Development of AI classification model for angiosome-wise interpretive substantiation of plantar feet thermal asymmetry in type 2 diabetic subjects using infrared thermograms. (December 2022)

• Record Type: Journal Article
• Authors: Evangeline N, Christy; Srinivasan, S.; Suresh, E.
• Is Part Of: Journal of thermal biology. Volume 110(2023)
• Publication Date: 2022-12
• Subjects: Artificial intelligence -- Diabetic foot syndrome -- Feature extraction -- Hotspot identification -- Infrared thermogram; Thermobiology -- Periodicals; Temperature -- Periodicals; Biology -- Periodicals; Thermobiologie -- Périodiques; Thermobiology; Periodicals; 571.46
• DOI: 10.1016/j.jtherbio.2022.103370
• Languages: English
• ISSNs: 0306-4565

Abstract

Abstract: Diabetic Foot Syndrome (DFS) is the prime impetus for most of the lower extremity complications among the diabetic subjects. DFS is characterized by aberrant variations in plantar foot temperature distribution while healthy subjects exhibit a symmetric thermal pattern between the contralateral and ipsilateral plantar feet. Thus, "asymmetry analysis" of foot thermal distribution is contributory in assessment of overall foot health of diabetic subjects. The study, aims to classify symmetric and asymmetric foot regions angiosome-wise, by comparing minimal number of color image features - color moments and Dissimilarity Index. Further, the asymmetric foot regions are assessed for identifying the hotspots within such angiosomes of the patients that characterize the possibility of onset of diabetic foot ulcer. The color feature based machine learning model developed, achieved an accuracy of 98% for a 10-fold cross validation, test accuracy of 96.07% and 0.96 F1-score thereby convincing that the chosen features are amplest and conducive in the asymmetry analysis. The developed model was validated for generalization by testing on a public benchmark dataset, in which the model achieved 92.5% accuracy and 0.91 F1 score. Highlights: Diabetic Foot ulcers may end up in infections and amputations in the lower limb. Asymmetry analysis helps to identify the regions of unusual temperature in the foot. The work confers minimal number of features that are amplest for asymmetry analysis. The study uses machine learning models to abet classification of at risk regions. The model is validated on public benchmark database for generalization.