Image segmentation using transfer learning and Fast R-CNN for diabetic foot wound treatments

Area of Science:

• Medical imaging
• Artificial Intelligence in Healthcare
• Diabetic Complications

Background:

• Diabetic foot ulcers (DFUs) present complex challenges due to their multifactorial nature.
• Effective management requires systematic patient evaluation and comprehensive treatment planning.
• Current assessment methods, like the PEDIS index, are qualitative and physician-dependent.

Purpose of the Study:

• To develop and evaluate an image recognition system for diabetic foot wound assessment.
• To support clinicians in executing effective DFU treatment plans.
• To automate the classification, location, and size analysis of diabetic foot wounds.

Main Methods:

• Application of deep neural networks, convolutional neural networks, and object recognition.
• Utilizing image analysis technology for wound feature extraction and labeling.
• Employing the Object Detection Fast R-CNN method for machine learning model training and evaluation.

Main Results:

• The developed system accurately analyzes wound classification, location, and size.
• Machine learning models achieved high effectiveness in wound image detection.
• Assessment accuracy reached up to 90% for wound image detection data.

Conclusions:

• Image recognition technology offers a promising tool for objective DFU assessment.
• AI-driven analysis can enhance the accuracy and efficiency of DFU management.
• This approach supports the systematic evaluation and treatment of diabetic foot ulcers.