|Home | About | Journals | Submit | Contact Us | Français|
Cardiac magnetic resonance (CMR) imaging has been widely used to assess myocardial perfusion and scar and is the non-invasive gold standard for identification of focal myocardial fibrosis. However, the late gadolinium enhancement technique is limited in its accuracy for absolute quantification and assessment of diffuse myocardial fibrosis by technical and pathophysiological features. CMR relaxometry, incorporating T1 mapping, has emerged as an accurate, reproducible, highly sensitive, and quantitative technique for the assessment of diffuse myocardial fibrosis in a number of disease states. We comprehensively review the physics behind CMR relaxometry, the evidence base, and the clinical applications of this emerging technique.
Cardiac magnetic resonance (CMR) imaging has been used widely to assess myocardial perfusion and scar (1–5). It is the non-invasive gold standard for left and right ventricular quantitation, as well as the assessment and quantitation of focal myocardial fibrosis (after infarction or due to other causes of cellular injury). Myocardial necrosis causes high signal on late gadolinium enhancement (LGE) inversion recovery (IR) T1-weighted images with excellent signal-noise ratios, and this has become the reference standard for non-invasive scar imaging in cardiomyopathies of various causes (1–4). However, LGE is limited in its ability to assess and quantitate diffuse (non-focal) myocardial injury and fibrosis. LGE is affected by inconsistencies in acquisition parameters, such as choice inversion time (TI), and in post-processing when signal intensity (SI) thresholds may be arbitrarily applied to distinguish normal myocardium from fibrotic tissue (6, 7). Moreover, the critical issue with LGE is that SI is expressed on an arbitrary scale (relative SI compared to “nulled” normal myocardium). Imaging of myocardial fibrosis using relative differences between scar and normal myocardium tissue is therefore qualitative. Semi-quantitative analysis of LGE can be performed using signal thresholding applied to LGE images; however, there are differences in technique for infarct quantitation (8), and this is only relevant when regional scar/enhancement is present; it does not allow quantitation of diffuse interstitial fibrosis.
Thus, in non-ischemic cardiomyopathies, such as hypertension or diabetes, LGE CMR is unable to detect signal differential where the collagen deposition is diffuse and widespread throughout the myocardium (9).
Cardiac magnetic resonance is an evolving technique, providing valuable and comprehensive data on the anatomy and functional integrity of both the heart and coronary blood vessels. Currently, CMR is performed at magnetic field strengths of 1.5 or 3T. MR images generate by exploiting the magnetic property (called spin) of nuclei that have an odd atomic number or mass number (10). A proton generates a small magnetic field much like a bar magnet, because the proton has mass, a positive charge, and spins. This small magnetic field is referred to as its magnetic moment. The single proton of the hydrogen molecule gives it a significant magnetic moment and combined with its abundance in the human body, makes it an ideal marker for clinical MRI.
In the absence of an applied magnetic field, the magnetic moments of the hydrogen nuclei are oriented randomly; when placed in a high static magnetic field (B0) they will align either parallel or anti-parallel to the magnetic field. Spins that aligned parallel to B0 have a lower energy than those aligned anti-parallel, and therefore more align parallel creating a net magnetization (M0) of the sample in the direction of the magnetic field B0 (11).
The interaction of a magnetic moment with B0 causes the magnetic moment to precess about the axis of the static magnetic field (B0), at a frequency specific to the strength of (B0)— the Larmor frequency. The Larmor frequency is defined as follows:
where ν is the frequency, in megahertz, B0 is the strength of the magnetic field, γ is the gyromagnetic ratio for hydrogen, and γ/2π=42.57MHz/T (11).
