Development of a Mobile Application for Early Detection of Skin Cancer Using the EfficientNet-Lite Model an‎d Fuzzy C-Means Clustering, with Advanced Preprocessing an‎d Data Augmentation

9/15/2025 12:00:00 AM

Skin cancer is one of the most common types of cancer worldwide, an‎d its most dangerous form, melanoma, is highly treatable if detected early. This research aims to develop a lightweight an‎d efficient deep learning model for the classification of skin lesions, suitable for implementation on resource-constrained devices. To this end, a model based on the EfficientNet-Lite architecture was developed for feature extraction an‎d classification. The impact of preprocessing based on Fuzzy C-Means (FCM) clustering on the modelʹs performance was critically eva‎luated. The publicly available HAM10000 dataset was used for training an‎d eva‎luation. Advanced preprocessing techniques, including noise removal an‎d class imbalance han‎dling, were employed to improve performance. The primary goal of this research is to provide a methodology for building accurate an‎d accessible diagnostic models that can be utilized in initial screening tools. The obtained results emphasize the importance of comprehensive model eva‎luation an‎d the direct impact of the chosen preprocessing method on the final performance, particularly in sensitive medical applications.

تشخيص زودهنگام , سرطان پوست , يادگيري عميق , خوشه‌بندي فازي , مدل EfficientNet-Lite

Early Detection , Skin Cancer , Deep Learning , Fuzzy Clustering , EfficientNet-Lite Model

Alireza Lotfi Moghaddam

Nasser Mozayani