Cardiac CT atrial motion to estimate atrial fibrosis burden

Project reference: DTP_SIE_20 This project has already been allocated and is no longer available for applications

First supervisor: Steven Niederer
Second supervisor: Ronak Rajani

Start date: October 2020

Project summary: The project will consist of four parts. 

  1. Using retrospective cardiac CT data set recorded in heart failure patients we will create a manually annotated dataset that will provide a gold standard for tuning and optimising cardiac CT atrial motion tracking. We will then evaluate and optimise conventional or machine learning feature tracking algorithms.

  2. We will obtain ethics to recruit and scan 20 patients as a proof of principal pilot study. Cardiac CT scans (as part of research) and electro-anatomical voltage mapping (as part of routine clinical care) will be performed in each patient. CT scans will be processed using feature tracking and area strains and principal strains will be calculated. Strains will be projected on to an atlas-based fibre field to provide estimates of atrial fibre strain. We will then compare different indexes of atrial strain (area strain, fibre strain and principal strain) against low voltage regions (that correspond with regions of fibrosis) recorded during the electro-anatomical mapping. This will provide a validation that fibre strains can be used as an index of atrial fibrosis.

  3. Patients will receive an atrial ablation and the fibrosis burden estimated by the strain indexes will be compared with their outcome after 6 months. This will provide pilot data to support the use of atrial CT mechanics for predicting atrial ablation outcome.

  4. CT images can be reconstructed at different timings (5,10 or 20%). We will evaluate how this affects motion analysis.

Project description: Atrial fibrillation (AF) is the most common arrhythmia. AF affects 5.2 million Americans, costs the US up to $26 billion per year, and increases the risk of cardiovascular disease, stroke, and death.

Guidelines recognize that AF is “complex and difficult for clinicians to manage”. Patients can be treated pharmacologically, by catheter ablation to isolate or destroy aberrant atrial tissue, or by AV node ablation, coupled with pacemaker implantation, to isolate the atria. Unfortunately, pharmacological treatments have profound side effects, AF may recur in ~50% of AF ablation cases, and pacemaker dependency has inherent risks. No single treatment is best in all cases. Consequently, selecting the optimal treatment for each AF patient remains a daily clinical challenge.
Precision therapies require accurate characterization of each patient’s specific disease phenotype. Pathological atrial fibrosis is a major contributor to sustaining AF, has been repeatedly implicated in its pathogenesis, and is proposed as a biomarker to guide personalizing treatment. Cardiac magnetic resonance (CMR) late gadolinium enhancement (LGE) currently provides the only non-invasive estimate of atrial fibrosis. However, widespread adoption of atrial LGE-CMR has been hindered by difficult and non-standardized image acquisition and analysis techniques and minimal validation.


To overcome these challenges, we propose to use mechanics-based measures to identify localized atrial fibrosis. Atrial fibrosis fosters chaotic electrophysiology,  attenuates local atrial mechanics, decreases contractility, and increases stiffness. This proposal exploits the mechanistic link between atrial fibrosis and atrial mechanics to develop and validate a mechanicsbased classifier of atrial fibrosis.


Fibrosis is both regulated by and alters cardiac mechanics. Global atrial mechanics correlate with fibrosis burden and local strain correlates strongly (cross correlation 0.85-0.9) with ablation scar and moderately (r=0.66 and AUC=0.66) with fibrosis. These preliminary studies demonstrate that left atrial mechanics can provide a measure of fibrosis burden. However, conventional ventricular 2D CMR does not capture the complex atrial anatomy and motion; the thin atrial wall is not easy to image; and the features that could be used for motion tracking are small, requiring high-resolution high contrast imaging. 


Drastic reductions in radiation doses with next generation cardiac computer tomography imaging combined with multiple detectors now make motion measurements with cardiac CT a viable clinical tool offering exceptional image quality with isotropic resolution. We wish to build on our use of cardiac CT to measure motion in the ventricles to measure atrial motion from cardiac CT and test if measures of local strain correspond to invasive measurements of atrial fibrosis, measured using electroanatomic mapping systems at the time of their procedure.