Authors:
A. Muthukumaravel, S. Silvia Priscila, B. M. Praveen
Addresses:
Faculty of Arts and Science, Bharath Institute of Higher Education and Research, Chennai, Tamil Nadu, India. Institute of Engineering and Technology, Srinivas University, Dakshina Kannada, Karnataka, India. Department of Computer Science, Bharath Institute of Higher Education and Research, Chennai, Tamil Nadu, India.
Abstract:
Skin disorders are a leading global health problem. The exact diagnosis of melanoma will help to improve patient outcomes. Methods of diagnosis, such as visual inspection by a dermatologist, are widely used for diagnosing various types of skin disease. However, they tend to suffer from experts' subjectivity. It is also time-consuming and limited by the availability of medical expertise. To overcome the problems, this research paper proposes an automated skin illness classification system using DL (Deep Learning) methods, which can assist in reliable and efficient diagnosis. The system suggested using a CNN to classify skin diseases from dermoscopic images. The HAM10000 dataset is an established, clinically verified dataset containing images of seven common pigmented skin lesion categories, used to train and test the model. Before training models, images are preprocessed using steps such as resizing, normalisation, colour normalisation, and noise reduction. A pre-trained CNN architecture has been used to extract visual features via TL (Transfer Learning), achieving good performance with limited medical data. The Adam optimiser trains the model with categorical cross-entropy loss. To evaluate robustness across all disease classes, accuracy, precision, recall, F1-score, and confusion matrix are used. The skin lesions are accurately classified, proving that the method works. The approach has great diagnostic potential for dermatologists. This can be expanded to online or mobile apps for early identification of skin disease and greater accessibility.
Keywords: Skin Disease; Deep Learning (DL); CNN Architecture; Medical Expertise; Noise Reduction; Dermoscopic Images; Colour Normalisation; Transfer Learning; Adam Optimiser.
Received on: 13/02/2025, Revised on: 07/06/2025, Accepted on: 12/09/2025, Published on: 05/01/2026
DOI: 10.64091/ATIHL.2026.000264
AVE Trends in Intelligent Health Letters, 2026 Vol. 3 No. 1 , Pages: 40-52