Related Experiment Video
Updated: Jul 3, 2025
06:08
Quantitative Visualization and Detection of Skin Cancer Using Dynamic Thermal Imaging
Published on: May 5, 2011
16.8K
Skin Cancer Segmentation and Classification Using Vision Transformer for Automatic Analysis in Dermatoscopy-Based
Galib Muhammad Shahriar Himel1, Md Masudul Islam1, Kh Abdullah Al-Aff2
1Jahangirnagar University, Dhaka, Bangladesh.
International Journal of Biomedical Imaging
|February 12, 2024
Summary
This study introduces a vision transformer model for accurate skin cancer classification, achieving 96.15% accuracy. This deep learning approach aids dermatologists in early skin cancer diagnosis.
Area of Science:
- Dermatology
- Computer Science
- Artificial Intelligence
Background:
- Skin cancer is a global health issue requiring early diagnosis for better patient outcomes.
- Deep learning models have shown promise in image classification tasks, including medical imaging.
Purpose of the Study:
- To develop and evaluate a vision transformer-based approach for skin cancer classification.
- To assess the effectiveness of the vision transformer model in distinguishing between benign and malignant skin lesions.
Main Methods:
- Utilized the HAM10000 dataset of 10,015 skin lesion images.
- Applied preprocessing techniques including normalization and augmentation.
- Employed the Segment Anything Model (SAM) for image segmentation and various pre-trained vision transformer models for classification.
Main Results:
- The vision transformer model demonstrated superior performance compared to traditional deep learning architectures.
- Google's ViT patch-32 model achieved 96.15% accuracy on the test dataset.
- The model exhibited a low false negative ratio, indicating high diagnostic reliability.
Conclusions:
- Vision transformer models show significant potential as effective tools for aiding dermatologists in skin cancer diagnosis.
- The proposed approach offers a promising advancement in automated skin lesion analysis.
- Further research can explore integrating this model into clinical diagnostic workflows.
