Education, tips and tricks to help you conduct better fMRI experiments.
Sure, you can try to fix it during data processing, but you're usually better off fixing the acquisition!

Friday, January 2, 2015

Potential of ultralow field T1 and high field T1ρ in evaluating brain trauma

Crazy Scientist sent me a link to a paper, "Neuroimaging after mild traumatic brain injury: Review and meta-analysis," (doi:10.1016/j.nicl.2013.12.009) and it prompted me to do something with a brief review I wrote this time last year as a way to plan some research activities on MRI of mild traumatic brain injury (mTBI). By a remarkable coincidence the review paper was made available online four whole days before I completed my own review. I've yet to read the published review so I can't yet tell if I wasted my time. In any event, the document I wrote was for internal consumption (for my collaborators, and to clarify my own thoughts) and was never designed to become a public document. But since our research direction changed mid-year I figured I might as well stick my review out there in case anyone can make use of it.

The title of my document is the same as the title of this post. You will find the contents pasted below, or if you prefer you can download a PDF from this Dropbox link. I have quickly re-read it to check for major bloopers, and I've added a couple of update notes highlighted in yellow. There may well be some direct copy-paste of parts of a few of the papers I reviewed, especially those with heavy neuroradiology content where I am generally a long way out of my depth and would prefer to accept charges of plagiarism than get the medical terminology wrong!

For the record, we are still interested in mTBI but the logistics of studying acute brain injury in a non-hospital setting, using a home-made machine (the ULFMRI) sitting in a second basement lab in a physics department, made it all too hard to pursue right now. We have shifted instead to studying chronic conditions where we have a fighting chance of getting a few people scanned in our unorthodox facilities.


 Document completed on 8th January 2014. A few updates highlighted in yellow, added on 2nd January 2015.

Potential of ultralow field T1 and high field T1ρ in evaluating brain trauma

Recent work at UC Berkeley has demonstrated that ultralow field (ULF) MRI of the human brain is readily achievable and provides a wide intrinsic contrast between brain tissue, CSF and blood (Inglis et al. 2013). The spin-lattice relaxation times (T1) observed at ULF are comparable to the relaxation times obtained at high fields using the so-called T1ρ contrast mechanism. (A brief review of the T1ρ mechanism appears below.) This similarity can be understood intuitively by recognizing that the static magnetic field strength B0 of the ULFMRI and the B1 spin locking fields used for T1ρ contrast are comparable, typically in the range 5-200 μT. Thus, we may reasonably expect that the applications mooted for T1ρ imaging at high fields will extend naturally to applications for ULF T1 imaging.

While T1ρ imaging at fields of 1.5 T and above has shown great promise in a wide array of novel applications, it is in the arena of human brain trauma and stroke that the largest potential seems to reside. There is a major hurdle to clear, however. At high magnetic fields the amount of tissue heating generated by the T1ρ contrast module – generally called a spin locking pulse – can be prohibitive and leads to severe restrictions to the peak amplitude of B1 that can be used. At 1.5 T the limit is around 80 μT, whereas at 3 T the limit is around 20 μT due to the quadratic dependence of heating (as measured by specific absorption rate, SAR) on the operating frequency. The SAR concern has drastically limited the application of T1ρ to human brain, even though a number of exciting results in animal models have shown that T1ρ contrast contains novel information on cellular processes that is not reflected in standard contrast mechanisms – in isolation or taken together - for hours or even days.

There are a number of “unmet needs” in the diagnosis and evaluation of such pathologies as mild traumatic brain injury (mTBI) and stroke where ULF T1 imaging holds promise. In the case of mTBI, for example, magnetic susceptibility weighted imaging (SWI) and diffusion-weighted imaging (DWI) have demonstrated their ability to characterize microscopic bleeding and white matter axonal damage, respectively. Yet subtle changes in brain tissue are often not detected on MRI scans in spite of cognitive deficits or psychiatric problems. And when lesions are detected, there is often limited correlation between acute measures derived from neuroimaging and the level of disability or eventual outcome. Predicting which patients are prone to cognitive or psychiatric disability is important since early therapeutic intervention has the potential to reduce the risk of long-term deficits.

Imaging of acute stroke is another area where MRI has yet to make a major clinical impact. Computed tomography (CT) is used to differentiate ischemic from hemorrhagic stroke, after which many patients receive an MRI scan only days after the event. MRI may be able to show the long-term pathology caused by the stroke. However, an earlier MRI scan that could determine the time since an ischemic stroke could be invaluable for determining the most appropriate treatment, e.g. thrombolysis, hypothermia or administration of neuroprotective drugs. Combinations of MRI parameters, in particular quantitative spin-spin relaxation time (T2) imaging together with DWI-derived measures, have had modest success of determining at a single imaging time point the duration of ischemia, yet there remains considerable variability in the results and little clinical application of acute MRI to date.

In this review the focus is on mild TBI. As will be shown, the reasons for this focus are four-fold: (1) there is a need to improve the detection of lesions in mTBI patients that have clinical symptoms but negative MRI scans; (2) there is a need to better differentiate lesions caused by mTBI from existing pathology or atrophy; (3) there is a need to develop biomarkers that can be used predictively in the acute and sub-acute stage of mTBI that might lead to improved treatment; and (4) the logistics of studying mTBI is amenable to the equipment and research subjects available at UC Berkeley. Other brain pathologies, most notably moderate to severe TBI, stroke and Alzheimer’s disease, are also of considerable interest to this proposal, but developing a research program tailored to mTBI should facilitate developments that can be redirected at these other conditions at the appropriate time.


