Introduction
Cardiovascular magnetic resonance (CMR) is a medical imaging modality whose wide applications include tissue characterization across different physiological and pathological conditions. CMR longitudinal relaxation time (T1) mapping has become particularly relevant since it differs between healthy and diseased tissue, and aims to play an important role in clinical decision-making in cardiovascular disease [
1‐
3]. T1 mapping provides pixel-wise T1 values by fitting images acquired during the T1 magnetization recovery at different times after application of an inversion pulse [
4], a saturation pulse [
5], or a combination of both [
6].
CMR T1 mapping is complex due to the heart’s motion, and several two-dimensional (2D) pulse sequences have been presented in recent years to tackle the problem [
7‐
9]. However, these sequences are designed to acquire only a single slice per breath-hold. The modified Look-Locker inversion recovery (MOLLI) sequence applies an inversion pulse followed by different single-shot steady-state free precession (SSFP) readouts over multiple heartbeats [
4]. MOLLI is precise and reproducible; however, it underestimates T1, mainly due to the magnetization transfer effect and imperfect inversion efficiency [
8]. To overcome this limitation, saturation recovery single-shot acquisition (SASHA) replaces the inversion pulses with saturation pulses, thus avoiding the underestimation of T1 values due to incomplete recovery of the signal after the inversion pulse [
5]. This approach improves accuracy at the cost of a lower signal–noise ratio (SNR). A different 2D saturation recovery sequence design is provided by modified Look–Locker acquisition using saturation recovery (MLLSR), which allows the saturation pulses to be shared between several readouts acquired in different heart beats to minimize T1 estimation error and provide high flexibility [
10]. Lastly, saturation pulse-prepared heart rate independent inversion recovery (SAPPHIRE) is a hybrid approach that uses inversion and saturation pulses and sits at the midpoint between the advantages and limitations of saturation and inversion schemes [
6].
Highly accurate T1 estimates can also be obtained with three-dimensional (3D) pulse sequences based on saturation recovery and developed to fully cover the left ventricle (LV). 3D SASHA combines a 2D SASHA-based pulse scheme with free-breathing imaging for 3D acquisition at 1.5T [
11]. A more recent free-breathing 3D T1 mapping sequence provides 3D acquisition at 3T, based on a new pulse scheme that acquires substantially fewer T1-weighted images than 3D SASHA [
12]. These 3D saturation recovery approaches offer higher SNR and good spatial coverage at the cost of longer acquisition times. They require navigator-triggered free-breathing and, therefore, rely on the respiratory navigation performance to achieve good image quality and acceptable acquisition time. Acquisition time is nevertheless in the range of minutes, even when denoising and optimization techniques are used [
13,
14]. Such long acquisition time compromises the feasibility of T1 mapping during contrast equilibrium and reduce clinical applicability. Furthermore, these 3D saturation recovery sequences acquire all T1-weighted images in the same RR interval of the saturation pulse, which compromises T1 estimation quality, particularly for long T1s and high heart rates [
15].
Independent of magnetic preparation, most proposed T1 mapping techniques are acquired using SSFP readout techniques. At 3T, SSFP sequences have been associated with higher energy deposition and increased off-resonance artifacts [
16]. Alternatives to SSFP sequences are spoiled sequences. These sequences use fast low-angle shot (FLASH) imaging readouts to avoid off-resonance artifacts and to eliminate transverse relaxation time (T2) dependence [
12,
17,
18]. The main limitation of FLASH schemes is the low SNR compared with SSFP sequences.
In addition to the development of pulse sequences, new techniques have emerged to accelerate acquisition. Compressed sensing exploits the sparsity or compressibility of CMR images and accelerates acquisition by undersampling without significantly degrading images [
19]. This technique has been successfully used to accelerate 3D cardiac imaging [
20‐
22]. Compressed SENSE is built on compressed sensing and incorporates components of the parallel acquisition technique SENSE [
23]. Studies showing the feasibility of compressed SENSE have recently been published for T1 calculation in the brain [
24] and for cardiac cine imaging [
25]. As expected, the method significantly reduced acquisition time, particularly in 3D acquisitions.
