Background
Detection of the location and extent of myocardial scarring is important for the prognosis of patients with myocardial remodeling [
1]. Post-infarct formation of myocardial fibrosis can lead to adverse cardiac remodeling and subsequently, to heart failure. The current
in vivo reference standard for detection of myocardial scar tissue is late gadolinium enhancement (LGE), where the prolonged retention of gadolinium contrast agent (CA) in regions of myocardial fibrosis results in increased signal intensity on T
1 weighted cardiovascular magnetic resonance (CMR) images. The presence and extent of LGE CMR carries important prognostic value as has been demonstrated in several ischemic and non-ischemic cardiomyopathies [
2]–[
4].
LGE is a validated method with a high sensitivity to discriminate infarcted from healthy myocardium. However, to guide and evaluate medical treatment, more information is needed about the heterogeneity of myocardial damage associated with diverse cardiac disease processes, as this damage is the substrate of arrhythmias, and a possible target for CMR-guided arrhythmia ablation. In order to provide this information, a shift from mere visualization to quantification of myocardial fibrosis is needed. Furthermore, a truly noninvasive method based on endogenous contrast without the requirement for an exogenous contrast agent would be preferable, since gadolinium enhanced CMR with some agents is off-label use [
5]. Allergic reactions after intravenous administration of gadolinium-based contrast agents are very rare but potentially life threatening [
6],[
7], and the method cannot be applied in patients with severe renal failure [
8]. A quantitative method capable of detecting myocardial fibrosis based on endogenous MR characteristics of the myocardium could, therefore, be a valuable tool, complementary to LGE [
9].
In the field of orthopedics, the MR relaxation parameter T
1 in the rotating frame (T
1ρ) is well established as measure for the collagen content in cartilage [
10],[
11]. Recent studies have hypothesized that this method may be applicable to directly image collagen in the heart, and therefore would be a promising candidate for detection of chronic myocardial infarction [
12],[
13]. The T
1ρ relaxation time describes relaxation while the magnetization is in the rotating frame, in the presence of a so-called spin-lock pulse. A spin-lock pulse is a low amplitude radiofrequency (RF) pulse on-resonance with the precessing transverse magnetization. By acquiring images with varying T
1ρ weighting, a so-called T
1ρ-map can be calculated.
In
ex vivo MI tissue, it has been shown that the T
1ρ relaxation time is sensitive to changes in macromolecular content, and that a significantly higher T
1ρ is found in the MI region [
12]. Studies in animal models of chronic MI showed the first
in vivo evidence for the ability to detect myocardial fibrosis with T
1ρ-mapping [
14],[
15]. Furthermore it has been shown that the addition of T
1ρ weighting to a conventional gradient echo sequence improves the contrast between acutely infarcted and noninfarcted myocardium in patients in an acute (63 ± 40 hours) myocardial infarction [
13]. However, this method has not yet been reported for assessment of chronic MI in humans.
The aim of the current study is to explore the potential of T1ρ-mapping to detect fibrosis in patients with chronic ischemic heart disease on a standard clinical MR scanner. Therefore, we first performed a study in a porcine animal model of chronic MI to reproduce initial results in literature and to validate the implementation of the technique. In a second study the method was translated to a clinical protocol, and the feasibility of detecting scar tissue with native T1ρ-mapping in patients with chronic MI was assessed and correlated with LGE CMR.
CMR methods
All subjects were imaged on a clinical 1.5 T MR scanner (Achieva, Philips Healthcare, Best, the Netherlands), using a 5-channel cardiac receive coil. T
1ρ-mapping was performed using a 2D T
1ρ-prepared balanced steady-state free precession (SSFP) gradient echo sequence, using the same spin-lock preparation scheme as in the animal study (Figure
1). In the first three patients a multi-shot gradient echo readout instead of a balanced SSFP readout was used, due to further optimization of the protocol. The amplitude of the spin-lock pulse was set to 750 Hz (17.6 μT), and four images with different spin-lock preparation times were acquired (SL = 1, 13, 27, 45 ms). Other parameters were: bandwidth/pixel = 530 Hz, TE/TR = 1.94/3.9 ms, resolution = 1.5 × 1.65 mm, slice thickness = 6 mm, slice gap = 4 mm, FOV = 288 × 288 mm
2, flip angle = 50 degrees, 2 TFE shots, NSA = 2, SENSE acceleration = 1.5, shot interval = 3 heart beats. In all participants, the T
1ρ-mapping was performed in 8 short axis slices, acquired in late diastole during expiration breath holds, covering the heart from apex to base.
