To develop robust free-breathing SASHA T1 mapping using high-contrast image registration and compare to MOLLI T1 mapping.T1 mapping has been used to assess diffuse myocardial fibrosis in a variety of cardiac diseases, but increased accuracy and precision are needed to confidently measure subtle changes in sub-clinical diseases. The SASHA technique (1) is more accurate than the commonly used MOLLI sequence (2), but SASHA has poorer precision because of the reduced dynamic range of saturation recovery. Motion-corrected free-breathing acquisitions have been used for T2* mapping (3), however robust motion correction is challenging for SASHA images due to poor blood-tissue contrast. We present a novel approach generating co-registered high-contrast images that improve the robustness of image registration and enable free-breathing SASHA T1 mapping.

Variable flip angle (VFA) SASHA (4) images have poor blood-tissue contrast because the center of k-space is acquired early on in the bSSFP image readouts (Fig. 1). Blood-tissue contrast improves later in the readout due to inherent bSSFP T2/T1 weighting, but T2-weighting would introduce systematic error in T1 values calculated from this data. Instead, additional low frequency k-space lines acquired following the primary SASHA-VFA image can be used to generate a secondary higher contrast (HC) image that can be used for registration. As the HC lines are acquired adjacent to and immediately following the high-frequency lines for the primary image with similar contrast, only a small number of HC lines are needed and the high-frequency lines can be can be shared with the primary image in a keyhole manner (Fig. 2). The difference image between primary and HC images (Fig. 3) provides consistent blood-tissue contrast between non-saturated and saturation-recovery images, improving the robustness of image registration. As the HC images are intrinsically co-registered with the primary images, registration performed on the HC or difference images can be used to directly motion correct the primary images used to calculate T1 maps.

A larger 120° maximum flip angle was used in order to maximize blood-tissue contrast for the HC image, however the VFA length was increased to maintain similar flip angles prior to the primary image k-space center as the previously described 70° SASHA-VFA. Other parameters include: 360×270 mm FOV, 256×150 matrix, 78% phase resolution, 7/8ths partial Fourier, 1.19/2.76 ms TE/TR, and GRAPPA R=2 acceleration, with 65 phase encodes for the primary image. High-contrast acquisitions consisted of 15 low-frequency lines at R=3 and were reconstructed using GRAPPA prior to combination with primary k-space data. POCS reconstruction was used on the combined high-contrast k-space data.

10 healthy volunteers (6 male, 33±8 yrs) were imaged on a Siemens 1.5T Avanto scanner with written informed consent. Breath-hold T1 data were acquired in a short-axis slice using 5(3)3 MOLLI, 70° SASHA-VFA, and 120° SASHA-VFA. The MOLLI sequence used 1.04/2.68 ms TE/TR and other parameters matched to SASHA. Free-breathing 120° SASHA-VFA data was acquired with 10 non-saturated images separated by >5 seconds and 30 saturation recovery images for a total imaging time of <90 seconds. 55% of the free-breathing images closest to end-expiration were automatically selected and aligned with a non-rigid image registration algorithm (5) using information from difference and primary images (HC-REG). Conventional registration was also performed using the primary images only (NORM-REG). SASHA T1 maps were calculated using a 2-parameter model and MOLLI T1 maps were calculated using a 3-parameter model with Look-Locker correction. The mean and standard deviation of myocardial T1 values were calculated for each T1 map.

Primary, HC, and difference SASHA-VFA images from one subject are shown in Fig. 3 and T1 maps are shown in Fig. 4. Mean T1 values (Fig. 5) were not statistically different between breath-hold 70° SASHA-VFA, breath-hold 120° SASHA-VFA, and free-breathing 120° SASHA-VFA with HC-REG (2-way ANOVA, p>0.05). The T1 standard deviation of free-breathing SASHA-VFA with HC-REG (46.1±3.8 ms) was lower than both breath-hold SASHA-VFA (55.3±7.7 ms, paired t-test p<0.05) and MOLLI (57.5±9.0 ms, paired t-test p<0.05).

Image registration of free-breathing SASHA-VFA data was robust with HC-REG and no apparent residual motion was observed. Residual motion was apparent in the majority of NORM-REG data, and myocardial T1 values were higher than HC-REG (1162.7±35.4 vs 1153.4±27.3 ms), consistent with blood pool contamination.

High-contrast images with improved blood-tissue contrast can be acquired in ~40 ms using keyhole sharing, enabling robust free-breathing SASHA T1 mapping. Myocardial SASHA-VFA T1 values were similar between free-breathing with HC-REG and standard breath-hold acquisitions, but with a 32% reduction in standard deviation. Free-breathing SASHA-VFA with HC-REG is an accurate T1 mapping technique with lower variability than the reference MOLLI sequence.