Web Application for Screening Malignant Melanoma in Digital Images Using Deep Learning
Kharittha Jangsamsi, Kasidet Samosorn, Puttimate Hirunuran, Prapawin Sakdapetchsiri, Wiriya Mahikul, Rithee Smithrithee, Krisana Chinnasarn · 2024
Malignant Melanoma (MM) is a highly lethal skin cancer and detecting it early is crucial, as it has a mortality rate of 48.08%. To address this issue, a web application was developed using deep learning models and image processing techniques. The application serves as a self-screening tool for individuals who are suspicious of any skin lesions. The study evaluated three deep learning models, namely AlexNet, MobileNet-V2, and ResNet-18, by using a comprehensive dataset from the MED-NODE dermatology database. The results showed that the ResNet-18 model, when combined with other approaches, demonstrated superior performance, with 86.11% accuracy, 88.10% precision, 88.10% sensitivity, and 83.33% specificity. The research highlights the potential of deep learning models and median filters in image processing techniques to facilitate effective melanoma risk screening, promote early detection, and improve patient outcomes.