Electrocardiographic imaging (ECGi) is an emerging non-invasive technique that computes unipolar electrograms (EGMs) at the epicardial surface from high-density body-surface ECG recordings and torso anatomy.
Activation mapping is an important post-processing step that allows the user to make sense of these reconstructed EGMs by imaging the heart’s activation sequence. The standard technique for constructing such maps is to annotate individual signals at the moment of sharpest downstroke. Although this annotation technique is well suited for the analysis of invasively recorded EGMs, it often yields poor quality maps in ECGi, since high frequency components of the signals are poorly reconstructed. In effect, this leads to lines of artificially strong activation time gradients.
To overcome this problem, we propose an algorithm that compares EGMs from neighboring recording locations and estimates delays between these points. Delay estimates depend on the overall signal, making them less sensitive to high frequency components. As delays alone cannot yield a complete map, we then describe a workflow and statistical framework to construct a spatially coherent activation map from the standard estimates of local activation times and these delay estimates. The method is optimized using simulated data and evaluated on clinical data from 12 different activation sequences.
Our results indicate that the standard methodology yields maps that display regions of erroneously high delays between contiguous points. The proposed workflow enhanced these maps significantly, correcting delays and smoothing the resulting maps. On our validation set, this resulted in a 19% reduction in relative error. Estimating delays between neighbors is therefore an interesting option for activation map computation and optimization in the setting of ECGi.