Year: 2026
Type: article
Abstract: Skin diseases are a significant contributor to the global burden of disease, with an estimated 900 million individuals affected worldwide. The diversity of skin conditions, coupled with the shortage of trained dermatologists, leads to delayed diagnoses and increased morbidity. The present study addresses this challenge by developing a computer- aided diagnostic framework leveraging Convolutional Neural Networks (CNNs) and transfer learning using ResNet50 architecture. A comprehensive dataset was curated from multiple sources including Dermnet, ISIC, and HAM10000, subjected to extensive preprocessing steps such as resizing, normalization, and data augmentation to enhance model robustness. The baseline CNN provided a foundational performance benchmark, but substantial improvements were realized through fine-tuning a ResNet50 network pretrained on ImageNet. Model training incorporated optimization techniques such as Adam optimizer, learning rate scheduling, and early stopping. The final model achieved 80% classification accuracy, balanced precision-recall scores, and strong AUC values across disease categories. Results highlight the potential for deploying AI-assisted systems in primary care and teledermatology platforms to enhance early detection and reduce healthcare disparities. Future research should focus on augmenting datasets with synthetic data, integrating patient clinical information, and conducting large-scale prospective trials to validate clinical utility. This research underscores the importance of interdisciplinary approaches combining dermatology, computer science, and public health to improve skin disease outcomes.
Institutions: Veritas University, Baze University
Language: English
Subfield: Oncology
Field: Medicine
Domain: Health Sciences
Open Access status: diamond
OpenAlex ID: W7215565857