Potassium Monitoring From Multilead T-wave Morphology Changes During Hemodyalisis: Periodic Versus Principal Component Analysis

Abstract

Background: End-stage renal disease (ESRD) patients undergoing hemodyalisis therapy (HD) experience blood potassium ([K+]) variations that are reflected on the T-wave (TW) morphology. Methods: We evaluated the performance of different lead space reduction (LSR) methods: principal component analysis (PCA), maximising the TW energy, and two derived versions of periodic component analysis (πCA) named πCAB and πCAT, maximising the QRST or TW beat periodicity. We applied these methods to 12-lead electrocardiogram (ECG) from 24 ESRD-HD patients. Then, we derived three markers of TW morphology changes (duw, dw and d^w,c), comparing an average TW derived every 30 min with that at the HD end, from the PCA, πCAB and πCAT based leads having the highest TW energy content. Similarities between these three methods were assessed by using Bland-Altman plots and the linear fitting error (∊) evaluated from the 12th to the 44th h of ECG recordings after the HD onset. Results: All series of duw, dw and d^w,c values showed good degree of mutual agreement (median bias ≤. 0.5 ms) and a small deviation from linearity in the [K+] increasing stage (median ∊ ≤ 3.3 ms). Conclusions: PCA andπCA can be used interchangeably to track TW changes in ESRD-HD patients, in this type of low noise contamination ECG recordings.