Session S64.5

Efficient Modeling of ECG Waves for Morphology Tracking

R Dubois*, P Roussel, M Vaglio, F Extramiana,
F Badilini, P Maison-Blanche, G Dreyfus

ESPCI-ParisTech
Paris, France

We develop a new approach to fully automatic ECG wave extraction and morphology tracking. It is based on Generalized Orthogonal Forward Regression (GOFR), which allows decomposing a one-dimensional signal into a set of appropriately chosen parametric functions by adjusting their parameters. Specific functions are designed, depending on the problem that is addressed. Two applications of GOFR to ECG modeling are presented. First, in order to delineate ECG characteristic waves, we describe a specific function, named Gaussian Mesa function (GMF), which is appropriate for fitting any characteristic cardiac wave. GOFR adjusts the parameters of one GMF for each cardiac wave, which is subsequently labeled and delineated. As a second example, we track the evolution of the T-wave morphology by introducing a Bi-Gaussian function (BGF), which is appropriate for T-wave modeling and characterization: its parameters can be linked to morphological characteristics of the T-wave (amplitude, position in time, rising and falling edge slopes).
The approach was validated in three experimental settings. We first used a database consisting of 100 digital 12-lead ECG Holter recordings acquired during three 24-hour periods (baseline and after 160 and 320 mg of sotalol) from 38 healthy subjects. QRS and T waves were automatically delineated using GOFR and GMF; the marker positions obtained were very close to those obtained from manual annotation (QRS onset: -9.5±7.7ms, QRS offset: 4±13ms, and T wave offset 14±10ms), and the automatic QT measurement was strongly correlated to the manual measurement (R=0.96, p<0.05). Subsequently, the morphology changes of the delineated T-wave were tracked to exhibit the influence of Sotalol. All parameters computed from the BGF model showed highly significant differences between the recordings performed at the time of peak plasma concentration and the baseline ECGs (p<0.05). Finally, we validate GOFR-BGF modeling on drug-free 12-lead resting ECGs from 100 genotyped long QT syndrome (LQTS) patients. The LQTS test confirmed the results obtained on the previous database; in particular, it was shown that the T-wave width and descending phase of repolarisation were more prolonged in type 2 LQTS than in type 1 (p<0.05) and control group (p<0.05).
The results confirm that the combination of GOFR and of specific, appropriate parametric functions is remarkably efficient for ECG wave modeling; in the present paper, we emphasize their application to wave morphology characterization and tracking.

(Abstract Control Number: 30)