Synergistic Feature Fusion for Accurate Skin Cancer Classification

Abstract

Skin cancer is considered one of the most dangerous types of cancer caused by damaged DNA, and it can be life-threatening. The abnormal DNA leads to uncontrolled cell growth, and the cancer can spread rapidly. Analyzing skin lesion images for cancer detection is challenging due to various factors such as light reflections, color variations, difficulty in identification and diagnosis, varying sizes and shapes of lesions, and the similarity between different skin diseases like melanoma and nevus. Automatic recognition of skin cancer can be beneficial in improving the accuracy and efficiency of pathologists in early detection. The proposed approach involves several steps to enhance the classification accuracy.First, the input images are normalized to account for variations in lighting and other factors. Then, features are extracted from the normalized images to aid in precise classification. Finally, feature fusion techniques are employed to improve the overall classification accuracy.In this investigation, models such as AlexNet, VGG-19, and VGG-16 were utilized. Compared to existing models, the results of the suggested model indicate higher reliability and robustness.By employing normalization, feature extraction, and feature fusion techniques, the proposed model aims to provide accurate and trustworthy skin cancer classification. To evaluate the performance of the proposed concept the Skin CancerInternational Skin Imaging Collaboration's (ISIC) dataset was used to test the model, and a testing accuracy of 86.8% was achieved. This can contribute to the early detection of skin cancer, ultimately improving patient outcomes and assisting pathologists in their diagnostic process.