INTRODUCTION

Free breathing 3D cardiac cine imaging techniques have been developed to overcome limitations of the current standard breath-hold 2D cine imaging. However, the associated scan time increases significantly with improved imaging settings and image quality relies heavily on compensation for respiratory motion, which can vary substantially between different patients. A robust and reliable method is highly desirable. In this study, we propose to achieve 3D cine imaging in one minute during free breathing, with improved motion compensation by using adaptive respiratory self-gating efficiencies.

MATERIALS AND METHODS

Previously we have developed an accelerated free-breathing 3D cardiac cine imaging technique [1] using a pseudo-random variable-density undersampling strategy, called CIRcular Cartesian UnderSampling (CIRCUS) [2-4]. Our previous results have demonstrated that choosing respiratory gating efficiency between 25%-50% can generate reasonable motion compensation for a scan of 2.5-3 mins using the CIRCUS acquisition strategy and an image reconstruction method that combines compressed sensing and parallel imaging (the so called k-t SPARSE-SENSE method) [5,6]. In this study, we propose to optimize the motion compensation strategy while further reducing scan time. Free breathing 3D cine imaging covering the entire left ventricle in a short axis view was acquired on 14 healthy volunteers on a 3.0T MR scanner (GE Medical Systems, Milwaukee, WI) with an 8-channel cardiac coil. The imaging parameters were: FOV = 34.0×25.5 cm2, TR/TE = 4.1/1.7ms, flip angle = 60°, readout bandwidth = ± 125kHz, slice thickness = 4-5 mm, image matrix = 256×144, and number of slices = 28-30. The scan time for 3D imaging was of the order of 2.5 mins. ECG triggers detected during the 3D scan were saved for retrospective gating in the image reconstruction. Temporal resolution was chosen to be 10×TR=41ms. With the variable-density CIRCUS acquisition, the k-space central line along kx (ky=kz=0) was repeatedly acquired and cross-correlation of corresponding signals was performed to obtain a respiratory motion signal. The principle correlation of that data was derived by applying principle component analysis (PCA), which yielded a 1D signal (self-gating signal) containing motion information within the FOV. The derived motion signal was used for sorting the 3D data into a set of bins according to the displacement of the self-gating signal. 3D cine images with the acquired data in the entire 2.5 mins scan time were reconstructed as the reference, with a fixed respiratory gating efficiency of 25%. We retrospectively mimic a shorter scan time of 1 min, by only using the data during the first 1 minute of the scan. Instead of choosing a fixed gating efficiency for the 1 min data sets, we used a series of gating efficiencies by accumulating the bins one by one (so called “adaptive”). All the data sets with different efficiencies were included in the image reconstruction. Thus, instead of discarding some data as happens in conventional gating, all the data was included in the reconstruction, which not only provides adaptive gating efficiencies but also improves the data fidelity.

RESULTS AND DISCUSSION

We selected data for review from 3 subjects whose heart rates were similar (~70 bpm) but with different respiratory motion. With the derived self-gating signal, data was sorted into bins and the data distribution varied according to the individual breathing patterns, which directly affects the image quality. Three cases are shown here (Figs. 1-4), with images from 2.5 mins and 1 min scans, at end-systolic (ES) and end-diastolic (ED) phases, and the signal profile across the left ventricle throughout the cardiac cycle (r-t). With the proposed adaptive respiratory gating efficiencies, we generated high quality and reliable 3D cardiac cine images even for data from one minute of acquisition, regardless of the varying respiratory motion patterns.

CONCLUSIONS

We have developed a highly accelerated, free breathing, 3D cardiac cine imaging approach which could be completed in one minute using adaptive respiratory self-gating efficiencies. This method can provide a significantly improved MR imaging tool for assessing cardiac function.

Acknowledgements

NIH K25 EB014914 (JL), NIH R56HL133663 (JL), GE Healthcare Research Grant (JL)