When a radio frequency (RF) pulse at the Larmor frequency is applied to the nuclei within the magnetic field, nuclei begin to resonate and those in the lower energy state absorb energy. Depending on the RF pulse length, the precession of affected nuclei will be moved into the transverse magnetization plane (xy axis) and be in phase. With cessation of the RF field, the nuclei will realign to their original orientation parallel to B0—a process referred to as relaxation. During the relaxation process, the net relaxation induces an RF signal, at the characteristic frequency, which can be measured by a receiver coil. This signal is known as free induction decay (FID).
Three different properties of the interaction of the magnetic moments with B0 can be measured. These are the longitudinal time constant T1, or “spin-lattice” relaxation, and transverse time constant T2, or “spin–spin” relaxation and , which is governed by a combination of the effect of spin–spin relaxation, and the homogeneity of the magnetic field. These constants are parameters that are used in MRI to distinguish between normal tissue types and pathological process. The SI of these times depends on the technical parameters that are used for image acquisition (12, 13) and the magnetic properties of a given tissue (14). At a given magnetic field strength, each tissue has a normal range for relaxation time. So, the variation of relaxation time from their normal value can be used to identify pathological process (e.g., edema and scar tissue).
T1 relaxation time refers to the tissue-specific time constant and is a measure of the time taken for protons to realign with the static field after perturbation by the RF pulse. This realignment with B0 is termed T1 longitudinal relaxation and the time in milliseconds. The above diagram T1 recovery curves demonstrate the fat has shorter T1 times than water molecules. This results from the fact that the fat nuclei lose their energy to lattice quickly, due to slow molecular motion, giving a relatively shorter T1 time. The quicker the system returns to the equilibrium state, such as occurs in fat, the greater the magnetization available to be excited by the next imaging pulse, producing more signals. Thus, fat appears bright in T1-weighted image (11).
T2 relaxation causes decay of signal arising from the dephasing of nuclear precession and consequent loss of net coherence (Figure (Figure1)1) (15). T2 decay due to the magnetic interaction that occurs between protons, results in an exponential decay of the transverse magnetization vector, also governed by tissue structure. Unlike T1 relaxations, T2 does not involve a transfer of energy but only a change in phase. The water molecules have a long T2 decay and appear brighter in T2-weighted images due to the property that more rapidly moving molecules have a lower tendency to transfer their spin leading to a slower dephasing of the transverse magnetization.
A third decay parameter, , describes the decay of transverse magnetization, due to spin–spin relaxation (T2), together with inhomogeneity in the static magnetic field (ΔB0), which occurs at tissue interfaces. This leads to a more rapid loss of phase coherence and the MR signal. These relaxation times are influenced by several factors, including field strength, blood iron content, blood volume, temperature, and blood oxygenation (15).
The conventional T1 mapping method can be generated with an inversion recovery spin echo (IR-SE) sequence. Spin echo uses a 180°RF pulse to invert the spins within the selected slice, followed by a 90°RF pulse rotate the recovered magnetization at TI into the transverse plane, and a further 180° RF pulse to form the spin echo (12). Sampling of IR curve repeats multiple times with different TIs. The repetition time (TR) of the sequence must be long enough for recovery longitudinal magnetization before the next 180°inversion pulse. The SI of the final image is proportional to the relaxed fraction of the magnetization during the TI as follows:
where the first exponential term relates to signal decay due to transverse relaxation (during TE), the second exponential refers to T1 relaxation (during TI). IR-SE is a gold standard to estimate T1 with a good accuracy. However, the drawback of IR-SE is that the TR of the sequence should be very long, i.e., TR should be at least five times longer T1 to allow full recovery of longitudinal magnetization. To reduce the acquisition time, inversion recovery turbo spin echo is currently used for routine clinical purposes (12).