AD             Alzheimer’s disease
ADC           Apparent diffusion coefficient
BOLD         Blood oxygenation level dependent (contrast)
CT              (X-ray) Computed tomography
DAI            Diffuse axonal injury
FA             Fractional anisotropy (
FLAIR        Fluid-attenuated inversion recovery
FLASH       Fast, low-angle shot imaging: a low flip angle, short TR gradient echo method
FSE            Fast spin echo: multiple spin echoes with separate phase encoding per echo
GCS           Glasgow coma scale
GRE           Gradient-recalled echo: a standard (spin warp) phase-encoded sequence
MCAO        Middle cerebral artery occlusion (stroke model)
mTBI          Mild TBI, synonymous with concussion.
PTSD         Post traumatic stress disorder
SAR           Specific absorption rate
                  (See for limits.)
SWI            Susceptibility-weighted imaging
TBI             Traumatic brain injury
TBSS         Tract-based spatial statistics

Traumatic brain injury in civilian and military populations

In the US mild TBI accounts for 75% of the 1.7 million patients annually who seek medical attention for acute head injury. However, to date, there are no reliable markers in neuroimaging for mTBI. Instead, mTBI is diagnosed based on a loss of consciousness (LOC) for less than 30 minutes, post-traumatic amnesia under 24 hours duration and/or a transient focal seizure, and a Glasgow Coma Scale (GCS) score between 13 and 15. Based on this diagnosis, limited intervention may be prescribed but in general patients are discharged without follow-up care.

In the last thirteen years more than 280,000 US service members have been diagnosed with TBI ( The rate of new cases now exceeds 30,000 a year. Most military or combat-related TBI occurs as a closed head injury as a result of an explosion (e.g. blast pressure wave) or from vehicle accidents, falls and athletic activities. Injuries during training are common, suggesting that the rate of mTBI will remain high in military populations even as deployments to Iraq and Afghanistan are reduced.

The predominant pathological signs of TBI include diffuse axonal injury (DAI) and micro-hemorrhage, with patterns varying depending on the cause and severity of injury. Evaluating someone who has experienced head trauma to determine the extent of the injury is critical to limiting further brain damage. As with civilian mTBI, many cases probably go undiagnosed, while many other cases that are diagnosed based on clinical instruments such as the Glasgow Coma Scale have negative radiological scans. The vast majority (>80%) of military TBI cases are classified as mild TBI (mTBI) and, in the absence of other trauma, the soldier may be considered for return to active duty eventually. Rapid assessment is needed to determine if and when it is safe for the soldier to return to his or her unit. In the military theater, screening is complicated because TBI occurs in the context of sleep deprivation, nutritional changes, emotional stress, polytrauma, and difficult environmental factors. Mission pressure also plays a role in returning military personnel to duty when they may have residual pathology.

Military brain injuries may be co-morbid with post-traumatic stress disorder (PTSD), and there is growing evidence that TBI and PTSD resulting from military exposures increases the risk of developing neurodegenerative diseases such as Alzheimer’s disease (AD) (Weiner et al. 2013), in addition to increased risk of depression, anxiety and substance abuse. Collectively, the prevalence and long-term consequences of mTBI in the military indicates the need to come up with better early detection procedures that rely on objective measures, and to be able to use them to develop improved interventions.

MRI of mild TBI

Emergency clinical neuroimaging is mostly conducted with CT because the procedure is sensitive to critical pathology such as intracranial bleeding, it readily detects skull fracture and it can be performed on patients with life support equipment or metallic fragments in the head or body. However, CT tends to be insensitive to many changes following mTBI and there is need for subsequent neuroimaging tests that could determine a baseline for evaluation and treatment. Between a quarter to a third of patients with negative CT findings are found to have pathology on subsequent MRI scans (Bigler, 2013).

Hunter et al. (2012) recently published a review of emerging neuroimaging tools for TBI research. The predominant neuropathological signs of TBI include diffuse axonal injury (DAI) and microhemorrhage. In mTBI, as determined by the Glasgow Coma Scale or other clinical assessment, these pathologies may or may not be measurable on MRI scans. The heterogeneity and complexity of TBI are recognized factors in using neuroimaging for evaluation of injury severity. These limitations can be problematic because good evaluation of brain trauma can be critical to limiting further damage, either by prohibiting certain activities, including a return to active duty for military personnel, or by prescribing an appropriate course of therapy. Furthermore, many recent studies have shown that, even when focal traumatic lesions are detected on an acute scan, measures derived from the lesions often do not correlate with clinical outcome. The implication is that while MRI can detect numerous small lesions, including axonal injury and microscopic hemorrhages, these may be clinically irrelevant. Better specificity and sensitivity is required if MRI is to make a better impact in the evaluation and treatment of mTBI cases.

Fluid-attenuated inversion recovery (FLAIR) imaging:

Fluid-attenuated inversion recovery (FLAIR) imaging – more correctly, T2-weighted FLAIR - is useful for identifying edematous lesions. Its suppression of cerebrospinal fluid (CSF) signal allows better evaluation of lesions adjacent to CSF than standard T2-weighted imaging. The T2 contrast mechanism relies on its sensitivity to changes in molecular mobility; most notably, cellular edema produces hyper-intense regions on a T2-FLAIR image.

