Detecting Diabetic Foot Complications using Infrared Thermography and Machine Learning

One in every ten adults suffers from diabetes. Diabetes Mellitus can cause foot ulcers which can lead to leg or foot amputation. Early identification depends on day-to-day risk assessment for diabetic patients. Healthy feet have an even temperature distribution while diabetic feet tend to have abnormal regions with higher temperatures. In our approach the abnormal regions of the foot plantar thermograms were segmented with lazy snapping. First Order Statistical features, GLCM based textural features and Wavelets based features were extracted from these regions of interest. Various classifiers were used to evaluate the proposed features. We concluded that the proposed features were able to discriminate normal and diabetic foot with an accuracy of 97.778%.