Related Experiment Video
Updated: Jul 29, 2026
06:08
Quantitative Visualization and Detection of Skin Cancer Using Dynamic Thermal Imaging
Published on: May 5, 2011
16.9K
Diabetic foot thermal image segmentation using Double Encoder-ResUnet (DE-ResUnet).
Doha Bouallal1, Hassan Douzi1, Rachid Harba2
1IRF-SIC Laboratory, Ibn Zohr University, Agadir, Morocco.
Journal of Medical Engineering & Technology
|May 31, 2022
Summary
This study introduces a deep learning model for precise diabetic foot segmentation using thermal and RGB images. The DE-ResUnet achieves 97% IoU, improving early detection of ulceration risk areas.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Diabetology
Background:
- Diabetic Foot (DF) ulceration is a significant complication of diabetes.
- Early detection of DF requires precise segmentation of foot thermal images.
- Current segmentation methods lack accuracy and reliability for clinical use.
Purpose of the Study:
- To develop a fully automated, robust, and accurate segmentation method for the diabetic foot.
- To improve the identification of high-risk areas for ulceration in diabetic patients.
Main Methods:
- A novel deep neural network, Double Encoder-ResUnet (DE-ResUnet), was developed.
- The model fuses thermal and RGB (Red, Green, Blue) image data.
- A dataset of 398 image pairs from healthy and diabetic subjects was utilized.
Main Results:
- The DE-ResUnet model achieved an average Intersection over Union (IoU) of 97%.
- The model demonstrated superior performance compared to state-of-the-art methods.
- Accurate delineation of high-risk regions like toes and heels was achieved.
Conclusions:
- The proposed DE-ResUnet offers a robust and accurate solution for automated diabetic foot segmentation.
- This technology has the potential to enhance early diagnosis and prevention of diabetic foot ulcers.
