Machine learning algorithm to evaluate risk factors of diabetic foot ulcers and its severity

Area of Science:

• Biomedical Informatics
• Computational Biology
• Diabetology

Background:

• Diabetic foot ulcers (DFU) pose a significant challenge in diabetes management.
• Early identification of risk factors is crucial for preventing DFU development and progression.
• Machine learning offers potential for developing predictive models in healthcare.

Purpose of the Study:

• To identify clinical and biochemical risk factors associated with DFU.
• To develop and compare machine learning algorithms for DFU prediction.
• To classify DFU subtypes using computational methods.

Main Methods:

• An observational study included 80 patients with type 2 diabetes mellitus (T2DM) with DFU and 80 without DFU.
• Machine learning algorithms including SVM, Naive Bayes, KNN, Random Forest, and ensemble learners (Stacking C, Bagging, AdaBoost) were employed.
• Models were developed for two stages: DFU presence/absence and ulcer type classification.

Main Results:

• Ensemble learning methods demonstrated superior performance over individual classifiers.
• Novel risk factors, ApoA1 and IL-10, were identified for DFU development.
• IL-10 and uric acid effectively discriminated ulcer severity grades.
• Stacking C algorithm achieved enhanced prediction accuracy for both classification stages.

Conclusions:

• Machine learning, particularly ensemble methods like Stacking C, can significantly enhance DFU prediction and classification accuracy.
• Identification of new risk factors (ApoA1, IL-10) advances understanding of DFU pathogenesis.
• Computational screening using these models can serve as a complementary tool for managing DFU and its subtypes.