Diabetic Foot Ulcer Classification Models Using Artificial Intelligence and Machine Learning Techniques: Systematic...

Area of Science:

• Medical Informatics
• Computational Biology
• Diabetology

Background:

• Diabetic foot ulceration (DFU) is a significant diabetes complication impacting survival and quality of life.
• Current DFU classification systems lack individual prognostication capabilities.
• Machine learning (ML) models are proposed as a solution for improved DFU characterization and prediction.

Purpose of the Study:

• To systematically identify and evaluate evidence on ML models predicting clinical outcomes in DFU patients.
• To assess the prognostic accuracy and limitations of existing ML-based predictive models for DFU.

Main Methods:

• Comprehensive literature search across major databases (MEDLINE, Scopus, Web of Science, IEEE Xplore) up to July 2023.
• Inclusion of anterograde analytical studies with at least 80% adult DFU population.
• Independent literature screening, data extraction, and risk of bias assessment using established tools (QIPS, PROBAST).

Main Results:

• Eleven studies (13 papers) analyzed 55 ML models predicting wound healing, amputation, or mortality.
• Models predominantly used clinical data and random forest, with most achieving Area Under the Receiver Operating Characteristic Curve (AUROC) ≥ 0.8.
• All included studies exhibited a high risk of bias due to inconsistent definitions, small sample sizes, and inadequate data handling.

Conclusions:

• Several ML models demonstrate good discriminatory ability for DFU outcomes.
• Widespread clinical application is hindered by a focus on model development, lack of external validation, and use of non-explainable techniques.
• Future research should prioritize external validation and development of explainable ML models for DFU care.