1. Will provide a comprehensive overview of current approaches to examining and characterizing brain microstructure and organization using both qualitative and quantitative imaging methods;

2. Strengths and limitations of imaging methods and associated tissue models will be examined;

3. Clinical utility and interpretability will be discussed.

The target audience of this presentation includes MR physicists as well as clinical researchers interested in the use of quantitative imaging data, including relaxometry, to interrogate tissue microstructure and composition.

While the majority of routine clinical and diagnostic imaging comprises qualitative T1, T2 and/or proton density (PD)-weighted imaging, additional information related to tissue microstructure can be gleaned through quantitative voxel-wise evaluation of the T1 and T2 relaxation times. Such imaging, commonly referred to as quantitative relaxometry, can minimize or eliminate hardware, acquisition, and patient-specific confounds that exist within conventional qualitative images, such as coil sensitivity differences, or differences in acquisition pulse sequences and imaging parameters. As such, quantitative imaging can afford improved reproducibility (both between different imaging centers, as well as with longitudinal measures on the same patient) and, in some cases, improved image contrast (Fig. 1).

Conventionally, calculation of T1 and/or T2 within each imaging voxel is performed by fitting a known signal intensity model to appropriately acquired data. While these approaches (and many others) provide precise and reproducible T1 and T2 values, they impose an inherent tissue model onto the data. That being, all water protons within the voxel have the same relaxation properties (i.e, they are all governed by the same T1 and/or T2 relaxation time). Since T1 and T2 are exquisitely sensitive to their biophysical and biochemical environment [1], this model, therefore, assumes each voxel contains a single homogeneous water pool (i.e., single component relaxation). Unfortunately, observation of tissue structure on the microscopic scale reveals a more complex picture, with water compartmentalized into multiple distinct environments, with differing physical and biochemical structures, and passive and active transportation processes that shuttle water between them [2] (Fig. 2).

While the complexity of tissue microstructure raises concerns about the validity of single component relaxometry, the sensitivity of relaxation processes to the local physical and biochemical environment offers the potential to glean additional microstructural information through the use of more complex relaxation (e.g., multicomponent) models.

Multicomponent T2 Relaxation

Proposed in the late 1980’s and further refined through the 1990’s, the first approach to tissue microstructure through relaxometry was through multicomponent analysis of T2 relaxation [3]. Expanding the Spin Echo signal expression to include multiple T2 relaxation species provides a framework for quantifying sub-voxel tissue compartments delineated by relaxation time. While this approach makes no assumption regarding the number of differing T2 species within the voxel, it does present a formidable analysis problem in that unconstrained it presents an underdetermined system. To avoid this problem, a T2 distribution consisting of M logarithmically-spaced T2 values are fit to the M echo time SE data using non-negative least squares and imposing a smoothing function to ensure a continuous distribution [4] (Fig. 3a).

Peaks within the T2 distribution are believed to correspond to unique physical environments [3.4]. In brain, for example, the short T2 peak (T2 < 50ms) corresponds to water trapped within the lipid bilayers of the myelin sheath; the intermediate peak (80ms < T2 < 200ms) to intra/extra-cellular water; and the longer peak (T2 > 250ms) to unrestricted free water. These assignments have been collaborated through histological comparison studies (Fig. 3b), as well as examination of their changes in known demyelinating diseases (such as multiple sclerosis, MS). The ratio of the area under the short T2 peak to the area under the full distribution has been termed the myelin water fraction (MWF), which has seen increased research activity of late due to its ability to provide improved sensitivity and specificity to myelin changes in MS and other degenerative disorders.

Multicomponent T1 Relaxation

Initial attempts to replicate results from T2 analysis in the context of T1 were, unfortunately, unsuccessful. This failure is believed to be due to the relative difference in the timescales between T1, T2, and the rate of water exchange between the various water compartments. Unfortunately, conventional multicomponent T2 analysis makes no mention of water exchange between, for example, the myelin-associated water and the intra/extra-cellular water pools, assuming instead that T2 is short relative to the exchange time, τ, such that each component can be considered in isolation. In contrast, T1 is long with respect to τ, such that the components appear as one “well-mixed” container. Thus, while T2 may be analyzed using multicomponent analysis, T1 appears to be mono-exponential.

However, in combined T1 and T2 analysis of peripheral nerve, Does et al., have shown unique T1 times associated with each T2 component [5]. Though difficult to observe and quantify directly, this underlying multicomponent T1 has led to the development of novel imaging methods designed to selectively isolate the T1-weighted signal associated with the myelin water [6]}.

Combined Multicomponent T1 and T2 Relaxation

While the conventional approaches to visualizing and quantifying multicomponent relaxation rely on spin echo or inversion recovery methods, any imaging sequence is sensitive and susceptible to these effects. Recently, a more rapid approach for quantifying multicomponent relaxation, termed mcDESPOT, has been proposed that utilizes rapid and time efficient steady-state imaging methods {Deoni:2008fe}. This approach differs from spin echo based measures in that water exchange and cross-relaxation effects are implicitly included within the signal model. This inclusion, however, requires knowledge of the tissue system (i.e., what water pools are in each with each other), which forces an upper limit on the number of tissue pools that can be modelled (3) (Fig. 4) [7].

While offering a potentially rapid approach to myelin water imaging, mcDESPOT remains to be fully understood. MWF values derived using mcDESPOT are universally higher than those obtained using spin-echo methods, though they are correlated with each other (Fig. 5a), and have been qualitatively validated against histology (Fig. 5b). Further, the effect of magnetization transfer remains to be fully addressed.

MRI relaxation data offers a potential wealth of information with increased specificity and sensitivity to tissue microstructure and composition. While numerous acquisition methods have been presented, and several mathematical tissue models and analysis approaches described, there remains significant limitations and voids in our knowledge regarding the ideal approach or the interpretation of derived results. Nevertheless, with more recent methodologies offering the potential for multicomponent relaxometry within clinically realistic scan times, the field is experiencing renewed interest and rapid growth.