On traditional T1-weighted images, the focal differences in T1 signal are measured qualitatively, assessed by visual inspection using relative units, and cannot be consistently compared between scans (16). For more precise measurement, an emerging technique for CMR T1 mapping has been applied to measure myocardial signal (in milliseconds) directly on a standardized scale within a single breath hold. This quantification of T1 mapping provides a benefit over the T1-weighted imaging of the myocardium by giving a quantitative measure of T1 time from multiple scans. A parametric map can be then reconstructed by calculating T1 values on a pixel-by-pixel basis (17), so, pixel intensities correspond to T1 values. Different cardiac MR acquisition sequences have been applied to create myocardial T1 maps, including Look–Locker (LL) and modified Look–Locker inversion recovery (MOLLI) (9, 16, 18–24).
The most common technique to measure spin-lattice T1 relaxation time values is the eponymously named “LL” sequence (also known as “TI scout”). It has been widely used to estimate the optimal TI for assessment of myocardial LGE (23, 25). It was originally proposed by Look and Locker in 1968 and developed more fully in 1970 (24). It consists of an initial inversion pulse, followed by a train of pulses with a constant, limited flip angle (7–15°). The inversion pulse prepares the longitudinal magnetization, which then recovers exponentially according to the T1 (24). The experiment is repeated until the k-space Cartesian map of data is filled, allowing an image to be reconstructed. By producing a train of absorption or dispersion signals (continuous wave magnetic resonance) or FIDs (pulsed magnetic resonance), it is possible to save time in spin-lattice relaxation measurements due to the fact that it is not necessary to wait for equilibrium magnetization before initiating the train (24) (see Figure Figure2).2). The total time required to acquire T1 would be significantly reduced compared to IR because the LL technique allows for multiple of MZ in a single measurement period (TM).
It was co-opted by Kaptein et al. in 1976 to quickly sample the recovery after a preparation pulse, during the recovery period or transient phase (27). This method was developed into the T one by multiple readout pulses (TOMROP) imaging sequence (28). In TOMROP, the multiple samples of a particular recovery after RF preparation each correspond to separate image. To acquire a complete data set for each image, the whole sequence must be repeated numerous times. Each repetition fills the next line of k-space for each image, and so on. Each image has a unique delay time. Early LL-based T1 techniques required the return to equilibrium of the spin system before the next application of an RF preparation pulse. Consequently, the acquisition time per slice of such implementations was long (28). Hinson and Sobol (29) used an LL method with no preparation pulse but the method suffered from poor accuracy, attributed to the slice profile. The late 1980s and early 1990s saw the LL method used for T1 measurements in a number of publications. Crawley and Henkelman (30) compared a number of one-shot and IR methods (LL, saturation recovery, IR, and stimulated echo) and concluded that the LL method was almost as efficient (in terms of dynamic range of the data and the proportion of the imaging time used to sample MR signals) as IR. Brix and colleagues used the TOMROP method with 32 gradient echoes to test for non-exponential behavior, found in fatty tissues (31), in a total acquisition time of 4min. The LL single-shot IR method has been optimized and refined (32, 33) including improved RF preparation pulses (34, 35).
Echo-planar imaging (EPI) was incorporated into the IR LL-based method (36), by interleaving EPI readouts for eight different slices after an inversion pulse. The sequence was repeated, and the slice order was changed to achieve a range of TIs for each slice, with a total acquisition time of 30s. LL with EPI was later applied in vivo in less than 3s (37), using a modified blipped EPI technique (38), sacrificing and accuracy to some extent. An entire image was acquired at each point on a single recovery of longitudinal magnetization after a saturation pulse. The technique was optimized in 1998 (39) and has found applications in pharmacokinetic modeling (25).
The development of LL technique, which is available on Philips, GE, and Siemens platforms, is summarized in Table Table11.