In the patients, LGE CMR was performed in the same 8 short axis slices, 15 minutes after intravenous contrast injection (0.2 ml/kg contrast agent (Gadovist, Bayer Healthcare). A look-locker scout was performed to optimize nulling of the remote myocardium. LGE imaging parameters were: TI = 300–340 ms, TE/TR = 3.5/7.1 ms, resolution = 1.5 × 1.65 mm2, slice thickness = 8 mm, slice gap = 4 mm, FOV = 288 × 288 mm2, flip angle = 25 degrees, 5 shots).
Post processing and image analysis
For both the animal and the patient studies, T1ρ-maps were calculated by pixelwise fitting of a mono-exponential decay function in Matlab (Release 2012a, Mathworks, Massachusetts, United States).
The LGE images and T1ρ maps were scored by a radiologist (TL) with over 10 years experience in the evaluation of CMR. Images were scored for image quality, location of the infarct, and transmurality.
The LGE images and T
1ρ maps were scored separately using the 17 segments AHA-model [
18]. The T
1ρ-maps were scored two times by the observer. The first time the observer was untrained, and blinded for clinical characteristics, results of conventional LGE imaging and the LGE images were not shown. The second time the observer was trained for looking at T
1ρ-maps, and the LGE images were shown along with the T
1ρ-maps, but the observer was blinded for clinical characteristics and the results on the scoring of the LGE images.
Segmentation of infarct and remote myocardium was based on the 2SD segmentation method on the LGE images, and this mask was then applied on the corresponding T
1ρ-maps to calculate T
1ρ-values for infarct and remote myocardium [
19].
Statistics
Statistical analysis was performed with GraphPad Prism (GraphPad Software, California, United States). Group comparison was performed using a two-way ANOVA analysis, and considered significant at p < 0.05.
Discussion
To our knowledge, this is the first report of
in vivo detection of chronic myocardial infarction in patients using native T
1ρ –mapping without the use of gadolinium contrast agents. Areas of myocardial fibrosis as identified with this approach corresponded reasonably well with conventional LGE images (Figure
6). While the results with the current implementation are promising, sensitivity and specificity is lower compared to the LGE method. We expect that improvements to the implementation will solve this issue, as discussed below. Since T
1ρ -mapping requires no contrast agent, it provides a truly noninvasive method, and therefore has the potential to become an alternative to the LGE method in patients with severe renal failure who are unable to receive a contrast agent.
In the animal model we found a significantly higher T
1ρ relaxation in the infarct region, compared to healthy myocardium (Figure
2). Histology showed that the collagen fraction in this area was higher compared to macroscopically normal remote myocardium. These results are in accordance with previous findings in other animal studies [
12],[
15]. Similar to the findings in the animal model, in patients with a chronic myocardial infarction we also found a significant higher T
1ρ value in the infarct area. The T
1ρ relaxation times found in patients were higher than in the animal model, which could be caused by multiple factors. It is known that the T
1ρ relaxation time depends on the strength of the main B
0 field, and decreases with a higher B
0[
20]. Also, the higher amplitude for the spin-lock B
1 pulse in the patient study leads to higher values for T
1ρ, as we also found in the T
1ρ dispersion results (Figure
4). Finally we used an increased trigger interval of 3 beats in the patient study, which enabled a better estimation of the true T
1ρ value because of reduced T
1 weighting due to incomplete relaxation of the magnetization prior to the next acquisition.
The
ex vivo T
1ρ dispersion results showed that the T
1ρ contrast between healthy myocardium and scar tissue increases with a higher B1 amplitude of the spin-lock pulse (Figure
4). This implies that for T
1ρ –mapping we should aim for the highest possible B
1 amplitude. However, on a clinical MR scanner the maximum B
1 amplitude is limited by the specific absorption rate (SAR) and the performance of the hardware (transmit coil and RF amplifiers). To stay within human SAR limits, a spin-lock amplitude of 500 Hz (11.7 μT) was used at a field strength of 3 T in the animal experiment. The patient study was performed on a 1.5 T system, enabling a higher B
1 amplitude for the spin-lock pulse of 750 Hz (17.6 μT), because of lower SAR values and a higher B
1 field available. Another reason to perform the patient study on a 1.5 T system, is that artefacts caused by B
1 and B
0 inhomogeneities, as can be seen in Figure
3, are reduced on a lower field strength. The dispersion data suggests that an even higher B
1 amplitude for the spin-lock pulse would generate more contrast between healthy and fibrotic myocardium. New CMR contrast developments such as relaxation along a fictitious field (RAFF) may have the potential to generate more contrast between infarct and remote tissue for a given B1 amplitude and SAR level [
21].
