Application of non-contact thermography as a screening modality for Diabetic Foot Syndrome – A real time cross sectional research outcome

Abstract

ObjectiveSubjects with uncontrolled diabetes for a prolonged span are susceptible to develop comorbidities like Peripheral Arterial Disease (PAD) and Diabetic Peripheral Neuropathy (DPN) – collectively the Diabetic Foot Syndrome (DFS), which may lead to ulcers, infections and amputations in the lower limbs. Thermal images of plantar feet are indicative of these conditions and can aid in understanding the overall foot health in such subjects. MethodsThis study aims to classify the thermal distribution patterns in feet of diabetic subjects versus normals, patterns in diabetic subjects into those having DFS versus not having DFS and patterns in DFS subjects into those with PAD versus DPN. The classifications were based machine learning techniques using plantar foot temperature features. ResultsThe thermal patterns were classified with cross validation accuracies of 98.89%, 95.2% and 97.1% respectively. The study compares the classifier performances prior-to and post thermal data augmentation to ruled-out type-1-error and accuracy paradox - practical issues faced in developing clinical decision support systems. Further, an empirical model was developed which showed to be a significant appraisal in mass screening of subjects. ConclusionThus the proposed model is an all-inclusive intelligent system using non-contact imaging modality that needs no external radiation and also, performs automatic classifications with just a single technique instead of multiple conventional methods. SignificanceAs a distinctive contribution, foot thermal pattern amongst normals were captured and presented as a polynomial model and the study proved the chosen features to be befitting in analysis DFS using artificial Intelligence.