Development of a decision support system to detect insulin-inducedlipohypertrophy in diabetic individuals

Development of a Decision Support System to Detect Insulin-Induced Lipohypertrophy in Diabetic Individuals INTRODUCTION AND AIM: This study aimed to develop a decision support system for detecting insulin-induced lipohypertrophy in individuals with diabetes (DM) and to evaluate the effectiveness of this system. MATERIALS AND METHODS: The sample of this methodological and randomized controlled study consisted of 249 individuals with DM. In light of the data obtained from the DM individuals via the "Patient Information Form", the abdominal regions of the individuals with DM included in the study were scanned for lipohypertrophy (LH) by an expert radiologist using an ultrasound device. The scan results detected LH in 130 individuals with DM, while no LH findings were found in 119 patients. SpO2, PPG, BIOZ, and temperature measurements were made on the LH region and a healthy quadrant of the patients with LH. The measurement was made only on the healthy quadrant in individuals with no LH detection. The obtained data were analyzed using two different computer programs in two computer environments. During the data evaluation, frequency distribution, independent sample t-test, chi-square analysis, logistic regression tests, and machine learning algorithms were applied. 70% of the obtained data was used in training and 30% in testing the data. RESULTS: The factors affecting LH formation were determined with logistic regression, such as rotation, needle length, and body fat ratio. As a result of the machine learning algorithms, the highest accuracy of 78.6% was obtained with the chi-square algorithm and PNN, SVM, and Ensemble classifiers using the six features most associated with the presence of LH. CONCLUSION: The developed decision support system is a valid and reliable method for detecting insulin-induced lipohypertrophy. It can be used in the clinic for LH detection by health professionals, especially diabetes nurses. Keywords: Diabetes, Nurse, Decision Support System, Lipohypertrophy, Machine Learning, Artificial Intelligence