The history of FLAIR for imaging TBI is mixed. While it has shown value in separating moderate and severe cases from mild TBI (Chastain et al. 2009), there are many clinical mTBI cases that produce negative FLAIR scans. Furthermore, while broad injury categorization is feasible, FLAIR is unable to produce a good predictor of eventual outcome on an individual patient basis. These limitations have provoked the testing of other sequences in the hope of securing better sensitivity in mTBI cases and better specificity and outcome prediction across the spectrum of TBI severity.

Susceptibility-weighted imaging (SWI):

SWI exploits the magnetic susceptibility differences of blood, iron and calcium deposits from normal gray and white matter tissue. Local variations in T2* are produced by hemorrhages, abnormal blood oxygenation, and by blood breakdown products. SWI uses a high-resolution, three-dimensional velocity compensated gradient echo sequence from which a final SW image is derived that uses both the magnitude and phase information. A filtered version of the phase image is multiplied with the magnitude image to augment contrast in the final SW image. In essence, then, SWI is an enhanced T2*-weighted imaging method. (Some reports refer to this method as BOLD imaging due to their similar T2* basis, but this is an inaccurate use of the BOLD label because concentrations of iron generate the dominant features.)

SWI has primarily been considered to be helpful in the identification of hemorrhagic axonal injury where smaller hemorrhages are invisible on CT or conventional MRI sequences. SWI may be especially helpful in the detection of microhemorrhages in early acute and sub-acute phases of injury, and in detecting areas of hypoxia-ischemia-induced secondary injury. For example, SWI at 1.5 T was able to detect a larger number of lesions and to define smaller areas of damage than CT, T2-weighted imaging, and FLAIR imaging when applied to a diverse TBI patient group an average of 5.6 days post-injury (Chastain et al., 2009). However, SWI at 1.5 T has not yet been shown to be superior to T2-weighted imaging or FLAIR in discriminating between good and poor outcomes.

SWI is also being used to locate chronic lesions. Spitz et al. (2013) found that SWI at 3 T was able to detect more lesions and be a better indicator than FLAIR for outcome, but these scans were obtained an average of 18 months after injury. SWI was able to identify TBI-related lesions in almost one third of patients for whom FLAIR was unable to detect any lesions; the lesion volume determined by SWI was also related to injury severity. However, earlier work by Geurts et al. (2012) using MRI an average of 7 weeks after injury, suggested that SWI and FLAIR are equally useful for identifying lesions. Clearly, the time since injury is a critical factor in establishing contrast and, therefore, the most appropriate pulse sequence to use.

A potential limitation of SWI is that the contrast mechanism tends to make the size of a susceptibility-induced feature appear larger than the actual lesion size. Nevertheless, lesion size and location may be poor markers for deficits or eventual outcome to the point that the limitation is irrelevant. Other challenges in using SWI for TBI include its susceptibility to artifacts at air–tissue and skull-tissue interfaces, including the orbital frontal areas, a region of distinct vulnerability in TBI. Finally, interpretation of SW images requires adequate training and familiarity with the technique in order to differentiate (normal) veins from small hemorrhages.

Diffusion tensor imaging (DTI):

DTI contrast is based on restrictions and hindrances to the free diffusion of water in tissue. One use of DTI is to compute scalar values that represent the bulk and anisotropic diffusion rates, such as the apparent diffusion coefficient (ADC) and fractional anisotropy (FA), respectively. ADC is a directionless estimate of the average bulk water movement in a voxel whereas FA estimates the preference for a directionality of water diffusion, e.g. due to cellular membranes and other barriers to free diffusion. ADC tends to increase with vasogenic edema, when water flows out of capillaries and into the interstitial space, but it decreases with cytotoxic edema, when a greater fraction of water molecules is retained in ischemic, swollen cells. FA, being a function of anisotropic restrictions to diffusion such as occurs along myelinated fibers, is especially sensitive to white matter (WM) tract damage.

According to a review by Benson et al. (2012), there are several mechanisms that will result in FA decrease in WM after injury, including (vasogenic) edema, ischemia, neurodegeneration and metabolic disruption, while few if any mechanisms result in FA increases. Understanding these characteristics permits ADC, FA and other scalar parameters derived from the DTI to be used together to characterize pathology. For example, decreased FA associated with increased ADC suggests vasogenic edema, whereas increased FA associated with decreased ADC suggests cytotoxic edema. Decreased FA associated with decreased longitudinal water diffusivity indicates impaired axonal transport. Decreased FA, together with a slight or no increase in ADC together with increased radial diffusivity (RD) is typical of sub-acute or more chronic axonal injury from trauma.

MacDonald et al. (2011) studied a group of soldiers who had clinical evidence of mTBI, including exposure to a blast, without abnormalities on conventional MRI, at a median of 14 days post-injury. A significant number of injured patients showed decreased FA and increased ADC compared to controls. Of particular note, the regions where white matter abnormalities were seen (middle cerebellar peduncles, orbitofrontal cortex) differed from the typical regions (cingulum, uncinate fasciculus, anterior limb of the internal capsule) observed in non-blast mTBI.

Importantly, changes in DTI-derived measures have shown correlation with injury severity, functional outcome, neurologic functioning, and cognitive ability. Longitudinal studies have also indicated that DTI might serve as a tool for revealing changes in the neural tissue during recovery from TBI. But while DTI derived measures are able to sort mild from more severe TBI, to date there has been no way to determine whether an individual patient has a mild TBI. Reasons for this lack of specificity include the diffuse nature of TBI-related anatomical changes, very small abnormalities caused by the TBI, and confounding features related to aging or other disease processes. As a consequence, although there have been several group studies, routine use of MRI-derived measures to diagnose mTBI have yet to be realized.

