Analysis of T Wave Nonlinear Dynamics for Serum Potassium Monitoring in End-Stage Renal Disease Patients Sabarathinam Srinivasan 1 , Hassaan A Bukhari 1,2 , Pablo Laguna 1,3 , Carlos S´ anchez 1,4 , Esther Pueyo 1,3 1 I3A, University of Zaragoza, IIS Arag´ on, Zaragoza, Spain 2 University of Bordeaux, IMB, UMR 5251, Talence, France 3 CIBER en Bioingenier´ ıa, Biomateriales y Nanomedicina, Spain 4 Defence University Centre (CUD), AGM, Zaragoza, Spain Abstract Non-invasive estimation of serum potassium, [K + ], is of major importance to prevent associated risks, but current ambulatory estimation methods are limited. We investigated changes in T wave nonlinear dynamics by quantifying a divergence-related marker ψ on electrocardiograms (ECGs) from 15 end-stage renal disease (ESRD) patients undergoing hemodialysis (HD) and we assessed the relationship between ψ and [K + ]. ECGs from 22 simulated transmural ventricular fibers were additionally calculated. In ESRD patients, ψ took the largest values at the beginning and end of the HD session, corresponding to the highest and lowest [K + ] values. The median correlation coefficient over patients between the change in ψ and the change in [K + ] was 0.92 and decreased to 0.74 after controlling for the effects of [Ca 2+ ] and heart rate. These associations were, however, highly patient-dependent. Both the strength and variability of the ψ-[K + ] relationship was reproduced in the simulations, with the variability explained by differences in transmural heterogeneities: 10% variations in the proportion of epicardial and midmyocardial cells led to more than 10% and 8% changes in ψ, respectively. In conclusion, changes in the nonlinear dynamics of the ECG T waves can be related to [K + ] variations in ESRD patients, despite the high inter-individual variability. 1. Introduction Around 10% of the worldwide population are affected by chronic kidney disease. The economic cost, increased mortality risk and decreased quality of life associated with this disease are remarkably high [1]. At the final stage, so called end-stage renal disease (ESRD), patients present an increasingly impaired ability to maintain potassium homeostasis. Serum potassium levels ([K + ]) outside normal ranges, in the form of hypokalemia or hyperkalemia, increase the risk for life-threatening arrhythmias and sudden cardiac death [2]. Investigating the effects of [K + ] variations on the electrical activity of the heart could allow deriving electrocardiogram (ECG)-based markers for continuous monitoring of [K + ], which would facilitate the delivery of more timely therapies for ESRD patients. Variations in [K + ] levels have been indeed shown to affect depolarization and repolarization features of the ECG [2]. Most of the investigated ECG markers in the literature are aimed at characterizing T wave properties [3]. However, such properties commonly refer to a specific time point or small portion of the T wave and are, thus, not necessarily representative of the whole T wave morphology changes in response to varying [K + ]. In previous studies, we characterized changes in the T wave shape at varying [K + ] by using time-warping techniques over clinical and simulated ECGs [4, 5]. The linear but also the nonlinear components of the measured warping markers showed a strong relationship with [K + ]. Based on those results, we hypothesized that evaluation of nonlinear dynamics of the T wave, as measured by other techniques quantifying the divergence of trajectories starting close to each other, could provide relevant information for [K + ] estimation. We quantified a divergence-related marker ψ over ECGs of ESRD patients during a hemodialysis (HD) session, where large [K + ] changes are expected to occur. Also, we simulated a set of ventricular fibers covering a wide range of transmural heterogeneities and we calculated pseudo-ECGs (pECGs) at different [K + ] levels. After confirming that simulated ψ reproduced the evidences from clinical ECGs, we performed a sensitivity analysis to assess the contribution of transmural heterogeneities to inter-individual differences in T wave nonlinear dynamics, particularly for [K + ] values out of normal ranges. Computing in Cardiology 2020; Vol 47 Page 1 ISSN: 2325-887X DOI: 10.22489/CinC.2020.461