Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

An End-to-End Deep Learning Framework for Arsenicosis Diagnosis Using Mobile-Captured Skin Images

Created by
  • Haebom

Author

Asif Newaz, Asif Ur Rahman Adib, Rajit Sahil, Mashfique Mehzad

Outline

This paper proposes a mobile-based image analysis framework for the early diagnosis of arsenicosis, a serious public health problem in South and Southeast Asia. We constructed a dataset of arsenic-induced skin diseases and other skin diseases containing over 11,000 images and compared and evaluated CNN- and Transformer-based models. The Swin Transformer model achieved the highest performance, achieving 86% accuracy. The model's interpretability was enhanced using LIME and Grad-CAM. The practical applicability was also demonstrated through a web-based diagnostic tool. This study demonstrates the potential of non-invasive, accessible, and explainable arsenicosis diagnosis using mobile images.

Takeaways, Limitations

•
Takeaways:
◦
Possibility of developing a mobile image-based early diagnosis system for arsenicosis
◦
Verifying the excellent performance of Transformer-based models and demonstrating their practicality through web-based tools.
◦
LIME and Grad-CAM enhance model interpretability, ensuring clinical transparency and supporting error analysis.
◦
Potential for early diagnosis and appropriate intervention in areas with low access to healthcare
•
Limitations:
◦
Further research is needed on the size and diversity of the dataset.
◦
Performance validation and long-term stability evaluation in actual clinical environments are required.
◦
Further research is needed to determine generalizability across different skin tones and disease types.
👍
Made with Slashpage