Interactive Patient-Specific Simulation of Cardiac Electrophysiology

Hugo Talbot 1, 2
1 SHACRA - Simulation in Healthcare using Computer Research Advances
LIFL - Laboratoire d'Informatique Fondamentale de Lille, Inria Lille - Nord Europe, Inria Nancy - Grand Est
2 ASCLEPIOS - Analysis and Simulation of Biomedical Images
CRISAM - Inria Sophia Antipolis - Méditerranée
Abstract : As in most of the medical departments, interns in cardiac electrophysiology follow a curriculum combining an intense theoritical learning with a long clinical practice. After years of theory (mainly book learning), junior electrophysiologists start practicing on patients under the supervision of a senior cardiologist. In the last decades, the improvement of computational technologies led to the development of numerical tools dedicated to training, planning or guiding of surgical procedures. The objective of this thesis is to construct a training framework, allowing junior electrophysiologists to practice radio-frequency (RF) ablation for the treatment of ventricular arrhythmias on virtual patients. Training on in silico models can not only shorten the electrophysiology curriculum, but it can also standardize it. Yet the development of such training systems raises several challenges. The first challenge consists in simulating the cardiac electrophysiology in real-time. Through the improvement of cardiac imaging, characterization of the normal and arrhythmic electrical activity of the heart using mathematical models has been an important research topic. We focus here on a model representing the electrophysiology at the organ scale: the Mitchell-Schaeffer model. A powerful GPU implementation is proposed to reach real-time performances. Our efficient electrophysiology model is coupled with a mechanical model of the heart. A realistic left bundle branch block can be simulated, thus inducing the associated late contraction of the left ventricle. For clinical application of electrophysiological mathematics, our virtual scenario of cardiac arrhythmias needs to be personalized. This crucial step aims at adapting all model parameters in order to fit patient data, acquired intra-operatively. After a detailed state of the art of optimization methods, the unscented Kalman filter deriving from a Bayesian approach is chosen and applied on a dataset of three patients suffering from ventricular tachycardia. Relying on our GPU electrophysiology model, the optimization process is achieved in about 20~minutes, while faithfully reproducing the pathology recorded in the operation room. Lastly, the construction of the first training framework dedicated to cardiac ablation is presented. The scenario reproduces the catheter navigation inside the vascular system using a physics-based approach, and the beating heart is modeled from patient data. In addition to the cardiovascular navigation, a case of an ectopic focus in the right ventricle is modeled using our GPU implementation. An innovative multithreading approach couples both simulations, thus offering performances close to real-time. The computational efficiency allows the trainee to interact with the simulation and perform all the clinical gestures, namely electrical catheter measurements, electro-anatomical mapping, electrical stimulation and eventually RF ablation. A clinical evaluation by electrophysiologists highlights the good performances and the realism of the training framework.
Document type :
Theses
Liste complète des métadonnées

Cited literature [115 references]  Display  Hide  Download

https://hal.inria.fr/tel-01097201
Contributor : Brigitte Briot <>
Submitted on : Friday, December 19, 2014 - 10:21:47 AM
Last modification on : Thursday, February 21, 2019 - 10:52:54 AM
Document(s) archivé(s) le : Monday, March 23, 2015 - 5:35:35 PM

Identifiers

  • HAL Id : tel-01097201, version 1

Collections

Citation

Hugo Talbot. Interactive Patient-Specific Simulation of Cardiac Electrophysiology. Computer Science [cs]. Université des Sciences et Technologies de Lille, 2014. English. ⟨tel-01097201⟩

Share

Metrics

Record views

1063

Files downloads

702