AI-Enhanced Imaging for Diabetic Foot Ulcer Risk Assessment and Diagnosis: A Retrospective Cohort Study

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Diabetic foot ulcer (DFU) assessment requires precise segmentation for accurate severity prediction.
• Existing segmentation tools often lack the necessary accuracy for effective wound boundary delineation.
• Wagner's grading system is a standard for classifying DFU severity.

Purpose of the Study:

• To develop and evaluate an artificial intelligence (AI)-driven method for DFU segmentation and severity classification.
• To improve the accuracy of DFU boundary delineation compared to conventional methods.
• To enable real-time, remote assessment of DFU severity using AI.

Main Methods:

• A novel AI-driven method combining an enhanced active contour model with Sobel edge detection for precise DFU boundary segmentation.
• Development of a lightweight classification model for predicting DFU severity based on Wagner's grading system.
• Utilized a retrospective cohort of 1339 DFU images (augmented to 6579) from 510 patients in India.

Main Results:

• The segmentation approach achieved a high Dice similarity coefficient of 0.99.
• The classification model demonstrated 95.58% accuracy, 95.58% sensitivity, and 99.16% specificity.
• The method exhibited a low false-positive rate (0.84%) and false-negative rate (4.83%), outperforming existing techniques.

Conclusions:

• The proposed AI method significantly enhances both segmentation and classification of diabetic foot ulcers.
• This advancement supports improved clinical decision-making and management of DFU.
• The developed AI-powered mobile application facilitates efficient remote DFU assessment.