Purpose

Introduced over 25 years ago, myocardial tagging is installed on all manufacturers’ MRI scanners^1-3. Despite extensive validation as the best method for clinical quantitative measure of myocardial wall deformation, tagging is limited to a research context due to the time-intensive analysis and need to acquire additional sequences. Cardiac MRI (CMR) sequences must provide maximum diagnostic information, and maximum patient throughput, with minimum operator effort and time (cost)⁴. We developed an accelerated CMR cine sequence with a “hidden” added capability to acquire cardiac strain data while requiring no extra operator effort. Called Cine Watermark (CWM), this sequence acquires normal cine magnitude images plus spatial-modulated tagging added only into the image phase (“watermark”) allowing quantitative cine strain calculation.

Methods

Figure 1 illustrates the ECG-triggered CWM sequence that acquires both normal and complimentary spatial modulation of magnetization (SPAMM/CSPAMM) prepared cine image sets. Before the SPAMM is enabled, a low-resolution pre-scan reference image is acquired for background phase correction. A classic “1-1” saturation SPAMM was modified were each non-selective rectangular RF pulse has a 45° flip angle. The cine readout uses a standard, low flip angle, spoiled GRE method with a parallel imaging acceleration factor of two. Within one breath-hold CWM acquires a full SPAMM image set then a full CSPAMM image set. CWM acquisitions were run with spatial modulation in the readout (RO) axis and phase encode (PE) axis directions to create orthogonal line-tags and combined for grid-tag patterns. Twelve healthy human subjects, 19-42 yo, 3 female, with informed consent, were scanned in a 3T Verio scanner (Siemens, Erlangen, DE) with a 32-chan anterior/posterior RF coil array (Invivo, Gainesville, FL). CWM scan slice locations included one long-axis (LA) 4-chamber and 1-3 mid-left ventricular (LV) short-axis (SA) views. Scan parameters included: field-of-view = 264x231 mm, slice thickness = 6 mm, matrix = 176x154, in-plane pixel size = 1.5x1.5 mm, flip angle = 8°, BW = 406 Hz/pix, averages = 1, cine phases = 20-25, and tag spacing = 6 mm. CWM image reconstruction was performed offline using custom Matlab programs (Mathworks, Natick, MA). Reconstruction included: 1) correct the SPAMM and CSPAMM background phase with the reference scan phase, 2) perform a complex image sum of the SPAMM+CSPAMM that eliminates the cosine modulation components to leave only the (central k-space) conventional T1-weighted component, from which, the magnitude images provide normal cine images without tags, 3) perform a complex image difference of SPAMM–CSPAMM that eliminates the central k-space T1-weighted component to leave only the cosine modulation components, from which, the phase images provide strong and persistent (0°/180°) phase-only tag lines that track cardiac displacement for strain calculation. Cardiac segmentation lines drawn onto CWM magnitude images also automatically segmented corresponding phase images (also Matlab). The cardiac strain analysis is described in detail in a separate submission. Briefly, cardiac strains were tracked by a novel non-linear least squares method combining quadratic b-spline contours, nearest neighbor optimized phase tracking, and physically-motivated regularizers. Cardiac LV SA, six sector, circumferential strain (Ecc) results were calculated and averaged around the LV.

Results

All CWM scans provided cine magnitude and spatial-modulated phase image sets suitable for strain analysis within a single breath-hold of 25 seconds. Figure 2 presents representative CWM LA and SA magnitude cine frames at end-diastole (ED) and end-systole (ES) time-points with subsequent LV segmentation. Figure 3 presents LA phase-tagged results at ED and ES with spatial-modulation applied in the RO, PE and combined RO+PE (grid) directions, where the segmented LV images show clear displacement and bending of the phase-tagging patterns at ES. Similar to Figure 3, Figure 4 presents SA phase-tagged results that also show clear displacement and bending of the LV phase-tagging patterns at ES. Figure 5 presents representative mean Ecc strain results averaged from 3 contiguous, mid-LV, SA slices for a human subject. All sectors, except the inferior sector, show nominal strain with consistent trajectories.

Discussion

CWM can acquire cardiac cine and phase-tagged cine images in a single breath-hold. The integration of parallel imaging acceleration was essential to achieve one breath-hold per scan capability and the decoupled RO & PE encodings improved the strain analysis. The Ecc strain results generally agree with the previous literature⁵; however, further work is needed to improve tracking robustness and overall accuracy.

Conclusions

Quantitative assessment of regional myocardial contraction is an essential element of diagnostic CMR. Cine “Watermark” MRI provides an approach to breath-hold cine imaging with integrated “hidden” capability to acquire strain data encoded into the phase. This work provides a platform for further sequence development to included further acceleration and the potential for SSFP imaging.

Acknowledgements

Special appreciation to Martha Forloines for her valuable management of MRI operations and technical assistance.