When DTI abnormalities have been detected in acute cases of mTBI the results have been ambiguous, with both increases and decreases in fractional anisotropy (FA) being observed within the first 24 hours after injury (Kou et al., 2013). This lack of specificity restricts the utility of DTI as a marker for outcome and has led to the suggestion that multiple factors, such as neuroimaging and blood protein levels, may provide more useful information.

Diffuse axonal injury may be detectable by DTI in moderate and severe TBI, but a recent study showed no significant differences for FA in mTBI patients compared to controls (Spitz et al., 2013). Moderate and severe TBI showed significant widespread reductions in FA. An issue raised by this study is how regions of interest are determined for quantitative comparisons. Anatomical ROIs as well as tract-based spatial statistics (TBSS) were tested. Use of TBSS was considered to offer more flexibility in evaluation than a priori ROIs. But TBSS was still unable to identify abnormalities in mTBI. Other methods for localizing pathology are required.

T1ρ relaxation: an overview

A starting point for understanding the origin of T1ρ relaxation is given in Wikipedia (

“(Water) molecules have a kinetic energy that is a function of the temperature and is expressed as translational and rotational motions, and by collisions between molecules. The moving dipoles disturb the magnetic field but are often extremely rapid so that the average effect over a long time-scale may be zero. However, depending on the time-scale, the interactions between the dipoles do not always average away. At the slowest extreme the interaction time is effectively infinite and occurs where there are large, stationary field disturbances (e.g., a metallic implant). In this case the loss of coherence is described as a "static dephasing". T2* is a measure of the loss of coherence in an ensemble of spins that includes all interactions (including static dephasing). T2 is a measure of the loss of coherence that excludes static dephasing, using an RF pulse to reverse the slowest types of dipolar interaction (in a spin echo). There is in fact a continuum of interaction time-scales in a given biological sample, and the properties of the refocusing RF pulse can be tuned to refocus more than just static dephasing. In general, the rate of decay of an ensemble of spins is a function of the interaction times and also the power of the RF pulse. This type of decay, occurring under the influence of RF, is known as T1ρ. It is similar to T2 decay but with some slower dipolar interactions refocused, as well as static interactions, hence T1ρ ≥ T2.”

Understanding the T1ρ phenomenon can thus be extrapolated from a consideration of spin echoes. We start with a simple spin echo sequence, 90x – TE/2 – 180y – TE/2. This single spin echo refocuses dephasing that arises from any process that is time-invariant across the two halves of the echo time, TE. The refocusing time scale can be shortened, however, by including additional refocusing pulses while measuring the signal at the same overall TE. This is the classic Carr-Purcell-Meiboom-Gill (CPMG) echo train. The apparent T2 that is measured will increase as the number of refocusing pulses is increased, for a fixed total TE, because the diffusion sensitization of the signal is reduced in concert. This well-known result has been utilized historically to study the process of molecular diffusion.

As more and more refocusing pulses are included in a CPMG echo train (of constant length, TE), the inter-pulse evolution delay becomes very small and the measured apparent T2 approaches the true T2. If we now simply discard the small inter-pulse delays we have replaced a train of refocusing pulses and inter-pulse delays with a constant “refocusing” field that persists throughout TE. Because there are no inter-pulse delays, and therefore no phase evolution periods absent an RF field, it is convenient to change nomenclature and use TL (to indicate a spin “lock”) for the duration of the long B1 pulse, rather than TE.

We now consider how the transverse magnetization decays under an effective constant field attained by the long duration B1 pulse. Viewed in a rotating reference frame at the Larmor frequency w0 = gB0, the effective field reduces to Beff(t) = B1(TL). The direction of B1(TL) is parallel with the magnetization direction established by the original 90x-degree excitation pulse; in this example Beff is along y’ of the rotating reference frame. Thus, the subsequent decay of coherent magnetization occurs along the direction of B1(TL). For this reason the loss of coherence under the effective field B1(TL) is characterized by a time constant T1ρ in order to disambiguate it from T1 and T2 relaxation, processes which occur in the absence of an RF field. It is also important to note that the T1ρ decay involves loss of transverse magnetization orthogonal to the B0 axis but parallel to the RF field axis. For this reason the T1ρ relaxation time constant is often referred to as “T1 in the rotating frame.” The longitudinal designation is in reference to relaxation parallel with B1(TL), not parallel with B0 as for T1. (For completeness it is worth nothing that there is also a T2ρ decay, which occurs during rotary echoes formed by periodically reversing the direction of the RF field and measuring the decay of magnetization that is transverse to the RF spin lock field. T2ρ is not considered further in this proposal because of the sparse literature using T2ρ in human brain or in animal models of brain trauma.)

The particular molecular interactions that dominate T1ρ relaxation are determined by the frequency of the RF pulse used to interrogate the spins. The interaction frequency of primary interest equates to 42.58 Hz/μT. Given an exemplar spin lock field magnitude of 20 μT, the frequency expected to dominate relaxation is ~850 Hz. At body temperature, much proton exchange kinetics, especially water-protein interactions, occurs in a low frequency range from a few hundred Hz to a few kHz. The exchange kinetics – and T1ρ - thus depends strongly upon cellular energetics, protein concentration, pH, temperature, viscosity and processes that mediate water concentration and mobility within cells (Makela et al. 2001). The spin lock field strength-dependence is one of the major features of T1ρ imaging. It is possible to tune the T1ρ relaxation time contrast to match expected proton exchange frequencies.

