Background
Late gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) imaging has become the gold standard for the assessment of myocardial viability in different cardiac pathologies, including myocardial infarction [
1,
2] and myocarditis [
3‐
5]. In addition, LGE imaging provides pre-interventional assessment of arrhythmogenic substrate in patients undergoing electrophysiology procedures as well as visualization of lesions after ablation [
6‐
8], and is gaining importance in the characterization of fibrosis in non-ischemic cardiomyopathies [
9‐
12]. LGE imaging is typically performed 10–20 min after the administration of a gadolinium (Gd)-based contrast agent using T
1-weighted inversion recovery (IR) sequences [
1,
13‐
15]. The inversion time (TI) is normally set to null the signal from the healthy myocardium, thus enhancing the contrast to noise ratio (CNR) between viable and diseased myocardial tissue. IR sequences, however, are prone to reduced scar to blood and scar to remote myocardium contrast when a sub-optimal TI is selected. Phase-sensitive IR (PSIR) LGE acquisitions have been introduced to provide intrinsic robustness with respect to the TI selection [
16]. Conventional PSIR sequences are based on the acquisition of an IR-prepared image (referred to as “magnitude image”), interleaved with a proton density image (referred to as “reference image”) that is acquired at a low flip-angle, which are then combined as described in [
16]. Although PSIR normally achieves excellent contrast between viable myocardium and scar tissue, the contrast between blood pool and LGE uptake is often suboptimal. This leads to difficulties in delineating sub-endocardial infarcts that are adjacent to the blood pool. Furthermore, unclear borders between scar tissue and blood affect the accuracy of scar segmentation that is crucial for infarct size and transmurality measurements as well as for the planning of electrophysiology procedures [
17,
18]. Black-blood PSIR LGE has been introduced [
19] to improve the contrast between the blood pool and scar tissue by exploiting an inversion pulse in combination with a T
2 preparation (T
2Prep) module (T
2Prep-IR) [
19‐
21]. However, a limitation of all PSIR frameworks is that the acquisition efficiency is intrinsically sub-optimal as the low flip-angle reference image has limited diagnostic value. Furthermore, most of the LGE PSIR implementations are limited to 2D acquisitions that are performed during a breath-hold to minimize respiratory motion artefacts. Recently, free-breathing whole-heart PSIR acquisitions have been introduced [
22,
23] and integrated with diaphragmatic navigator gating [
24]. The use of diaphragmatic navigator gating, however, leads to reduced scan efficiency and unpredictable acquisition times that can make the selection of the correct TI challenging. In addition, residual imaging artefacts may be observed as a result of the combination of the inversion pulse with the diaphragmatic navigator [
25]. In order to overcome these drawbacks, we propose the extension of a 3D whole-heart Bright-blood and black-blOOd phase SensiTive inversion recovery (BOOST) sequence [
26] – that has been recently introduced for non-contrast enhanced visualization of coronary lumen and thrombus – to black-blood LGE imaging. The proposed post-contrast BOOST sequence exploits a T
2Prep-IR module for the acquisition of the magnitude image, enabling black-blood LGE PSIR reconstruction. Furthermore, the acquisition of the reference image is designed to provide a complementary and fully co-registered bright-blood dataset for the visualization of the heart anatomy, the great vessels, and the coronary lumen. The entire framework has been integrated with image-based navigation [
27] to achieve 100% scan efficiency and predictable scan time. In this study, the feasibility of BOOST for post-contrast simultaneous black-blood LGE imaging and bright-blood heart anatomy, great vessels, and coronary lumen visualization was tested in a cohort of cardiovascular patients at the end of a clinical CMR examination.
Discussion
In this study, we extended the use of a novel PSIR-like framework, referred to as BOOST, to post-contrast applications for simultaneous 1) black-blood LGE assessment and 2) bright-blood heart anatomy, great vessels, and coronary lumen visualization. With the BOOST framework, the acquisition of the magnitude image (T
2Prep-IR BOOST) is based on a T
2Prep-IR module for optimal contrast between the blood pool and scar tissue after PSIR computation (black-blood PSIR BOOST). Furthermore, the acquisition of the reference image (bright-blood T
2Prep BOOST) is performed with a high flip-angle and it is preceded by a T
2Prep module. This ensures adequate signal and tissue contrast for the visualization of heart anatomy, great vessels, and the coronary lumen. In contrast to previously published approaches providing a single bright-blood dataset for the simultaneous visualization of LGE and proximal coronary arteries [
40], our framework generates two separate yet co-registered 3D volumes, each one being specifically designed and optimized for the visualization of the coronary lumen (bright-blood T
2Prep BOOST) and myocardial scar (black-blood PSIR BOOST).
