Integration of Feature Fusion Strategy on Efficient net for Skin Cancer Detection and Classification

The increase in prevalence over the last decade has made skin cancer one of the top 10 most common cancers. The investigation of skin cancer lesions is challenging because of their complex expression, comparatively similar forms, and exposure to subjective diagnosis. The primary aim of the study is to explore human skin cancer by combining a feature fusion technique on EfficientNet B7. This study integrates a feature fusion with EfficientNet B7 for enhanced human skin cancer classification. We have preprocessed and augmented skin lesions, then extracted and fused features from multiple layers of EfficientNet B7. The study’s results demonstrate strong performance in skin cancer classification, achieving an accuracy of 93.4%, with precision at 92.3%, recall at 94.8%, specificity at 92.1%, and an F1-score of 93.5%. Overall, these results underscore the efficacy of the feature fusion plus EfficientNet B7 in delivering reliable classification results. As a result, dermatologists can inspect and classify the type of skin cancer in less than a minute with more accuracy and efficient decision-making thanks to the proposed study, which also helps to shorten the time it takes to diagnose the disease.