T1ρ imaging:

The standard T1ρ contrast module consists of a 90-degree excitation pulse followed by a long duration, low amplitude “spin locking” B1 field placed on-resonance. Once the spin locking relaxation period has ended the remaining magnetization may be returned briefly to the z-axis with a second 90-degree pulse. The net result of a 90x – B1SL(TL) – 90-x preparation sequence is thus a longitudinal magnetization that is dependent upon T1ρ. Regular imaging sequences may then be applied to sample the T1ρ-dependent signals. The imaging scheme can be any imaging pulse sequence, although the long duration of the spin locking composite suggests that 3D sequences will be preferred. (In this regard, spin locking is similar to T1-weighted imaging using inversion recovery.) Pulse sequence issues pertinent to T1ρ imaging are considered in more detail below.

The simplest approach to T1ρ imaging is to establish a weighted contrast, as with conventional T1 and T2-weighted imaging. More often, however, T1ρ is quantified for a particular magnitude of B1SL using several different TL periods; four or five TL periods are commonly used in generating a T1ρ map. Finally, it is also possible to measure a range of T1ρ for different amplitudes of B1SL, a process termed T1ρ dispersion imaging that is analogous to (static) field cycling relaxometry.

T1ρ imaging of TBI and stroke

Whereas FLAIR imaging reflects cellular water mobility and water concentration via T2, and diffusion imaging attempts to measure bulk water motion directly, T1ρ imaging is concerned with the interaction between water protons and other molecules and is expected to contain unique information on tissue status. This sensitivity to proton exchange suggests that it may be a useful modality to measure cellular dysfunction, e.g. pH changes. T1ρ contrast has been shown to be sensitive to some molecular processes earlier or more specifically than other MRI contrast methods, especially in pathologies involving disruption of cellular energetics.

T1ρ contrast MRI should be extremely useful in the acute phase of brain injury, when cascades of biochemical changes are occurring extremely rapidly. NMR spectroscopy, most notably 31-phosphorus NMR, has demonstrated its utility in studying rapid energetics – even during exercise – and so the potential for conducting metabolic imaging by proxy, based on the vastly more sensitive 1-H nucleus, is an intriguing possibility.

In an experimental model of TBI (fluid-percussion induced-TBI in rat brain), Immonen et al. (2009) found that both T1ρ (using an 80 μT spin lock field) and T2 measured at 4.7 T three days after injury correlated with final hippocampal atrophy determined post-mortem 6 months later. Due to the similar temporal characteristics of the two contrast mechanisms, Immonen et al. concluded that the relaxation time changes reflected the severity of vasogenic edema, a phenomenon seen in prior stroke studies. At 23 days post-injury, T1ρ, T2 and ADC increases predicted the long-term neurodegeneration of the ipsilateral hippocampus. However, in a subsequent report (Immonen et al. 2009b), the authors noted that T1ρ measured at the injury site exhibited significant differences from control animals as few as three hours after the injury, whereas T2 changes were insignificant at that time. These acute phase results are in general agreement with other studies comparing T1ρ to T2 and T1.

Perhaps the most compelling demonstration of T1ρ utility has been shown in a rat model of stroke using middle cerebral artery occlusion, MCAO (Jokivarsi, et al. 2010). In that work, T1ρ was shown to provide a linear index of post-ischemia delay by comparing T1ρ in infarcted tissue to tissue on the contralateral hemisphere. The T1ρ difference correlated with time since MCAO. The T2 difference between the ipsilateral and contralateral tissue was initially negative, however, becoming positive 1.5-3 hours post-MCAO. A non-linear response is also generally reported for diffusion imaging studies of stroke. Thus, while DWI may be a sensitive way to localize tissue damage, the quantities that are computed from DWI do not have the specificity to permit accurate quantitation of ischemic duration.

Very little work has been done on TBI using T1ρ in either animal models or clinically. A recent paper (Immonen et al. 2013) used a fluid percussion injury model in rat and demonstrated that T1ρ is highly sensitive for predicting increased seizure susceptibility and epileptogenesis after TBI. At 9 and 23 days post injury, a change in T1ρ of the perilesional cortex showed the greatest predictive value for increased seizure susceptibility at 12 months. But further work using T1ρ for trauma pathology has been hampered by the potential for heating arising from the long duration spin lock pulses.

Pulse sequences for T1ρ imaging in human brain

There are two parts to achieving successful T1ρ imaging of human brain: an acceptable contrast module and a rapid image acquisition scheme. With regard to the T1ρ contrast module, it should be robust to B0 and B1 heterogeneities expected across the head, and it must be applied within safety limits for tissue heating. Heating from B1 is measured using a specific absorption rate (SAR) in units of W/kg. Even moderate B1 amplitudes applied for tens of milliseconds can approach or exceed SAR safety limits and the problem is compounded by the use of higher static magnetic fields because SAR scales quadratically with frequency.

Until very recently, most attempts to do T1ρ imaging in human brain used a standard spin locking pulse as described previously. (Some publications refer to this approach as “continuous wave” spin locking because the locking B1 field is applied at constant amplitude.) The first studies on human brain generally focused their efforts on the imaging readout in an attempt to make the total acquisition time practicable. A rapid imaging method is required to sample the T1ρ-prepared spins, both to permit acquisition of data in a clinically acceptable time but also to permit extension of repetition times in order to offset some of the SAR limitations arising from the contrast module.

