Non-invasive method to analyse the risk of developing diabetic foot

Area of Science:

• Medical Informatics
• Public Health
• Diabetes Management

Background:

• Diabetic foot complications are a leading cause of lower extremity amputations and significant healthcare costs.
• Diabetic foot ulcers often precede amputations, highlighting the need for preventative strategies.
• Early identification of high-risk individuals is crucial for effective diabetes mellitus management.

Purpose of the Study:

• To propose an automatic, non-invasive method for identifying diabetic patients at high risk of developing diabetic foot complications.
• To develop a triage tool for basic health units to assist healthcare teams in managing diabetic foot risk.

Main Methods:

• Utilized K-means clustering algorithm with social scope and self-care data from 153 diabetic patients.
• Classified patients into high-risk and low-risk groups based on Euclidean distance to cluster centroids.
• Evaluated the method using both simulated and real-world data, with specialist classification as the gold standard.

Main Results:

• Achieved high accuracy (0.97 ± 0.06) with simulated data.
• Demonstrated promising accuracy (0.68 ± 0.16) with real data when compared to specialist classifications.
• The method requires simple computational processing, making it suitable for resource-limited settings.

Conclusions:

• The proposed non-invasive method effectively identifies patients at high risk for diabetic foot complications.
• This tool can assist basic health units in triaging patients, potentially reducing the incidence of diabetic foot and amputations.
• Simple computational requirements facilitate widespread adoption in primary healthcare settings for improved diabetes care.