#4110 BIO-CONDUCTIVITY MEASUREMENTS ENABLE PORTABLE AND SELF-ADMINISTRATED CHRONIC KIDNEY DISEASE SCREENING AND MONITORING

Abstract

Abstract Background and Aims Chronic kidney disease (CKD) is an escalating health concern. The identification of CKD during mass public screening by laboratory tests constitutes important resource implications. Home-based and self-administrated approach to screen and monitor disease severity in both healthy and CKD subjects are eagerly awaited. Frequency-difference electrical impedance tomography (fdEIT) reconstructs the interior of a body by measuring the electrical responses to a small alternating current applied at the surface of the subject at various frequencies. Kidney fibrosis is a cardinal feature of CKD, which is associated with alterations in renal molecular composition and hence electrical conductivity. Therefore, this study aims to investigate the feasibility of a portable, self-administrated approach to assess CKD severity using fdEIT in the kidney region. Method Clinical subjects are recruited at Queen Mary Hospital, Hong Kong. EIT data was collected via a PVC belt acting as an electrode holder placed circumferentially on the upper abdominal region of the subject. The belt was connected to a portable console to collect EIT data. 24 frequencies ranging from 10KHz to 300KHz are used to stimulate electrical responses in the body. (Figure 1A) Paired blood and urine samples were collected for measured of eGFR. Group source separation was implemented to extract the kidney region of interest (ROI) and extract conductivity features (Figure 1A); these features, together with the age of the subject, are then input into a regression model to estimate the eGFR and the CKD stage of the subject according to the following classification scheme: stage 1 CKD (eGFR > 90) as healthy subjects, stages 3, 4, 5 CKD (eGFR < 60) as unhealthy subjects, stage 2 CKD (60 < eGFR < 90) as borderline cases. Results 75 subjects were recruited (54 with CKD and 21 were healthy volunteers). We obtained an eGFR estimation model with R2 score of 0.40. We also obtained a CKD stage classifier with sensitivity of 87.5% and specificity of more than 99.9%. (Figure 1C) The mean conductivity in the extracted kidney signal comprises 40% weighting in the regression model (Figure 1B), showing comparable importance as the age in predicting the eGFR. Conclusion The results demonstrate that measuring bio-conductivity anomalies through fdEIT is highly accurate and non-invasive, and can be developed into a portable and self-administrable device to screen and monitor CKD