An initial study of T1ρ in human brain was performed at 0.1 T (Ramadan et al., 1998). SAR issues were largely avoided by the use of a low static field, permitting measurement of T1ρ in 50 μT steps up to 250 μT. Image acquisition was performed with a single slice GRE acquisition. In 2004, Borthakur et al. acquired T1ρ of human brain at 1.5 T using a conventional spin lock module followed by a fast spin echo (FSE) readout using centric phase encoding. Two-dimensional readout was used in that report although the authors noted that a 3D gradient echo readout scheme could have been used, as they had applied previously in skeletal imaging. SAR considerations limited the spin lock amplitude to a maximum of 11.7 μT. T1ρ dispersion data were reported for 1.5-11.7 μT.

In 2004, Wheaton et al. again used a single slice GRE scan at 1.5 T to measure T1ρ in human brain, but in order to reduce SAR the amplitude of the spin lock field was reduced to 4.7 μT for the outer thirds of phase-encoded k-space while the full amplitude of 11.7 μT was used only for the central third of (low frequency) k-space. The T1ρ values produced by this sequence were within 2% of those using 11.7 μT for the entire k-space. The same group demonstrated faster T1ρ imaging in 2006 by the use of EPI readout at 1.5 T (Borthakur et al., 2006), permitting the use of a constant amplitude spin lock of 11.7 μT.

Very recently, 3D turbo spin echo (TSE) readout with CSF suppression was demonstrated at 3 T in a study of age-related T1ρ changes in human brain (Watts et al., 2013). Whole brain coverage was obtained with 1.8 mm isotropic resolution in a total acquisition time of 14 min. SAR was rendered acceptable – 69% of the safety limit of 3 W/kg average for the head - using a spin lock amplitude of 11.7 μT.

The initial attempts to modify the standard spin lock contrast module tended to focus on artifact reduction rather than the SAR issue. Witschey et al. (2007) used phase alternation, a variant of rotary echoes, within the spin lock to overcome B0 and B1 imperfections. Michaeli et al. (2006) introduced a fundamentally different approach to the spin lock. Instead of constant amplitude pules, spin locking is effected via a train of adiabatic full passage (AFP) sweeps; the dynamics of the AFP sweeps maintains the magnetization precessing around the effective field in the desired manner. The original AFP work was performed at 4 T and used single slice readout, either segmented 2D Turbo FLASH or segmented 2D spiral. Even so, the peak B1 amplitude in the AFP sweeps was 59 μT, a high value made feasible with the use of a small surface transmit coil approximately 14 cm in diameter.

The concept of a variable spin lock field during the preparation period has been extended very recently. A gradient-modulated variant of the AFP train was presented by Andronesi et al. (2013) and used to reduce SAR at 3 T for whole brain imaging with 3D Turbo FLASH readout. The isotropic resolution was 1.3 mm and the whole acquisition of T1ρ at 9 μT took 5 min 40 sec, generating a SAR of approximately 1.8 W/kg. They used the standard body transmit coil and a 32-channel head coil for reception.

It now seems feasible to acquire whole brain T1ρ images in clinically acceptable times, using the combination of new spin lock contrast modules coupled with efficient 3D imaging readout. In the most recent applications of 3D T1ρ imaging (Watts et al., 2013 and Andronesi et al., 2013), extensive use was made of methods that reduce the total data acquisition needed, e.g. parallel imaging (SENSE, GRAPPA) and partial Fourier acquisition, and also methods that reduce SAR, e.g. variable flip angle TSE. It should be noted that a body transmit RF coil was used in both of these studies, limiting the peak spin lock field to between 9 and 12 μT for acceptable SAR. (The Siemens Trio body coil can attain maximum field strength of B1 = 23.5 μT.) Whether this low range of spin lock field amplitudes is useful for imaging acute neuropathology has yet to be demonstrated.

ULF T1 imaging of human brain

In their study on a rat model of irreversible stroke, Jokivarsi et al. (2010b) offered this observation:

“To supersede T2 in the detection of irreversible ischemia, T1ρ MRI must be measured with B1 in the range of 60 to 170 μT, which does not cause tissue heating with the data acquisitions used here, even in flow-compromised tissue (Grohn et al, 2000), but yet needs to be proven in humans within the SAR guidelines.”

In that study they had used a spin lock of 38 μT in order to emulate what they estimated might be feasible clinically, but as a result the dynamics of T1ρ had generated only marginal sensitivity advantage over T2. In their earlier studies they had used higher spin lock fields and had demonstrated very significant advantages of T1ρ contrast. That low spin lock amplitudes should be less beneficial is understandable because as the spin lock amplitude approaches zero, T1ρ ~ T2.

At present, the range of spin lock fields suggested by Jokivarsi et al. is beyond the technical and safety limits at 3 T and probably at 1.5 T as well, suggesting that exploring alternative approaches to obtaining T1ρ-like contrast is warranted for human brain. Fortunately, there is an alternative to high field T1ρ in assessing proton exchange-dominated contrast: ultralow field (ULF) B0 imaging. Spin-lattice relaxation in B0 is a function of the thermal fluctuations occurring at the Larmor frequency. Thus, T1 at B0 of tens to hundreds of microtesla will have similar frequency dependence as T1ρ relaxation under the B1SL used to date in spin locking methods on high B0 scanners. (Here, high B0 means any static magnetic field greater than 0.1 T.) (Update on 2nd January 2015: In recent work we have actually determined that ULF T1 and T1ρ have different dependencies, with T1ρ showing strong sensitivity to chemical exchange and perhaps water diffusion through microscopic magnetic susceptibility gradients, mechanisms that seem to be non-existent for ULF T1. This work is being prepared for submission.) Indeed, it is worth mentioning that the first paper on T1ρ imaging (Sepponen et al., 1985) set out to develop a method specifically to permit ultralow field relaxation time contrast but with the benefits of higher SNR attainable with a larger polarizing field. (We can think of high field T1ρ imaging as an attempt to mimic ULF MRI contrast on a high field scanner.) It seems as if the benefits of high field polarization may now have reached the point where T1ρ cannot be applied optimally, however. Fortuitously, with the development of pre-polarization and a SQUID detector at ULF, we are finally in a position to achieve high SNR as well as low field contrast directly.

