1st supervisor: Martin Bishop, King’s College London
2nd supervisor: Steven Niederer, King’s College London
Personalisation of anatomically-detailed patient-specific models
LGE MR imaging data will be used to construct anatomically-detailed computational ventricular and whole-torso models using existing pipelines. In this initial proof-of-concept, patients without implanted devices will be used; specifically, patients with ischemic heart disease and large, arrhythmic scars. We will also work closely with MR Physicists in the Dept (Seb Roujol) to develop sequences that enable LGE MR of patients with implanted devices for later incorporation into the modelling pipeline. Pacing data from the implanted device (paced at known site in the RV apex) will be used along with simultaneously recorded ECG. Simulated activation sequences and simulated ECGs will be used along with recorded patient pacing data to parameterise tissue conductivities of the ventricular models in a patient-specific manner. As only activation (QRS of ECG) is used in pace mapping, T-wave and action potential duration personalisation is not required, which represents a key advantage of our approach for accurate personalisation.
Simulated arrhythmia episodes
Sustained episodes of ventricular tachycardia (VT) will be simulated in the ventricular models. Multiple episodes will be simulated in each patient model by applying induction protocols from different sites. During these episodes, torso ECGs will be simulated using full bidomain simulations. This will allow detailed ECG recordings of the clinical VT. A trained clinician will assess the simulated VT episodes and identify the optimal ‘exit’ sites for catheter ablation targets, which will be used later for comparison.
Simulated pace mapping
Sites surrounding the scarred regions will be chosen and repetitive pacing performed at the cycle length of the recorded clinical VT (as done in the clinic). Sites will be chosen, not only on the endocardium (as in procedures), but also transmurally and epicardially, representing a distinct advantage of this in-silico approach. It is expected that many 1000s of points may be chosen (limited only by CPU time) to enable accurate identification of targets. Simulations of pacing will be combined with novel reaction eikonal simulation methods, along with the lead-field approach to enable rapid simulations of ECGs in whole-torso models. At each site, simulated ECGs will be correlated with those recorded during the simulated clinical VT; as performed clinically, those sites having the highest correlation will be identified as ablation targets. The in-silico identified ablation targets will be compared with those identified by the clinicians previously.
Adaptation to use implanted device recordings
Current pace-mapping methods use only ECG recordings of the clinical VT. This is a major disadvantage, as this is often not available and difficult to obtain. The above pipeline will be repeated, but instead of simulating ECG recordings during the VTs, electrogram (EGM) recordings will be simulated from the exact location of the (multiple) sensing electrodes in the implanted devices. As less information-rich than ECG data, different and multiple recording/sensing vectors will be investigated from multipolar devices to obtain the optimal sensing settings to be used in future developments.
Validation
The constructed computational pipeline will be used with the novel MR sequences to obtain real clinical patient data in order to validate this method.