Double-blinded qualitative scoring of the LGE images and T
1ρ maps in patients was performed to investigate if the native T
1ρ maps can be used to assess the presence and location of myocardial scar accurately. We found a 72% agreement between both methods in the patients (Table
2a). However compared with
in vivo gold standard LGE imaging, the sensitivity of T
1ρ –mapping to detect scar tissue was found to be lower using the present implementation.
The most important reason for the lower sensitivity is that the contrast to noise ratio between healthy myocardium and scar tissue is much higher in LGE imaging. Although the difference in native T
1ρ between healthy and infarct tissue is significant, especially smaller infarcts might be more difficult to detect with this method. The infarct size of the patients in this study was small, which can also be concluded from the mean LV ejection fraction of 55.7 ± 7.4% [
22].
Important to keep in mind is that the underlying principle to discriminate myocardial scar tissue from normal myocardium with LGE imaging and T
1ρ mapping is different. In LGE imaging the difference in contrast agent washout between normal and diseased myocardium is used to identify scar tissue, which reflects changes in perfusion and extracellular volume in the scar area [
23]. On the other hand, native T
1ρ mapping directly measures the effect of tissue damage and scar tissue formation on the T
1ρ relaxation time. T
1ρ is known to be sensitive to changes in macromolecular content, and the histology results show that in the infarct region both a significant increase in T
1ρ time and fibrosis percentage is found. It is unknown, however, if the increase in T
1ρ directly reflects an increase of collagen in scar tissue, or that other changes in tissue composition after MI are involved. Studies in cartilage and protein solutions suggest that other factors such as cellular content and exchange might be involved, since in agarose gels an increase in macromolecule content leads to a decrease in T
1ρ time, which is in the opposite direction of our findings [
24],[
25]. Further research should be performed on the mechanism and relation between myocardial fibrosis formation and the myocardial T
1ρ relaxation time.
Another reason for the higher sensitivity of LGE, can be that the radiologist is trained to assess myocardial scar on LGE images, but has no experience in looking at T1ρ maps. After the T1ρ maps were scored double blinded, the radiologist was trained to look at T1ρ maps, and scored again with the LGE images shown alongside. This resulted in a significant increase of sensitivity for the detection of infarct area with T1ρ mapping compared to LGE imaging. We also observed that endocardial infarcts were more difficult to detect on T1ρ maps, since partial volume effects, combined with the high T1ρ relaxation time of the blood, make it difficult to distinguish the transition from myocardium to blood on a T1ρ map. Furthermore, since we used multiple breath holds to calculate a T1ρ map, misregistration between the different breath hold positions leads to problems with the calculation of the T1ρ map at the edges of the myocardium.
Future work should aim to overcome these limitations that reduced image quality and assessability of the T1ρ maps in our study. Most important step is the development of a single breath hold T1ρ mapping sequence to obtain a high quality T1ρ map, without registration errors due to multiple breath hold acquisitions. This requires faster cardiac T1ρ mapping sequences, using acceleration methods. Furthermore, a black-blood readout will enable the detection of an endocardial infarct close to the blood, by reducing partial volume effects.
One of the main drawbacks of LGE imaging is the lack of the ability to measure myocardial fibrosis quantitatively. Currently there is a lot of interest in quantitative imaging of myocardial tissue to overcome these limitations, by measuring quantitative contrast enhanced T
1-maps and extracellular volume (ECV)-maps [
26],[
27]. We believe that native T
1ρ –mapping could provide additional quantitative information on myocardial fibrosis in cardiomyopathies, also in patients with diffuse interstitial myocardial fibrosis. Native T
1ρ –mapping requires no separate pre- and post-contrast scan, no hematocrit measurement, and is therefore easier to incorporate in a clinical protocol, compared to ECV-mapping. Here we have shown the first evidence that T
1ρ mapping can provide quantitative information on myocardial fibrosis in patients with a chronic myocardial infarction. Though speculative, the finding that the remote myocardium of the patients was about 1 SD above the T
1ρ value of the healthy subjects, may suggest a slight increase in diffuse interstitial fibrosis in this area, which could become significant if a larger group with more statistical power would be studied. Future work should focus on further validation of the relation between myocardial fibrosis and T
1ρ, and on the relation with native and contrast-enhanced T
1 and ECV-mapping.
Competing interests
FV is an employee of Philips Healthcare, Best, The Netherlands.
Authors’ contributions
JG, SJ, SC performed the LAD occlusion and histology. JvO, MF, JZ, FV performed the animal CMR experiments and processed the CMR data. JvO, HeA, JZ, FV performed the patient CMR experiments and processed the CMR data. TL scored the CMR data. All authors read and approved the final manuscript.