Sequence simulations and phantom acquisitions showed that the proposed post-contrast PSIR BOOST dataset achieves improved scar-blood contrast when compared to a more conventional PSIR sequence for LGE imaging [
16]; this was confirmed by in vivo measurements in patients. While the PSIR BOOST volume provided adequate LGE depiction in most of the patients with positive findings, phantom experiments indicate higher CNR
scar-myo in the T
2Prep-IR BOOST datasets, where precise viable myocardial nulling is achieved; this can be qualitatively appreciated in vivo as shown in Fig.
5. As such, referring to the T
2Prep-IR BOOST dataset for the detection of subtle, non-ischemic, fibrosis patterns might be preferable; this aspect, however, needs further investigation and will be analyzed in future studies. Furthermore, sequence simulations show that the bright-blood T
2Prep BOOST dataset provides SNR
blood and CNR
blood-myo similar to those of a dedicated T
2-prepared post-contrast CMRA acquisition. In vivo acquisitions showed that respiratory motion corrected bright-blood T
2Prep BOOST datasets allowed visualization of the origin and the proximal course of the coronary arteries (LM, LAD, LCX, and RCA) with high diagnostic quality. A trend of improvement was observed in comparison to the conventional CMRA; respiratory motion compensation performed with diaphragmatic navigator gating assumes a fixed linear correlation between the respiratory motion of the liver and that of the heart. The fixed correlation factor of 0.6 [
37] that was used in this study might have been inexact for some of the subjects, thus leading to sub-optimal motion compensation. Conversely, with the use of image-based navigation, respiratory motion information can be directly extracted from the heart itself, thus avoiding the risk of imprecise approximations. In addition, with image-based navigation, it is possible to correct for movements along both SI and RL directions [
27]. These aspects may have been a contributing factor of the improved coronary delineation that was obtained with BOOST. Furthermore, and as predicted by sequence simulations, the black-blood PSIR BOOST reconstruction provided visualization of LGE with diagnostic quality in most cases and significantly improved CNR
scar-blood was quantified in comparison to clinical 2D PSIR acquisitions [
16].
The PSIR reconstruction performed with the proposed framework exactly follows that described in [
16], with the exception of the intensity normalization step that is conventionally performed at the end of the PSIR pipeline. In contrast to previously published post-contrast PSIR sequences [
16], the reference image (T
2Prep BOOST) acquired in our approach exhibits high tissue contrast, thus preventing the application of surface coil intensity normalization. In fact, the presence of high tissue contrast in the reference image significantly alters the resulting contrast of the normalized PSIR reconstruction (Fig.
3). The use of surface coil intensity normalization is typically exploited to compensate for large variations in the intensity of the image caused by rapid fall-off of the surface-coil fields, thus improving the local tissue contrast. This was shown to be particularly beneficial for the visualization of subendocardial infarcts, given the fact that the contrast between scar tissue and blood is particularly reduced in conventional PSIR acquisitions [
16]. With this new sequence configuration, however, intrinsically enhanced contrast between blood and scar tissue is provided using a T
2Prep-IR module for the acquisition of the magnitude image; in addition, the use of pre-scan based normalization readily available on commercial scanners can be exploited to compensate for variations in signal intensity. This might alleviate the need for surface coil intensity normalization, however further validation may be needed to corroborate this point.