The LL sequence has been widely applied in CMR due to its fast acquisition with minimal breath-hold requirements. The LL sequence has been used to measure T1 values in patients with myocardial fibrosis (23). However, it suffers from significant limitations: low flip angle RF pulse exciting the magnetization and the two RR intervals in the LL sequence are not sufficient for the magnetization to return to equilibrium. This causes underestimation of true T1 values using LL. Furthermore, the LL T1 images with different TIs are acquired at different cardiac phases. Therefore, images are “cine” with cardiac motion effect, which requires tedious manually tracking of the myocardial borders for each phase, a labor-intensive and error-prone process which will is challenging in clinical practice. The drawing of regions of interest (ROI) in myocardial segments requires adjusting for cardiac motion, which result in including blood pool (partial volume averaging) and artificially increasing the measured T1 (40). T1 times between patients may vary due to differences in Gd kinetics (such as in renal impairment), or with different contrast agents; correction factors have been proposed using kinetic modeling for the LL technique (41).
To address these shortcomings, several myocardial T1 mapping sequences have been created, including MOLLI.
Currently, the most evaluated sequence for myocardium T1 mapping is an MOLLI sequence (22, 42). The T1 mapping identifies a significant variation between normal and abnormal myocardium. It demonstrates that the myocardial fibrosis among different myocardial disorders includes ischemia (18), acute/chronic infraction (19), amyloidosis (20), diabetic (21), dilated and hypertrophic cardiomyopathy (17), and heart failure (9).
Modified Look–Locker inversion recovery is a CMR pulse sequence that is used for accurate T1 mapping of myocardium with high spatial resolution. A T1 map of the myocardium is a reconstructed image, where the T1 relaxation value is computed for every pixel of the corresponding myocardial voxel. Signal recovery from each myocardial voxel is acquired at different TIs following a single inversion pulse, all gated to the same cardiac phase, thereby enabling a pixel-based T1 quantification in the myocardium. MOLLI has introduced two variations to the standard LL sequence; selective data acquisition at a given time of the cardiac cycle over successive heartbeats, and merging of image sets from multiple LL experiments with varying TIs into one data set (22, 40). While selective data acquisition effectively decreases the number of images acquired in each LL experiment to one per heartbeat, the use of multiple LL experiments with different TIs increases the number of samples of the relaxation curve to a value that is sufficiently high for accurate T1 estimation.
Modified Look–Locker inversion recovery is an ECG-gated pulse sequence scheme and uses three prepared LL experiments consecutively within one breath hold over 17 heartbeats to reconstruct 11 images with different TIs. Three successive ECG-triggered LL experiments (LL1, LL2, and LL3) are carried out with three, three, and five single-shot readouts, respectively, at end diastole of consecutive heartbeats to sample the recovery of longitudinal magnetization after the inversion pulse. MOLLI pulse sequence scheme is illustrated (Figure (Figure3).3). T1 maps can be generated any time before or after contrast agent (e.g., gadolinium) administration (40).
Reconstruction of T1 maps from MOLLI source images is performed offline using purpose written customized software, or inline using vendor-specific processing, with ROI (septal or endo–epicardial) able to be analyzed in a pixel-wise quantitative fashion (Figure (Figure44).
Image data are sorted by their effective TI, which are given by
where N=image number within the LL experiment and RR=heartbeat interval. Three-parameter non-linear curve fitting using a Levenberg–Marquardt algorithm is performed for corresponding pixel, which is given by
where y denotes SI and corresponds to the apparent, modified T1 in an LL experiment.
Correction for readout-induced attenuation of the relaxation curve is attempted by using the three-curve fitting parameters , A, and B for the calculation of T1. Then, T1 estimates as in conventional LL methods by
The MOLLI sequence has been fully described, optimized, tested/re-tested, in phantoms and in large cohorts of healthy volunteers (22, 40) as well as being applied in cardiomyopathies (9, 17, 18, 20, 43). In addition, the T1 mapping with MOLLI has been validated against histopathology for assessment of myocardial fibrosis. It demonstrated that the pre-contrast “native T1” has a linear correlation with percentage of myocardial fibrosis as measured histologically on invasive myocardial biopsy. T1 times post-contrast administration (10–15min) had an inverse linear relationship with collagen content in myocardial fibrosis subjects (9, 44, 45).
