Image-Based Biophysical Simulation of Intracardiac Abnormal Ventricular Electrograms

Abstract

Abstract-Goal: In this work, we used in silico patientspecific models constructed from 3D delayed-enhanced magnetic resonance imaging (DE-MRI) to simulate intracardiac electrograms (EGM). These included electrically abnormal electrograms as these are potential radiofrequency ablation (RFA) targets. Methods: We generated signals with distinguishable macroscopic normal and abnormal characteristics by constructing MRIbased patient-specific structural heart models and by solving the simplified biophysical Mitchell-Schaeffer model of cardiac electrophysiology. Then, we simulated intracardiac electrograms by modelling a recording catheter using a dipole approach. Results: Qualitative results show that simulated EGM resemble clinical signals. Additionally, the quantitative assessment of signal features extracted from the simulated EGM showed statistically significant differences (p<0.0001) between the distributions of normal and abnormal electrograms, similarly to what is observed on clinical data. Conclusion: We demonstrate the feasibility of coupling simplified cardiac EP models with imaging data to generate intracardiac EMG. Significance: These results are a step forward in the direction of the pre-operative and non-invasive identification of ablation targets to guide RFA therapy.Index Terms-cardiac electrophysiology modelling, intracardiac electrogram modelling, radiofrequency ablation planning, electroanatomical mapping.

Model-Based Feature Augmentation for Cardiac Ablation Target Learning From Images

“…Also the red CARTO signal presents a steep and high amplitude potential occurring after the QRS complex, a LAVA, which does not happen in the blue signal. This characteristic is well reproduced in the simulated LAVA signal, whereas the simulated blue signal presents a clean depolarization complex and (due to the model characteristics as detailed in [23]) appears noise-free.…”

“…Table II shows the set of parameters used for our personalized simulations. Details of the modifications leading to these parameter values can be found in [23]. EGM recording model: The cardiac source model with the tissue-specific parameters and the electrogram recordings were computed simultaneously using a dipole approach as done in Nevertheless, it is important to note that the blue CARTO signal presents a certain degree of noise, which to the untrained eye might seem as abnormal activations.…”

“…Furthermore, we presented in [23] the use of patient-specific anatomical models constructed from DE-MR images coupled with a biophysical electrophysiology model for the synthetic simulation intracardiac electrograms with distinguishable healthy and LAVA-like characteristics.…”

“…We build up on our previous publication on signal simulation [23] in order to show that model-based feature augmentation can introduce physiological knowledge from modelling into the data-driven capabilities of machine learning. The contributions of this manuscript are:…”

“…A preliminary version of this work was presented in [23]. -The analysis of the relative impact of using only imaging data or imaging data together with model-based feature augmentation on classification performance.…”