In this study, we propose and validate the 3D saturation recovery compressed SENSE rapid acquisition (3D SACORA) imaging sequence, a new 3D T1 spoiled saturation recovery mapping technique for acquisition of the entire LV in a single breath-hold at 3T. The proposed technique combines flexible saturation time sampling, compressed SENSE, and sharing of the saturation pulses between two readouts acquired at different RR intervals. This approach aims to achieve good quality single breath-hold saturation recovery 3D T1 mapping and stability over a wide range of heart rates (HRs).
Discussion
3D SACORA (3D saturation recovery compressed SENSE rapid acquisition) was developed as a new 3D T1 mapping sequence to speed up T1 mapping acquisition of the whole heart. The proposed sequence was validated in phantoms against the gold standard technique (IR-SE) and in vivo against the previously proposed sequence 3D SASHA.
3D SACORA successfully acquired a whole-heart 3D T1 map in a single breath-hold at 3T, estimating T1 values in agreement with those obtained with the IR-SE and 3D SASHA sequences. Thus, 3D SACORA’s use of 5 saturation times for T1 fitting as well as k-space under-sampling via compressed SENSE enabled very short acquisition times (15s, for a heart rate of 60 bpm) without significantly compromising T1 estimation accuracy or image quality.
3D saturation recovery T1 mapping sequences have been developed recently because they do not require full longitudinal magnetization recovery and produce highly accurate T1 values. Acquisition times with these 3D sequences are, however, much longer than 20 s, and scans, therefore, cannot be performed under single breath-hold conditions. 3D SACORA was optimized to keep scan time shorter than 20 s without compromising T1 estimation accuracy, image quality, or LV coverage. Conditions established for this optimization included (i) allowing enough time for the PD to achieve full magnetization recovery between readouts, (ii) limiting the number of turbo field echo shots as much as possible, (iii) optimizing trade-off between readout length and the compressed SENSE factor, (iv) acquiring T1-weighted images with enough SNR for proper application of compressed SENSE, and (v) acquiring a T1-weighted image with a saturation time as close as possible to the native cardiac T1 values at 3T.
3D SACORA is able to acquire a 3D T1 map in 15 heart beats (heart rate, 60 bpm) at 3T. The main constraint for acquiring T1 maps with a short scan time at 3T is the recovery beats required by the proton density. To mitigate this, 3D SACORA was designed to have just two readout shots. This was achieved by combining a shot length of ~210 ms with a compressed SENSE factor of 4.5, providing good image quality without major deterioration or blurring. A compressed SENSE factor of 4.5 guarantees good T1 estimation accuracy and precision, with comparable results to lower compressed SENSE factors, as shown in Fig. S3. A shot length of ~210 ms is similar to the conventional 2D MOLLI shot length of ~190 ms [
4] and shorter than the diastolic time length at very high heart rates (250 ± 59 ms at 128 ± 22 bpm) [
34]. The PD readouts were separated by 6 s to guarantee full magnetization recovery. 3D SACORA acquires two T1-weighted images with long saturation times to enhance fitting quality for relevant cardiac T1 values at 3T. In addition, these images are ideal for applying compressed SENSE due to their high SNR; whereas, T1-weighted images with shorter saturation times tend to be noisier, as shown in Fig.
4.
3D SACORA and our own implementation of 3D SASHA were validated in phantom experiments. 3D SASHA is the reference sequence for in vivo experiments in this study and has an acquisition time roughly 13 times longer than 3D SACORA. The phantom results show that 3D SACORA and 3D SASHA acquired 3D T1 maps with high accuracy and precision and in good agreement with IR-SE measurements. However, slight differences were found between 3D SACORA and 3D SASHA in specific cases (Fig.