The integration of the framework with image-based navigation enabled data acquisition during free-breathing with 100% scan efficiency and predictable scan time. The acquisition time for BOOST (approximately 12 min) was similar to that of a conventional CMRA acquisition with diaphragmatic navigator (approximately 13 min), considering an average scan efficiency of 50% and 2× parallel imaging acceleration. The BOOST framework, however, provides both a bright- and black-blood dataset in the same acquisition time, whereas the overall acquisition of conventional CMRA and 2D PSIR sequences was about 20 min in our cohort of patients. This intrinsic efficiency of the BOOST framework holds potentials for reducing the scan time that is currently needed to perform a complete CMR examination. This might be particularly beneficial in the case of claustrophobic, anxious, or clinically unstable patients. Additionally, reducing the overall examination time would imply economic benefits and reduction of patients waiting lists. Future technical developments of the BOOST sequence will include the integration of acceleration techniques [
36,
41,
42] to improve both the nominal acquisition time as well as the spatial resolution. Furthermore, improvements in the acquired spatial resolution might enable isotropic acquisitions that would, for instance, allow for more robust visualization of the mid and distal coronary arteries. Similarly, the achievement of higher spatial resolution could benefit tissue characterization, allowing for a more accurate delineation of scar tissue and enabling a more accurate image fusion between the bright-blood T
2Prep BOOST and the black-blood PSIR BOOST datasets for the assessment of scar location and transmurality. Additionally, the framework will be integrated with algorithms for arrhythmia rejection that could further improve the image quality that was obtained in this study. Currently, BOOST is combined with image-based navigation enabling in-line translational motion correction along the SI and RL directions. However, the breathing pattern in patients is often more complex and involves translation, rotation, and non-rigid deformations [
43‐
45]. Therefore, future technical developments will aim at combining the BOOST framework with strategies for non-rigid respiratory motion correction [
46] that might be particularly beneficial in very sick patients who often have irregular breathing patterns [
47]. In addition, the use of non-rigid respiratory motion correction may help to reduce ghosting artefacts that may originate from rigid translation of static tissues such as the chest wall and arms during the motion correction process. Similarly, a rigid registration between the T
2Prep-IR BOOST dataset and the T
2Prep BOOST dataset is currently performed prior PSIR computation to compensate for residual mis-registration errors; this may be also sub-optimal and the use of non-rigid registration could further improve the quality of the resulting PSIR BOOST dataset and additionally reduce the risk of phase errors that may originate in portions of the image where phases are not varying smoothly (e.g. in correspondence to the interface between different tissues).
In this study, the BOOST acquisition was performed at the end of a clinical CMR examination as it was considered unethical to potentially jeopardize the acquisition of conventional LGE data at the expense of a novel sequence at this stage. Injection timing was optimized to provide optimal contrast agent concentration during the acquisition of the clinical 2D PSIR sequences, thus providing suboptimal contrast conditions for the BOOST scan. This was noticed particularly in two specific cases (Patient 08 and Patient 10), where LGE uptake could not be depicted despite the absence of motion artefacts and the achievement of optimal blood signal suppression. Furthermore, as the BOOST acquisition was performed at the end of the clinical scan, there may have been more respiratory or heart-rate irregularities that might have had an additional detrimental effect on the image quality that was obtained with BOOST. However, scan time was not prolonged by more than 15 min at most. Therefore, future studies are warranted to rigorously compare the proposed post-contrast 3D BOOST sequence and conventional 2D PSIR acquisitions by, for instance, randomizing the order of the two acquisitions and by performing separate Gd injections to ensure equivalent contrast conditions. Kellman et al. [
19] demonstrated that black-blood LGE provides improved conspicuity of subendocardial infarcts; future studies will aim at investigating the accuracy of black-blood PSIR BOOST for the quantification of scar transmurality and, thus, regional viability assessment. Furthermore, accuracy in the detection and quantification of ischaemic scar will be validated. Similarly, further clinical validation of BOOST is needed in comparison to conventional CMRA in patients with angiographically confirmed coronary artery disease.
The improved contrast between blood pool and scar provided by the proposed black-blood PSIR BOOST images may facilitate scar segmentation. However, the nulling of the blood and viable myocardium signal reduces the depiction of the heart anatomy and might challenge the localization of the scar itself. This challenge can be addressed by fusing the co-registered black blood PSIR dataset with the bright blood whole heart dataset, which then allows both scar and myocardial anatomy visualization as shown in Fig.
7. This characteristic makes this framework particularly suitable for the planning of electrophysiology procedures. Similarly, the framework may be beneficial for the visualization of lesions after ablation and follow-up of patients. This further enlarges the spectrum of potential clinical applications of post-contrast BOOST, that will be tested in dedicated studies in the upcoming future.