CNN-Based Multi-Class Classification of Fungal, Scabies, and Allergic Skin Diseases Using Image Processing

Abstract

Neglected Tropical Diseases (NTDs) affect over a billion people globally, with skin conditions like fungal infections, scabies, and allergies often overlooked due to overlapping symptoms and limited diagnostic resources. To address this, we propose a CNN-based multi-class classification model using image processing techniques to distinguish six skin disease classes: tinea, candidiasis, pityriasis versicolor, scabies, contact dermatitis, and eczema. A dataset of 300 images was curated from DermNet, a credible dermatology resource, and preprocessed via normalization, augmentation, and batch-wise training. The designed CNN architecture achieved 93% testing accuracy, with 92% precision, 95% recall, and 93% F1-score, significantly outperforming benchmark models. By integrating image processing (e.g., noise reduction, flipping) with a 10- layer CNN framework, the model mitigates challenges posed by symptom similarity and dataset limitations. This work aligns with the WHO’s 2030 NTD roadmap by offering a scalable tool for early detection and reduced transmission of neglected skin diseases.