ESTRO 2024 - Abstract Book

S3886

Physics - Image acquisition and processing

ESTRO 2024

Quantitative T1 mapping in the presence of an implantable cardiac device on a 1.5T MR-linac

Beau P. Pontré 1,2 , Stefano Mandija 2,3 , Osman Akdag 2 , Tim Schakel 2,3 , Cornelis A. T. van den Berg 2,3 , Martin F. Fast 2

1 University of Auckland, Department of Anatomy and Medical Imaging, Auckland, New Zealand. 2 University Medical Center, Department of Radiotherapy, Utrecht, Netherlands. 3 University Medical Center, Computational Imaging Group for MR Diagnostics and Therapy, Image Sciences Institute, Utrecht, Netherlands

Purpose/Objective:

Cardiac arrhythmias are a common precursor for sudden cardiac death. In cases where anti-arrhythmogenic medication or therapy through an implantable cardiac defibrillator (ICD) proves insufficient, catheter ablation is used as a treatment to form scar tissue that disrupts erroneous electrical conduction in the myocardium. Unfortunately, the procedure is invasive, not always effective and introduces procedural risks. Cardiac radioablation is emerging as salvage treatment of cardiac arrhythmia. The key for this treatment is the identification of the scar tissue. To this end, quantitative mapping of myocardial longitudinal relaxation times (T1) can be useful in guiding the identification of treatment targets [1], with a T1 increase of 7% reported in chronic myocardial infarction [2]. Previous research has demonstrated the feasibility of T1 mapping on an MR-linac [3]. Standard techniques for myocardial T1 mapping are typically based on a balanced steady-state free precession (bSSFP). However, these sequences are highly sensitive to metal-induced artifacts making them inappropriate for imaging during arrhythmia radioablation on an MR-linac.

This study compares the use of balanced vs spoiled gradient echo (GE) sequences for myocardial T1 mapping in the presence of an ICD on an MR-linac in phantoms and in a volunteer.

Material/Methods:

Two cardiac-triggered, inversion recovery (IR) prepared sequences were acquired on a 1.5T Unity MR-linac (Elekta AB, Sweden) from which T1 maps were obtained by means of two-parameter least-squares fitting:

1. IR-prepared, single-shot bSSFP, TR/TE/FA = 3.8ms/1.9ms/35° 2. IR-prepared, single-shot spoiled GE, TR/TE/FA = 3.7ms/1.8ms/7°

Images were acquired with in-plane resolution 2 x 2mm at five 4mm thick transverse slices through an 8-tube phantom (Eurospin TO5) with known T1. Seven images were acquired for each slice with inversion times (TI) = 200, 400, 600, 1600, 2600, 3600ms, and one without IR-preparation. A simulated ECG trace with a heart rate of 60 bpm was used for the trigger and a delay of five heart cycles between each slice was used to allow for sufficient recovery of longitudinal magnetisation. Acquisition time for each sequence was about 5 minutes. All scans were repeated with an ICD placed as indicated in Figure 1a. The experiment was repeated on a healthy volunteer, with images from three contiguous short axis slices acquired using a peripheral pulse unit trigger. Each set required a series of seven, 12-second breathholds. Signal acquisition blocks were timed to occur during mid-diastole for all inversion times. All scans were repeated with an ICD placed on the left clavicle of the volunteer.