A Convolutional Neural Network based system for classifying malignant and benign skin lesions using mobile-device images

• DOI: 10.1101/2023.12.06.23299413
• Publication Date: 2023-12-07T07:25:25Z
• Authors: Rim Mhedbi; Hannah O. Chan; Peter Credico; Rakesh Joshi; Joshua N. Wong; Collin Hong
• Keywords: Skin lesion; Actinic keratosis

Abstract The escalating incidence of skin lesions, coupled with a scarcity dermatologists and the intricate nature diagnostic procedures, has resulted in prolonged waiting periods. Consequently, morbidity mortality rates stemming from untreated cancerous lesions have witnessed an upward trend. To address this issue, we propose lesion classification model that leverages efficient net B7 Convolutional Neural Network (CNN) architecture, enabling early screening based on camera images. is trained diverse dataset encompassing eight distinct classes: Basal Cell Carcinoma (BCC), Squamous (SCC), Melanoma (MEL), Dysplastic Nevi (DN), Benign Keratosis-Like (BKL), Melanocytic (NV), ‘Other’ class. Through multiple iterations data preprocessing, as well comprehensive error analysis, achieves remarkable accuracy rate 87%.