A Real-Time Skin Lesion Classification System Using Deep Learning and Flask Framework for Remote Healthcare Diagnostics

Abstract

Skin disorders are among the most prevalent medical conditions worldwide, particularly affecting individuals in underdeveloped and low-resource regions where access to dermatologists is limited. Delayed or inaccurate diagnosis often leads to disease progression, secondary infections, increased treatment costs, and prolonged patient discomfort. Although deep learning techniques have demonstrated strong performance in medical image analysis, many existing models are computationally expensive and unsuitable for real-time or web-based deployment [1][2]. This paper presents an efficient skin disease classification system based on the MobileNetV2 architecture for accurate real-time diagnosis of eight common skin diseases, using a dataset of 1,247 clinical images. The model leverages transfer learning with ImageNet-pretrained weights, extensive data augmentation, and class-balancing strategies to improve performance and generalization. The trained model is deployed through a Flask-based web application to support remote healthcare diagnosis. Experimental results demonstrate high classification accuracy across all disease categories, with F1-scores ranging from 0.91 to 1.00, and low computational complexity, making the proposed system well-suited for telemedicine and practical clinical applications.