Vitiligo Detection Using Machine Learning

Abstract

Skin disorders are widespread worldwide, encompassing various conditions such as skin cancer, vulgaris, ichthyosis, and eczema. Among these, vitiligo stands out as it can appear anywhere on the body, including the oral cavity, and significantly affect overall Heath, leading to cognitive issues, hypertension, and mental health problems. Traditional diagnostic methods employed by dermatologists, like biopsy, blood tests, and patch testing, have limitations, particularly in cases where lesions progress from macules to patches. To address this, machine learning (ML) and deep learning (DL) models have emerged to expedite diagnosis. This research introduces a DL-based model specifically designed for predicting and categorizing vitiligo in healthy skin. Leveraging a pre-trained Inception V3 model, image features are extracted and utilized alongside classifiers such as naive Bayes, convolutional neural network (CNN), random forest, and decision tree. Evaluation metrics including accuracy, recall, precision, area under the curve (AUC), and F1-score are employed. Results show that Inception V3 coupled with naive Bayes achieves high accuracy, recall, precision, AUC, and F1-score values of 99.5%, 0.995, 0.995, 0.997, and 0.995, respectively. Inception V3 with CNN achieves even higher accuracy at 99.8%, along with impressive recall, precision, AUC, and F1-score values. Similarly, Inception V3 paired with random forest exhibits exceptional performance across all metrics, achieving 99.9% accuracy, 0.999 recall, 0.999 precision, 1.00 AUC, and 0.999 F1-score. Although Inception V3 combined with the decision tree classifier shows slightly lower performance, it still achieves respectable results. Notably, Inception V3 coupled with random forest demonstrates superior performance across most metrics, with both Inception V3 models achieving identical AUC outcomes of 1.00, indicating excellent predictive capability.