SkinSAM: Empowering Skin Cancer Segmentation with Segment Anything Model

TL;DR

This paper introduces SkinSAM, a fine-tuned adaptation of the Segment Anything Model, achieving high accuracy in skin cancer segmentation on dermatoscopic images, demonstrating its potential for medical diagnostics.

Contribution

The paper presents SkinSAM, a fine-tuned SAM-based model specifically adapted for skin cancer segmentation, with superior performance on the HAM10000 dataset.

Findings

Fine-tuned SkinSAM achieved a pixel accuracy of 0.945.

The model obtained a dice score of 0.8879.

Vascular lesions had the best segmentation results.

Abstract

Skin cancer is a prevalent and potentially fatal disease that requires accurate and efficient diagnosis and treatment. Although manual tracing is the current standard in clinics, automated tools are desired to reduce human labor and improve accuracy. However, developing such tools is challenging due to the highly variable appearance of skin cancers and complex objects in the background. In this paper, we present SkinSAM, a fine-tuned model based on the Segment Anything Model that showed outstanding segmentation performance. The models are validated on HAM10000 dataset which includes 10015 dermatoscopic images. While larger models (ViT_L, ViT_H) performed better than the smaller one (ViT_b), the finetuned model (ViT_b_finetuned) exhibited the greatest improvement, with a Mean pixel accuracy of 0.945, Mean dice score of 0.8879, and Mean IoU score of 0.7843. Among the lesion types, vascular lesions showed the best segmentation results. Our research demonstrates the great potential of adapting SAM to medical image segmentation tasks.