Session PB9.2

Quality Evaluation and Effect of Time Synchronization on the Digital Recovery of Intracardiac Electrograms

M Sanroman-Junquera, I Mora-Jimenez*, E Everss, J Almendral-Garrote, A Garcia-Alberola,
F Atienza, L Castilla-SanJose, JL Rojo-Alvarez

University Rey Juan Carlos
Fuenlabrada, Spain

Background: Digital recovery of cardiac signals from paper printed recordings is required when analyzing historical ECG signals which are not available in digital format. Also, in implantable cardioverter defibrillators (ICD), it is not always possible to program the adequate setting thresholds for sampling and storing device-recorded intracardiac electrograms (EGM), due to the tachycardia cycle programming thresholds or to patient condition.
Aim: Our aims were: (1) to propose an image processing algorithm suitable for recovering ICD stored EGM; (2) to evaluate its performance in different cardiac rhythms; (3) to use an adequate time synchronization processing for this application.
Method: EGM from simultaneously acquired tip-ring and can-coil ICD lead configurations were scanned and separated into two streams, and signal tracking was made for each stream by: (1) detection of the midpoint of the column black pixels width; (2) detection of the most likely tracking angle according to a Least Squares (LS) fit of the precedent slope; (3) filling of the image gaps. Recovered EGM (from the same manufacturer) were compared to their gold standard (ICD sampled and stored recordings), in 4 sinus rhythm, 8 sustained tachycardia, 10 ventricular fibrillation, from 17 patients. Performance was characterized according to Percentage Root mean square Difference (PRD), Normalized Root Mean Square (NRMS), and maximum amplitude error (MAE). Three different techniques for time alignment(LS, spline, and matched filter) were benchmarked. Statistical significance was given by paired t-test yielding alpha<0.05 (*).
Results: Consistent lower performance was observed in tip-ring than in can-coil recovered EGM (for LS alignment, PRD 16.8+-9.8 vs 9.5+-3.1*, NRMS 9.2+-3.6 vs 5.3+-2.0*, MAE 1.0+-0.5 vs 0.3+-0.2*). Performance was improved by LS alignment with respect to spline and matched filter (for instance, PRD 9.5+-3.1, 10.3+-3.0*, 14.3+-5.2*, respectively) in discharge EGM, but not in tip-ring EGM. Tip-ring recordings were better recovered from SR than from T and from VF (NRMS 6.13+-0.44, 5.77+-2.21, 12.02+-3.75*, respectively).
Conclusions: Our algorithm allows to automatically recovering ICD EGM from paper printed recordings. Quality evaluation shows that the performance depends on the lead type (can-coil or tip-ring) and on the use of an adequate time synchronization.

(Abstract Control Number: 85)