Nacif et al. (46) compared n=168 myocardial T1 maps using LL and MOLLI at 1.5T, showing that pre-contrast (native) T1 values had good agreement, but LL had wider limits of agreement, and post-contrast T1 maps also had good agreement but with LL giving higher values than MOLLI, hence they are not interchangeable (46).
T1 mapping can be generated for different segments of the myocardium (base, mid-cavity, and apex) within a single breath hold of about 15–20s. However, the apex T1 values with MOLLI are slightly higher than basal and mid-cavity. The increasing in T1 values may be caused by partial volume effect and some degree of overestimation effect in apical level of left ventricle. These effects occur as a result of the ventricular wall being somewhat tilted toward the apex and no longer being aligned perpendicular to the short axis images (47–49).
Furthermore, T1 mapping with MOLLI has a greater reproducibility, accuracy, and an excellent overall inter- and intraobserver agreement over a wide range of TIs in pre- and post-contrast agent administration compared to the LL technique (22, 42). However, the T1 mapping with MOLLI sequence is sensitive to extremes of heart rate (bradycardia or tachycardia) (22) leading to slightly underestimation of T1 values. This may be corrected though “heart rate correction” by changing in the timing of the readouts with respect to the inversion pulses at different heart rates. This variation introduces various degrees of disturbance of the T1 relaxation curve. The heart rate affects the T1 value if the T1 value is higher than 750ms or less than 200ms with MOLLI (22, 42).
Moreover, MOLLI is limited by long breath hold about 15–20s (17 heart beats to acquire the final T1 maps). This may be difficult for elderly and pulmonary compromised patients and generates respiratory and motion artifacts (50). However, modern inline processing provides registration tools to reduce motion artifacts before the computation of final T1 maps (motion corrected or “MoCo MOLLI”). This will minimize the sensitivity of T1 mapping to motion artifacts and heart rate. Also, various acquisition sequences with short breath hold, such as shortened modified Look–Locker inversion recovery, have been validated and recently applied for cardiomyopathies (51, 52). At 1.5T, the pre- and post-contrast (10min) T1 times of normal myocardium are 980±53 and 470±26ms, respectively (Figure (Figure5)5) (22). Pre-contrast T1 values of myocardial fibrosis (infarction scar) are significantly longer than those of normal myocardium (1,060±61 vs. 987±34ms) (43). The longer pre-contrast T1 values in myocardial fibrosis patients have been reported in different cardiomyopathies. Infarction, myocarditis, and interstitial diffuse fibrosis, all have high pre-contrast T1 values when compared with normal myocardial T1 times (17, 20, 43). However, longer T1 values may also be noticed in different pathologically important processes, such as edema (53). Previous phantom, animal, and human tissue based studies lend insight into the effects on T1 signal by showing that T1 increases with increased water content and amount of extracellular fibrillar macromolecules (14, 54). Messroghli et al. concluded that increased myocardial T1 mapping corresponds to the areas of the human myocardial infarction, which showed on LGE (43). However, the links between the exact molecular mechanism in healthy tissue and myocardial fibrosis and corresponding pre-contrast T1 are less well understood. On the hand, the decrease in post-contrast T1 values in diffuse myocardial fibrosis has been previously related to increased extracellular space. It has been well described in acute, chronic ischemic, diffuse myocardial fibrosis, and inflammatory myocardial injury. The post-contrast T1 times (10min) were significantly shorter in chronic infarct scar compared with normal myocardium at 0.15mmol/kg (390±20 vs. 483±23ms, respectively) (43).