Detailed comparisons of the relaxation mechanisms underlying T1ρ at high field and T1 at ULF have yet to be performed. While we might expect some small differences, e.g. due to chemical shift and magnetic susceptibility effects for T1ρ at high B0 (Makela et al., 2004), the T1ρ and ULF T1 are expected to be similar in magnitude and have similar dependence on molecular dynamics such as proton exchange. (See above highlighted comment.)

Qualitatively similar ULF T1 and high field T1ρ values have been reported for human brain. Inglis et al. (2013) found T1 at B0 = 130 μT of 61 ms and 141 ms for WM and GM, respectively. Ramadan et al. (1998) found T1ρ at B1SL = 150 μT (measured at B0 = 0.1 T) of 120 ms and 137 ms for WM and GM, respectively, but also reported anatomical variation of T1ρ for WM that was not observed for GM. Borthakur et al. (2004) reported T1ρ at B1SL = 11.7 μT (measured at B0 = 1.5 T) of 85 ms and 99 ms for WM and GM, respectively. Finally, Andronesi et al. (2013) determined T1ρ at B1SL = 9 μT (measured at B0 = 3 T) of 79 ms and 96 ms for WM and GM, respectively. Clearly, variation due to different subjects as well as experimental methods suggests that in-depth comparison of these results is unwise. But the results are reasonably consistent and suggest that ULF T1 imaging is a good alternative to high field T1ρ imaging of brain.

In pilot experiments on excised, fixed rat brains we have compared directly T1ρ measured with B1SL = 35 μT at 9.4 T to ULF T1 at B0 = 130 μT. Three of the rats had MCAO surgery (one hemisphere) 8 or 24 hours before sacrifice; the other two rats had sham surgery. (We are still blind to the assignment.) The fixed rat brains were sectioned to enable the hemispheres to be measured separately on the ULF system. Cerebellum and cervical spinal cord was excluded. We found good reproducibility of T1 at ULF (two sets of measurements) albeit with a systematic reduction of ULF T1 compared to high field T1ρ:

In spite of a low number of samples it is evident that there is a linear relationship between the two sets of relaxation times. Systematic differences could arise from methodology, e.g. the very high field used in measuring T1ρ. Chemical shift effects as well as magnetic susceptibility gradients across a small, irregular sample could have profound effects on T1ρ measured at 9.4 T. Another possibility is the presence of small amounts of free fixative. At ULF we were careful to remove as much free fixative as possible after noting that any visible fluid tended to bias the T1 towards artificially high values (closer to that for free water in fixative). It is conceivable that slight differences in sample preparation for the T1ρ measurements at 9.4 T, which were performed in Finland, could be the largest source of discrepancy. This work is ongoing. (Update on 2nd January 2015: We haven’t pursued this track recently. We can’t do animal model work ourselves and the ULFMRI scanner resolution is limited to ca. 2 mm voxels, making images of small rat brains rather poor. So instead we have refocused our efforts on human brain and we have no current plans to scan more fixed rat brains.)

Potential of ULF T1 and HF T1ρ for brain trauma imaging

There is now compelling literature to suggest that low field relaxation times are sensitive to the rapid changes occurring in the acute phase of traumatic injury. If we assume that low field relaxation, whether T1ρ or ULF T1, does indeed offer a novel window on traumatic injuries such as stroke and TBI, the question arises as to how to make the technology available to the patient.

A recent study of mTBI by Yuh et al. (2013) included this perspective:

“Routine performance of brain MRI on mTBI patients may not currently be cost-effective. However, smaller, less costly head-only MRI scanners are under development. These, among other continuing advances in MRI technology, may ultimately render the expense and logistics of acute MRI scans less prohibitive.”

It is worth emphasizing that a scanner optimized for imaging of acute brain trauma (including stroke) may need to be different than one optimized for methods (such as SWI and DTI) currently used for chronic studies. Higher B0 leads to greater contrast in SWI, while for DTI higher B0 improves SNR. In the case of T1ρ imaging, however, the SNR benefit that arises as B0 is increased is offset by the considerable SAR, leading to severe restrictions on the spin lock field amplitudes that can be utilized. It could be preferable to use T1ρ contrast at a relatively low field of 0.5-1 T.

There may also be utilitarian and safety reasons to develop a dedicated MRI scanner for acute trauma assessment. Chronically, the benefits of high B0 can be utilized with ample time to consider contraindications to the high magnetic field, any incidental pathology, atrophy and so on. In the acute phase, however, and especially when the injury is suffered on the battlefield, the initial consideration shifts from optimal contrast to optimal safety and ease of access. Whereas it is typical for an emergency room to have immediate access to a dedicated CT scanner, few centers have short-notice access to MRI for patients that may be on life support or otherwise may present contraindications for any procedure presently considered to be non-essential.

