Modeling Skin Cancer Detection and Analysis using Improved YOLOv5 Towards Early Diagnosis and Clinical Applications

Skin cancer is among the most common malignant tumors, often diagnosed at advanced stages due to subtle symptoms and limited public awareness. To address this challenge, this paper presents an improved YOLOv5-based framework for skin cancer detection, incorporating SimAM and SEAttention mechanisms to enhance feature extraction capabilities. The proposed model adopts a lightweight design suitable for deployment on resource-constrained devices, thereby improving accessibility for non-expert users. Initial experimental results demonstrate that, compared with the original YOLOv5, the proposed model achieves a 0.2 GFLOPs reduction, 1,312 fewer parameters, and a <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{1 0 \%}-\mathbf{1 5 \%}$</tex> improvement in detection accuracy. The enhanced capacity for fine-grained lesion analysis highlights its potential for early diagnosis and clinical application.