Image and Simulation Guided Lead Less Endocardial Physiological CRT  

Project summary:  In dyssynchronous heart failure patients, pacing devices can recover synchronous activation and drastically improve patient outcomes. New pacing devices have leadless systems with a small remote electrode and a wireless power transmitter. This device has the advantage of giving many potential pacing sites, including the intrinsic conduction system (physiological pacing). However, which patients will benefit from physiological pacing, what device settings should be used and where to optimally place the remote power transmitted remain open questions. 

 

This project will use image processing to identify structural, functional, and anatomical predictors of response to physiological pacing. Patient specific simulations will be used to predict optimal device settings. Anatomical models will be used to identify the optimal location for the power transmitter. The project will require working with retrospective and prospective clinical data and working as part of an inter-disciplinary team of clinicians, academics and industry. 

 

Project description:   

Background: EBR systems makes a leadless pacing device that can be placed endocardially. The electrode needs to be placed in a position that achieves a synchronous activation pattern but also must be placed in a position where it can receive ultrasound energy delivered by a remote transmitter. Optimizing the location of the electrode and transmitter by trial and error during a procedure is difficult as the transmitter needs to be implanted first and cannot be easily moved. Using pre-procedural imaging may allow the ideal location for the electrode to be determined and then the location of the transmitter that best covers the electrode location can be worked out before any device is implanted. During our recent clinical trial, we demonstrated (EHJ case report accepted) that we could perform physiological pacing using the EBR device. We are now planning a multi-centre study to evaluate endocardial wireless physiological pacing. This project will combine modelling and image analysis to test if guidance can label the septum during the procedure, test if we can identify patients who will benefit from the physiological pacing and optimise the transmitter location in patient receiving the therapy.   

 

Data: The project will initially work with retrospective data (n=10) collected as part of the CT Guided WiSE-CRT clinical study at KCL. Prospective data will be collected as part of international multi-centre the Prof. Rinaldi led, EBR systems supported, trial of leadless endocardial physiological CRT.  

 

Methods: Existing scar detection networks in Pytorch (https://pytorch.org/) and Monai (https://monai.io/) will be used for analysing cardiac CT images. CARP (https://opencarp.org/) will be used for simulating cardiac electro-mechanics. GPErks (https://github.com/stelong/GPErks) will be used to train Gaussian Process Emulators for physics-based model parameter calibration. 

 

Resources: Simulations will be run on Tom2, any network training will be run on local GPU machines, technology evaluation will be performed in the XMR catheter lab during a live procedure (A clinical fellow is already employed, coordinating the centres and writing the ethics for this project). 

 

Year 1-2: Using retrospective patient data a virtual cohort of patients’ will be created. These are four chamber biomechanical models based on CT anatomy, motion and perfusion data, ECG and pressure measurements. The study will introduce synthetic scar locations, pacing locations and device settings to study which patients are likely to benefit most from physiological pacing. Testing specifically if septal scar, lateral wall scar, a very dilated ventricle or slow Purkinje network conduction impact patient response.  

 

Year 1-3: Our developed deep neural networks for identifying scar in CT anatomical images will be applied to retrospective and prospective EBR data sets to evaluate if this method can be used to prospectively screen for scar that may impact physiological pacing. These anatomical based scar classifier methods will be extended to also consider motion derived from retrospective gated CT images. 

 

Year 2-3: Using four chamber heart simulations we will predict which patients are most likely to benefit from physiological pacing relative to lateral wall endocardial pacing. We will test these predictions against the acute haemodynamic response for lateral and physiological pacing in each patient.  We will predict the optimal transmitter location for each case, including how much energy is delivered to the transmitter. We test these predictions by plotting the predicted energy delivered and the battery life at 6 months follow up. 

 

Year 2-3: The septum guidance software will be demonstrated in the XMR catheter lab by leveraging the current Siemens image overlay system. 

​

Year 3-4: Thesis and publication write up

 

​