That said, given the widespread availability of clinical MRI scanners there is clear merit in attempting to evaluate T1ρ contrast for studying TBI with existing technology. A parallel approach is to develop ULFMRI for studying acute brain trauma. Compared to T1ρ at high fields, using ULFMRI could offer several possible benefits for acute trauma imaging. It should permit easier use of life support apparatus, and should be better suited to patients at risk of contraindications for high field MRI. ULFMRI could also offer significant performance benefits over high field T1ρ imaging, such as allowing imaging in the presence of metal in the patient’s head, e.g. surgical implants to stabilize facial fractures, shrapnel, etc. Finally, a ULFMRI scanner designed specifically for acute trauma injury would likely be smaller, cheaper and easier to site than a high field scanner using a superconducting magnet.

Deleted on 2nd January 2015: Sections on Future Work and a Timeline.


Andronesi et al. (2013).
Whole brain mapping of water pools and molecular dynamics with rotating frame MR relaxation using gradient modulated low-power adiabatic pulses.
NeuroImage, Epub (2013).

Benson et al. 2012
Detection of hemorrhagic and axonal pathology in mild traumatic brain injury using advanced MRI: Implications for neurorehabilitation.
NeuroRehabilitation 31, 261–279 (2012).

Bigler, 2013.
Neuroimaging biomarkers in mild traumatic brain injury (mTBI).
Neuropsychol Rev 23, 169-209 (2013).

Borthakur et al. (2004).
In vivo measurement of T1ρ dispersion in the human brain at 1.5 tesla.
J Magn Reson Imaging 19, 403-9 (2004).

Borthakur et al. (2006).
A pulse sequence for rapid in vivo spin-locked MRI.
J Mag Res Imaging 23, 591-6 (2006).

CA Chastain, UE Oyoyo, M Zipperman, E Joo, S Ashwal, LA Shutter, and KA Tong. Predicting Outcomes of Traumatic Brain Injury by Imaging Modality and Injury Distribution.
J Neurotrauma 26:1183–1196 (2009).

Geurts et al. 2012
The reliability of magnetic resonance imaging in traumatic brain injury lesion detection.
Brain Inj. 26, 1439–1450 (2012).

Hunter et al. 2012
Emerging Imaging Tools for Use with Traumatic Brain Injury Research.
J Neurotrauma 29, 654-71 (2012).

Immonen et al. (2009)
Quantitative MRI predicts long-term structural and functional outcome after experimental traumatic brain injury.
NeuroImage 45, 1-9 (2009).

Immonen et al. (2009b)
Distinct MRI pattern in lesional and perilesional area after traumatic brain injury in rat – 11 months follow up.
Exp Neurol 215, 29-40 (2009).

Immonen et al. (2013)
MRI biomarkers for post-traumatic epileptogenesis.
J Neurotrauma 30(14), 1305-9 (2013).

Inglis et al. (2013)
MRI of the human brain at 130 microtesla.
PNAS 110(48), 19194-201 (2013).

Jokivarsi et al. (2010)
Estimation of the onset Time of cerebral ischemia using T1rho and T2 MRI in rats.
Stroke 41, 2335-40 (2010).

Jokivarsi et al. (2010b)
Correlating tissue outcome with quantitative multiparametric MRI of acute cerebral ischemia in rats.
J CBF Metab 30, 415-27 (2010).

Kou et al., 2013
Combining biochemical and imaging markers to improve diagnosis and characterization of mild traumatic brain injury in the acute setting: results from a pilot study.
PLOS One 8(11), e80296 (2013).

MacDonald et al. (2011)
Detection of blast-related traumatic brain injury in U.S. military personnel.
N Engl J Med 364(22), 2091-100 (2011).

Makela et al. (2001)
Proton exchange as a relaxation mechanism for T1 in the rotating frame in native and immobilized protein solutions.
Biochem Biophys Res Commun 289(4):813-8 (2001).

Makela et al. (2004)
B0 dependence of the on-resonance longitudinal relaxation time in the rotating frame (T1rho) in protein phantoms and rat brain in vivo.
Magn Reson Med 51, 4-8 (2004).

Michaeli et al. (2006)
T1rho MRI contrast in the human brain: Modulation of the longitudinal rotating frame relaxation shutter-speed during an adiabatic RF pulse.
J Mag Res 181, 135-47 (2006).

Ramadan et al. (1998).
On- and off-resonance spin-lock MR imaging of normal human brain at 0.1T: Possibilities to modify image contrast.
Mag Res Imaging 16(10), 1191-9 (1998).

Sepponen et al. (1985).
A method for T1rho imaging.
J Comput Assist Tomogr 9(6) 1007-11 (1985).

Spitz et al. 2013
White matter integrity following traumatic brain injury: the association with severity of injury and cognitive functioning.
Brain Topogr. 26, 648-660 (2013).

Watts et al. (2013)
In vivo whole-brain T1rho mapping across adulthood: Normative values and age dependence.
J Mag Res Imaging, Epub (2013).

Weiner et al. 2013
Military risk factors for Alzheimer’s disease.
Alzheimer’s and Dementia 9, 445-51 (2013).

Witschey et al. (2007)
Artifacts in T1rho-weighted imaging: Compensation for B1 and B0 field imperfections.
J Mag Res 186, 75-85 (2007).

Wheaton et al. (2004)
Method for reduced SAR T1rho-weighted MRI.
Magn Reson Med 51, 1096-1102 (2004).

Yuh et al. 2013.
Magnetic resonance imaging improves 3-month outcome prediction in mild traumatic brain injury.
Ann Neurol 73, 224-35 (2013).

1 comment:

  1. Hello -- I'm a reporter with KTVU in Oakland. I'm hoping to reach you regarding one of your YouTube videos about MRI for a story we're preparing. If you could write me back, I'd be grateful. Thank you.