A Skin Cancer Detection System Using Wavelet-Based Feature Extraction and Deep Learning: A Simulation-Based Approach

Abstract

Skin cancer is one of the most prevalent and potentially life-threatening diseases, where early diagnosis is critical for effective treatment. This work presents a real-time embedded skin cancer detection system using wavelet-assisted deep learning, implemented on a Python based platform. A camera module captures dermoscopic skin lesion images, which are processed locally on the Python based to enable portable and low-cost diagnosis. The acquired images undergo preprocessing for noise reduction, resizing, and illumination normalization. Wavelet transform–based feature extraction is then applied to enhance texture, edge, and color characteristics of lesions. The enhanced features are fed into a deep learning model that classifies skin lesions into malignant cancers such as Melanoma and Basal Cell Carcinoma, as well as benign conditions including Nevus, Seborrheic Keratosis, Actinic Keratosis, and Pigmented Benign Keratosis. The diagnostic outcome is displayed in real time on an output screen, indicating the presence or absence of skin cancer. Experimental results demonstrate that the proposed system achieves high classification confidence, with Melanoma detected at an average confidence of 96%, Basal Cell Carcinoma at 93%, and benign lesion classes at approximately 94%. Overall, the system maintains detection confidence above 90% across all tested categories, confirming its robustness and reliability. The proposed approach offers a portable, cost-effective, and accurate solution for early skin cancer screening and remote healthcare applications, highlighting the effectiveness of combining wavelet-based feature enhancement with embedded deep learning.