Also, T1 mapping with high magnetic field (3T) has been reported in a few studies of interstitial myocardial fibrosis. It was similar to 1.5T, the pre-contrast T1 was longer, and post-contrast T1 was shorter in myocardial fibrosis patients compared to normal myocardium. Puntmann et al. (55) reported higher pre-contrast T1 values for hypertrophic and non-ischemic-dilated cardiomyopathies at 3T compared to controls (hypertrophic 1.254±43ms and non-ischemic-dilated cardiomyopathy 1.239±57ms vs. healthy 1.070±55ms). Also, the post-contrast T1 values (10min) at 3T were shorter in hypertrophic and dilated cardiomyopathies compared to healthy (hypertrophic: 307±47ms, dilated cardiomyopathies: 296±43ms vs. controls: 402±58ms) (55).
The myocardium is made of densely packed myocytes, contributing to approximately 90% of myocardial mass. Signal from pre-contrast imaging reflects the majority of signal from the myocytes themselves. Conversely, in post-contrast imaging, the majority of Gd-based contrast signal arises from the interstitium (as Gd is an extracellular contrast agent and does not, therefore, reside within myocytes unless they are damaged); thus, post-contrast T1 mainly reflects interstitial or replacement fibrosis (see Box 1, below), hence the ability to calculate extracellular volume (ECV) from the difference between these two parameters.
Pre-contrast “Native” T1=predominant signal from myocytes (replacement fibrosis, or intracellular accumulation, e.g., Fabry disease). Post-contrast T1=predominant signal from interstitial space (interstitial fibrosis).
T1 mapping data can be used to calculate ECV fraction, using the pre- and post-contrast T1 times and hematocrit. Details of this technique and its uses are beyond the scope of this review; SCMR guidelines exist to guide application of T1 mapping and ECV quantitation (65).
Challenges remain with myocardial relaxometry for T1 mapping. These include technical challenges such as variations of T1 times at different field strength and across different vendors, and the rapidity in growth of pulse sequences being released as product and as works-in-progress, calling into question both the inherent accuracy and the level agreement between these techniques. Furthermore, the variations in T1 relaxometry values with different contrast doses and image timing require further investigation, to establish the test–retest and inter-site reproducibility of this technique. Next, the challenges to application of T1 mapping to clinical practice include establishment of robust normal ranges in large cohorts across multiple ethnic groups, and the observation that T1 mapping appears to be a highly sensitive technique, with the ability to discriminate healthy normal myocardium and identify very early changes in substrate. However, this technique lacks specificity; a wide variety of conditions prolongs native T1 and/or shortens post-contrast myocardial T1. Therefore, further clinical data are required in order to establish the use of these parameters in relation to disease (e.g., early detection of target organ damage in systemic conditions such as hypertension or diabetes), to inform treatment decisions, and their ability to predict or alter clinical outcomes.
Myocardial T1 mapping using quantitative relaxometry is an emerging and important tool in the assessment of global myocardial fibrosis. It is a highly sensitive marker of disease but is not specific, with changes in myocardial T1 occurring in many different conditions. Nevertheless, the high sensitivity and excellent reproducibility of the technique offers a tool for the early detection of myocardial damage, over-and-above techniques such as the CMR LGE technique and other modalities such as speckle tracking echocardiography, pulse wave velocity, and tissue tagging. Native T1 mapping is proving to be a robust indicator of early myocardial disease in many conditions, and normal ranges and guidelines for post-processing have been published by the Society of Cardiovascular Magnetic Resonance (65). Myocardial T1 mapping is a rapidly evolving technique, now with longitudinal prognostic data emerging, and normal ranges established at 1.5 and 3.0T in healthy humans and in aging. Further questions remain as to the standardization of pulse sequences across field strengths and between vendors, the affect of contrast type, dose and timing, the post-processing software, and the interpretation of T1 mapping results to inform clinical practice.
CH-C, MS, and GG: main contributions to the conception or design of the work; the interpretation of data for the work; drafted the work and revised it critically for important intellectual content; final approval of the version to be published; and agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The authors thank Dr. Qurain Alshammari for his input.
CH-C was funded by the DSITIA Smart Futures Fellowship Early Careers Grant.