, and provide equitable dermatology assistance at scale.