3). First, 3D SASHA is more precise than 3D SACORA for short T1s, due to the denser sampling of short saturation time T1-weighted images present in 3D SASHA. Second, 3D SACORA T1 estimation is, by design, less heart-rate sensitive than 3D SASHA. Heart-rate sensitivity in 3D SASHA acquisition is due to its lack of long saturation images, which can undermine T1 estimation quality, especially at high heart rates. Thus, the 3D SACORA sampling strategy was validated over a wide range of heart rates (50 bpm–120 bpm). For heart rates outside this range, the sequence keeps acquiring saturation times at the low saturation time area of the T1 relaxation curve and close to the T1s of interest, as shown in Table S1. This sampling strategy of the saturation times makes 3D SACORA robust for a very wide range of heart rates.
The pig heart is an established model in cardiology due to its similarity to the human heart [
35,
36]. In this study, we acquire in vivo data before and after contrast administration. The image quality in 3D SACORA was close to that obtained with 3D SASHA despite the much shorter acquisition time. All septal T1 measurements were similar in the two sequences (Fig.
6a, b), despite the differences in sequence design and protocol. Furthermore, the mean septal pre-contrast T1 of 1453 ms estimated by 3D SACORA is in good agreement with published saturation recovery T1 measurements in pigs at 3T [
29]. There was a slight discrepancy between 3D SACORA and 3D SASHA in T1s measured in the lateral and anterior segments of the myocardium (Fig.
5a, b), probably caused by movement artifacts, which were more frequent in 3D SASHA due to the longer acquisition time.
In pigs, cardiac acquisitions performed in free-breathing are free of major respiratory artifacts [
36,
37]. This was crucial for the successful comparison of 3D SACORA with 3D SASHA in a model without respiratory motion compensation (e.g., respiratory navigator [
12] or motion correction [
38]) or breath-hold acquisition. Nevertheless, the results might be improved by taking appropriate measures to minimize respiratory artifacts. For example, T1 map quality could be improved by reducing respiration-induced motion using registration approaches such as non-rigid image registration [
39].
In this study, we used an in-house version of 3D SASHA for 3T [
11]. As the reference sequence for the in vivo experiments, the sampling strategy of the saturation times was similar to conventional 3D SASHA to acquire gold standard in vivo T1 values. In addition, 3D SASHA was implemented with a shorter shot length than 3D SACORA, minimizing the effect of cardiac motion and reducing partial volume averaging, which are especially relevant for high heart rates. This sequence was successfully validated against IR-SE with phantom experiments, and the in vivo results were in good agreement with published saturation recovery T1 measurements at 3T [
12,
29].
Despite the good performance of 3D SACORA in post-contrast imaging, the sequence was primarily designed for pre-contrast imaging. Although post-contrast T1 values obtained with the proposed technique do not differ significantly from those obtained with 3D SASHA, the accuracy and precision of short T1 could be improved by increasing the number of short saturation images. The additional acquisition time could be compensated by decreasing the number of PD recovery beats.
One of the main limitations of this feasibility study is the lack of in vivo human data. Nevertheless, pig model is a well-stablished model for cardiac research and in vivo pig acquisitions can be performed under free-breathing conditions without significantly decreasing image quality [
35‐
37]. Additionally, feasibility studies of new cardiac acquisitions have been successfully performed on pigs [
40‐
43]. A clinical study will be required to evaluate the performance of 3D SACORA (single breath-hold of 15s, for a heart rate of 60 bpm) in patients with ischemic and non-ischemic cardiomyopathies.
In conclusion, the proposed 3D SACORA sequence acquired pre-contrast and post-contrast T1 maps of the whole heart with good accuracy, precision, and image quality for LV analysis at 3T. The sequence was optimized for speed and can acquire a 3D T1map in 15 heart beats for a heart rate of 60 bpm.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.