Robust free-breathing SASHA T1 mapping with high-contrast image registration

Abstract

Background: Many widely used myocardial T1 mapping sequences use breath-hold acquisitions that limit the precision of calculated T1 maps. The SAturation-recovery single-SHot Acquisition (SASHA) sequence has high accuracy with robustness against systematic confounders, but has poorer precision compared to the commonly used MOdified Look-Locker Inversion recovery (MOLLI) sequence. We propose a novel method for generating high-contrast SASHA images to enable a robust image registration approach to free-breathing T1 mapping with high accuracy and precision.

Methods: High-contrast (HC) images were acquired in addition to primary variable flip angle (VFA) SASHA images by collecting an additional 15 k-space lines and sharing k-space data with the primary image. The number of free-breathing images and their saturation recovery times were optimized through numerical simulations. Accuracy and precision of T1 maps using the proposed SASHA-HC sequence was compared in 10 volunteers at 1.5 T to MOLLI, a breath-hold SASHA-VFA sequence, and free-breathing SASHA-VFA data processed using conventional navigator gating and standard image registration. Free-breathing T1 maps from 15 patients and 10 volunteers were graded by blinded observers for sharpness and artifacts.

Results: Difference images calculated by subtracting HC and primary SASHA images had greater tissue-blood contrast than the primary images alone, with a 3× improvement for 700 ms TS saturation recovery images and a 6× increase in tissue-blood contrast for non-saturated images. Myocardial T1s calculated in volunteers with free-breathing SASHA-HC were similar to standard breath-hold SASHA-VFA (1156.1 ± 28.1 ms vs 1149.4 ± 26.5 ms, p >0.05). The standard deviation of myocardial T1 values using a 108 s free-breathing SASHA-HC (36.2 ± 3.1 ms) was 50 % lower (p <0.01) than breath-hold SASHA-VFA (72.7 ± 8.0 ms) and 34 % lower (p <0.01) than breath-hold MOLLI (54.7 ± 5.9 ms). T1 map quality scores in volunteers were higher with SASHA-HC (4.7 ± 0.3 out of 5) than navigator gating (3.6 ± 0.4, p <0.01) or normal registration (3.7 ± 0.4, p <0.01). SASHA-HC T1 maps had comparable precision to breath-hold MOLLI using a retrospectively down-sampled 30 s free-breathing acquisition and 30 % higher precision with a 60 s acquisition.

Conclusions: High-contrast SASHA images enable a robust image registration approach to free-breathing T1 mapping. Free-breathing SASHA-HC provides accurate T1 maps with higher precision than MOLLI in acquisitions longer than 30 s.

Keywords: Extracellular volume; Fibrosis; Free-breathing; Image registration; SASHA; T1 mapping.

Fig. 1

Bloch equation simulation of a sinusoidal-ramped variable flip angle (VFA) bSSFP readout for native myocardium (T₁/T2 = 1175/50 ms) and blood (T₁/T2 = 1650/240 ms). a Readout flip angles are increased in a sinusoidal ramp pattern for 120° SASHA-VFA (green circles) and 70° SASHA-VFA (dashed blue line). b Signal evolution of myocardium (blue) and blood (red) for a non-saturated image using 120° SASHA-VFA. Typical k-space center locations for the primary and high-contrast images are marked with light blue and dashed red boxes respectively. c Signal evolution for a saturation recovery image using 120° SASHA-VFA

Fig. 2

a Partial sequence diagram of the proposed high-contrast SASHA sequence showing a non-saturated image and a saturation recovery image. Image acquisition is extended by ~40 ms by appending additional k-space lines (dashed red) to primary image readouts (light blue). b k-space coverage is shown for primary and high-contrast images. The lower half of k-space is shared between primary and high-contrast images and the additional k-space lines cover the central portion of k-space

Fig. 3

Overview of proposed analysis for free-breathing SASHA-HC. The displacement of the heart in free-breathing data is estimated using rigid image registration and used to select a subset of images for further analysis. This subset is aligned with a non-rigid image registration algorithm using both primary and difference images, and the resulting deformation fields are applied to the primary images. A T₁ map is calculated from the aligned images using non-linear least squares curve fitting

Fig. 4

Bloch equation simulation of tissue-blood contrast for primary (thin green), high-contrast (dashed red), and difference (thick blue) images as a function of saturation recovery time. Contrast is characterized by the difference between myocardial and blood transverse magnetization (M_XY)

Fig. 5

a Simulated precision for native myocardium for an acquisition duration of 45 heartbeats. The optimal TS times, corresponding to the minimum coefficient of variation, are marked for sampling scheme with various NS images. b The optimal TS times for each sampling schemes are plotted as a function of total acquisition duration. c The optimal coefficients of variation (CV) is plotted as a function of acquisition time, using fixed TS times from b). All simulations assume a 100 % acceptance rate and a heart rate of 60 bpm

Fig. 6

Primary, high-contrast, and difference images are shown for non-saturated and saturation recovery preparation in a healthy volunteer. Brightness and contrast levels in displayed images are matched between the NS and TS images

Fig. 7

T₁ maps from a healthy subject using different techniques. Breath-hold techniques are shown in the top row and the bottom row shows T₁ maps calculated from free-breathing 120° SASHA-VFA data using various approaches

Fig. 8

T₁ maps from a clinical patient undergoing evaluation for coronary artery disease using MOLLI, breath-hold SASHA-VFA, and free-breathing SASHA-HC. T₁ maps were acquired before and 10 min after administration of gadolinium contrast. ECV maps were calculated using an assumed hematocrit of 0.40

Fig. 9

Measured myocardial T₁ precision of free-breathing SASHA with different sampling schemes in healthy volunteers with normalization to MOLLI. The mean normalized precision (over all subjects) is shown with solid lines and the corresponding coloured shaded region indicates ±1 standard deviation. Acquisition time is shown on the x-axis for various different gating efficiencies at heart rates of 60 and 75 bpm