Development of Mobile Application for Skin Cancer Detection Using Deep Learning Models

Skin cancer remains the leading cause of cancer-related mortality worldwide, emphasizing the critical need for accurate and timely diagnosis. The current clinical treatments are time-consuming and susceptible to human error, which can delay treatment and affect patient outcomes. This research addresses these challenges by making two major contributions to the field. Firstly, four fine-tuned deep neural models; MobileNetV2, MobileNetV3, EfficientNetV2, and ResNet50; were evaluated across three distinct experimental settings to explore their potential in accurately classifying skin lesions. Among the models tested, MobileNetV3 outperformed the others, achieving the highest accuracy of 80.17% in categorizing four types of skin cancer in dermoscopic images and 79.17% in clinical images. Secondly, the study indicated that the development of an accessible mobile application capable of detecting skin cancer at its early stages is crucial. This app leverages the MobileNetV3 model to provide a quick and reliable tool for early diagnosis, thereby broadening access to life-saving interventions and reducing the burden of treatments on healthcare systems.