Prediction of chronic kidney disease using urinary dielectric properties and support vector machine

• Published in: The Journal of microwave power and electromagnetic energy Vol. 50; no. 3; pp. 201 - 213
• Main Authors: Mun, Peck Shen, Ting, Hua Nong, Mirhassani, Seyed Mostafa, Ong, Teng Aik, Wong, Chew Ming, Chong, Yip Boon
• Format: Journal Article
• Language: English
• Published: Taylor & Francis 02.07.2016
• Subjects: chronic kidney disease; classification; support vector machine; Urinary dielectric properties
• ISSN: 0832-7823
• DOI: 10.1080/08327823.2016.1230927

In this study, we aim to classify the urinary dielectric properties of subjects with chronic kidney disease (CKD) and normal subjects, at microwave frequency between 1 GHz and 50 GHz using support vector machine (SVM). The dielectric properties of urine were measured at room temperature (25°C), 30°C and body temperature (37°C). Urinary dielectric behaviour differences were observed between respective diabetic kidney disease (DKD) and non-DKD compared to normal subjects. Two-group classifications obtained the highest accuracy of 75.91% and 70.02%, respectively, in differentiating DKD and non-DKD group from normal group. The highest classification accuracy was achieved at 63.94% for three-group classifications. The best classification accuracies were obtained at 30°C for two-group and three-group classifications.