Skin Cancer Detection using Image Enhancement Techniques and Resnet50 from Dermoscopy Images

Automated Skin Cancer Detection (ASCD) systems leveraging Artificial Intelligence (AI) and Deep Learning (DL) have emerged as transformative solutions. These technologies offer potential for rapid, accurate, and cost-effective diagnosis, significantly reducing the dependency on invasive procedures like biopsies. Improvements in accuracy and efficiency in melanoma diagnosis are shown in the research using latest DL models, such as Resnet50, together with image enhancing approaches. The overarching goal of this study is to provide the groundwork for more accurate computer-assisted skin cancer diagnostics. On the ISIC 2018 dataset, the research achieved a segmentation accuracy of 98.7% by employing the segmentation phase for image segmentation. Using the Skin Cancer Classification Phase model, which outperforms existing methods, the research achieves a classification accuracy of 98.8% on the ISIC 2018 dataset. By tackling problems with low contrast and noisy datasets, image enhancement algorithms improved segmentation accuracy and provided good results. The results demonstrated the value of combining latest DL models with preprocessing methods, leading the way towards a clinically applicable, fully ASCD system.