Introduction

Conventional cardiac MRI uses ECG and respiratory sensors (belt, navigator) to prospectively/retrospectively compensate the cardiac and respiratory motion. In the context of radiofrequency ablation of cardiac arrhythmia, a catheter is inserted into one or several cardiac chambers for contact electrophysiology recording and/or ablating the arrhythmogenic substrate [1,2]. These instruments may include position sensors whose readings may be used to sort k-space data to improve image quality, particularly in presence of arrhythmia where ECG may be distorted. Moreover, respiratory compensation using navigator-based techniques may suffer from different scaling factors between displacement of the diaphragm and actual displacement of the heart, depending on the slice location (ex atrium vs apex) [3]. This study presents a method that exploits MR-compatible microcoils located near the catheter tip to retrospectively sort k-space data and retrieve a correct image without using conventional physiological sensors or echo-navigation. The method was evaluated in vivo in pig.

Methods

After sedation of the animal and mechanical ventilation at 12 breaths/min (protocol approved by ethic committee), a 9Fr MR-compatible steerable catheter (Vision-MR, Imricor, Burnsville, MN, USA) was inserted through the femoral vein and navigated until it reached the right ventricle (RV) wall close to the apex. Conventional surface ECG were used to record cardiac rhythm of the animal (85bpm). To perform active catheter tracking during MRI scans and determine the 3D coordinates of the micro coils, a pulse sequence integrating a tracking module interleaved with a radial Gradient Echo (FLASH) imaging sequence was implemented [4]. Radial acquisition was chosen for its intrinsic robustness to motion. The active tracking sequence comprised three non-selective projection acquisitions along the respective axis followed by acquisition of N k-space projections (Figure 1a-b). Imaging parameters were slice thickness=3mm, TE/TR=4.6 /24.96ms, FA = 15°, FOV=169mm, matrix=112*112, N=3 segments per TR, radial views = 6000, TA = 98s), interleaved with tracking data acquisitions (FOV = 450mm, 512 sampled points, time of tracking = 24.04ms). All MRI scans were performed on a 1.5 T MR scanner (MAGNETOM, Aera, Siemens Healthcare, Erlangen, Germany). Following the acquisition, phase of the k-space of the tracking data for each microcoils was analyzed and converted into relative displacement. Directions where a dominant modulation due to cardiac and/or respiratory motion were filtered using a low pass Gaussian filters (sigma = 9) to extract both respiratory and cardiac motion components. Data were binned into 10 cardiac phases then each cardiac phase was sorted into 2 respiratory motion phases. Images were then reconstructed after Cartesian regriding and inverse Fourier transform of the corresponding k-space data. A cine image was reconstructed by playing out the 10 cardiac phases for a fixed respiratory phase.

Results

Figure 2a shows an example of raw signals derived from the most distal microcoil of the catheter on X and Y axes. From these signals, cardiac and respiratory motion were extracted through filtering Figure 2b. The resulting respiratory rate was 12.8 breath/min and ECG 87 bpm, consistent with ventilator frequency and ECG readings (around 85 bpm). Figure 3 displays the segmentation of signals in Figure 2b in 2 respiratory and 10 cardiac phases. As a result, between 246 to 492 projections were combined together to create each image. Figure 4 displays each image at different cardiac phase reconstructed from the proposed method for one respiratory state (Figure 5 is an animated gif of images of Figure 4). A good image quality is obtained with contraction-relaxation of the myocardium.

Discussions and conclusions

This proposed retrospective gating method exploits intra-cardiac position sensors integrated into the catheter. The time penalty introduced by the tracking module appears acceptable (around 20 ms). Although the results presented here show image quality with 3 segments per TR, the sequence implementation offers flexibility between tracking update time and the number of segments acquired between two consecutive tracking modules. Such a method relies on local measurement of displacement of the catheter during cardiac contraction and respiration. Thus it is expected to provide alternative means for retrospective image reconstruction in presence of arrhythmia where conventional ECG and navigator-based approaches may fail. The method has been implemented on a radial gradient echo image acquisition for proof of concept but may be extended to other acquisition sequences.