Methods for signal processing and interpolation of body surface ECG signals and the impact on the inverse reconstructed cardiac electrical activity

Abstract

Non-invasive electrocardiographic imaging has been developed to overcome the challenges and limitations of the standard 12-lead ECG and invasive cardiac mapping. It utilizes high density body surface ECG signals and patient specific heart-torso geometry to reconstruct cardiac electrical activity. Body surface ECG signals are contaminated with different types of noise such as baseline wander, line frequency
noise, and high frequency artifacts that impact the inversely reconstructed cardiac electrical activity. In addition, inevitable missing electrodes due to electrode disconnection, applying defibrillation pads, and/or applying CARTO electrode patches affect the inverse reconstruction of cardiac electrical activity. This thesis aims to study the impact of different filtering techniques that comply with international recommendations for ECG diagnostic and medical devices on the inverse
reconstruction of cardiac electrical activity. In addition, the impact of using different
interpolation methods including the hybrid method that is developed in this thesis on
the reconstructed cardiac electrical activity is evaluated. Furthermore, this thesis compares between different methods of estimating the activation time including the temporal and spatiotemporal methods from the reconstructed electrograms.
Although there was no substantial difference in the inverse solution between the filter that has a bandwidth of 0.5 – 150 and notched at 50 Hz and the filter with a bandwidth of 0.5 – 40 Hz, it is recommended to use the former because the latter resulted in smoother reconstructed electrograms and a shifted activation time map.
The developed hybrid interpolation method estimated the missing data in body surface
ECG signals better than other interpolation methods. The ECG median CC values were
0.992, 0.979, and 0.977 for the hybrid interpolation, Laplacian interpolation, and PCA
interpolation respectively. However, the Inverse forward interpolation performed best when the reconstructed cardiac electrical activity was considered for comparing the performance especially when the missing electrodes has low peak-to-peak amplitudes.
The median CCEGM increases from 0.619 without interpolation, to 0.656 when
interpolating 11 electrodes using the IF interpolation.

Date of Award	Oct 2021
Original language	English
Sponsors	INTERREG IVA administered by the SEUPB.
Supervisor	Jim McLaughlin (Supervisor) & Dewar Finlay (Supervisor)