Abstract:
"Diabetic foot ulcers (DFUs) are becoming more common and present serious problems for patients
and healthcare systems around the world. Open wounds known as DFUs are typically found on
the lower extremities of people with diabetes. These wounds can cause serious complications,
including gangrene, infections, and the need to remove a lower limb. Peripheral neuropathy, poor
circulation, and abnormalities in the structure of the foot are the causes of these ulcers, highlighting
the urgent need for preventive measures and individualized foot care interventions.
A novel solution utilizing thermal imaging and ML technologies was developed to address the
problem of DFUs. We trained a model to analyze thermal images of diabetic patients' feet and
identify foot temperature distribution as indicators of high-risk areas of DFUs by using
ML algorithms, specifically CNNs. Furthermore, we incorporated machine learning-based
decision support systems into the current healthcare workflows, enabling physicians to make wellinformed choices based on instantaneous thermal image analysis.
Several data science metrics were utilized to evaluate the performance of our machine learning
model for DFU detection. The model's ability to accurately classify DFU cases and non-DFU cases
was revealed by these metrics, which included accuracy, precision, recall, and F1-score. In
addition, we computed the AUC and examined the ROC curve, which gave us a thorough
evaluation of the model's ability to discriminate across various thresholds. We proved the model's
adaptability and its potential to be a useful tool in early DFU detection and preventive care
strategies through extensive testing and validation procedures using clinical datasets."
