The distinct contrasts observed in magnetic resonance (MR) imaging of biological tissues are strongly influenced by the composition of biomolecular constituents and their hierarchical organization [1], [2]. Multiple relaxation mechanisms—including (residual) dipolar interactions, chemical exchange, and (anisotropic) susceptibility effects—contribute to these image contrasts. For example, both susceptibility-based pathways and the magic angle effect (MAE) have recently been considered [3] for characterizing orientation-dependent transverse relaxation in brain white matter (WM). However, MAE has not been adequately addressed, largely due to an oversight of the detailed organization of bound water adjacent to or within biomacromolecules. Historically, the concept of “structured” water in tissues has been met with skepticism [4]; nevertheless, its recent application via MAE to orientation-dependent transverse relaxometry in neural tissues [5], [6] has highlighted the role of microstructural water organization, challenging previously prevalent susceptibility-based models [3], [7], [8].

Approximately two decades ago, the increasing availability of whole-body MR scanners operating at ultra-high magnetic B0 fields led to the discovery of unprecedented varying T2∗ contrast in gradient-echo (GRE) images of human brain WM in vivo at 7 T [9], [10]. This characteristic T2∗ contrast (or its reciprocal, R2∗ contrast) was subsequently found to depend on the orientation of axon fiber tracts in an animal study [11], where the anesthetized brain was positioned at different angles relative to the B0 field. Shortly thereafter, a clearer picture emerged from a study conducted on healthy human brains at 3 T [12], where the R2∗ values in fiber tracts were found to be higher along either the left-right or anterior-posterior directions compared to the inferior-superior direction, which aligns with the B0 field. Moreover, the observed R2∗ values appeared to be proportionally correlated with diffusion fractional anisotropy (FA) derived from diffusion tensor imaging (DTI).

The complete profiles of R2∗ orientation dependence in human brain WM in vivo at 3 T were eventually established by leveraging orientation information from DTI, specifically using the principal diffusivity (λ1) direction within each imaging voxel [13], [14]. More importantly, an orientation-dependent function of sin2θ (or equivalently sin2θ [15]) was proposed, where θ is the angle between the λ1 direction and the B0 field. In a subsequent study involving two coronal formalin-fixed brain slabs at 7 T [16], an enhanced orientation-dependent function was developed, incorporating an additional sin4θ term. This refined function was derived from R2∗ measurements of brain specimens rotated at 18 different angles around the brain slab's norm direction, spanning from 0° to 170° relative to the B0 field. In this ex vivo study, the investigators also introduced a phase shift (φ0) into the proposed R2∗ orientation-dependent function. However, the biophysical basis of φ0 was not adequately elucidated. As a result, this phase shift was often overlooked in subsequent applications of the function to previously reported studies, where orientation information was derived from λ1 in DTI [17], [18], [19].

To better understand the origins of the observed orientation-dependent transverse relaxation in WM, a follow-up investigation was conducted at 7 T using rotating brain specimens, following a workflow similar to that described above [3]. In this comparative study, unlike the pronounced R2∗ orientational anisotropy, the measured R2 values exhibited minimal, if any, discernible orientation dependence. This observation led investigators to conclude that magnetic susceptibility effects were the dominant mechanism underlying anisotropic relaxation. However, the temperature-dependent anisotropies of R2∗ observed in the brain samples could not be adequately explained by the proposed relaxation mechanism, because the diamagnetic susceptibility of myelin is essentially independent of temperature and the static dephasing theory invoked therein also neglects spin motion. These conflicting findings clearly illustrate that the true origin of orientation-dependent R2∗ in WM remains elusive.

Meanwhile, researchers developed a hollow cylinder fiber model (HCFM) for WM based on magnetic susceptibility theory [7], [20]. In this model, the annular region of the cylinder is filled with molecular constituents exhibiting anisotropic magnetic susceptibilities, allowing prediction of R2∗ orientation dependence using an analytical function that includes a sin4θ or sin4θ term [7]. However, this analytical function alone was insufficient to accurately characterize the shape of orientation-dependent R2∗ profiles measured from rotating pig brain samples, unless an additional heuristic sin2θ term was included. Interestingly, a similar orientation-dependent R2 function, which incorporates the term sin4θ, with and without the term sin2θ, was also proposed for human brain WM in vivo at 3 T [8]. The underlying biophysical principle of the proposed model for R2 orientational anisotropy was the same susceptibility theory for R2∗ anisotropy but applied at a mesoscopic scale. Specifically, diffusion-mediated decoherence, arising from local susceptibility variations across axon fiber bundles, enhances the transverse relaxation measured at the macroscopic voxel level.

Another potential anisotropic relaxation mechanism is the well-known MAE [21], originated from rotationally restricted water molecules in highly ordered biological tissues [22], but this effect has not been adequately considered in existing orientation-dependent R2∗ or R2 relaxation models [16], [23]. This oversight is partially due to the markedly different profiles of measured R2∗ or R2 orientation dependence in WM, where enhanced relaxation rates are observed when axon fibers are oriented perpendicular to the B0 field, compared to parallel [12]. In contrast, conventional MAE profiles, such as those observed in tendons and deep-zone cartilage, exhibit the opposite pattern of transverse relaxation rate enhancement between these two orientations [21]. Recently, this seemingly contradictory R2∗ or R2 anisotropy in WM has been reconciled by employing a generalized MAE function that incorporates a phase shift [5], [6]. Remarkably, the proposed MAE-based function can be reformulated in terms of previously developed orientation-dependent R2∗ and R2 models, thereby suggesting a potentially ambiguous mechanism underlying anisotropic transverse relaxation in WM, as reported in prior literature [3], [15].

An orientation-dependent transverse relaxation rate provides essential specificity in neuro-microstructural imaging, as it directly reflects the extent of (de)myelination in WM [3], [5], [20], [24]. To accurately capture the underlying biophysical processes, it is essential to disentangle the previously proposed relaxation mechanisms—namely, the susceptibility effects and the MAE—in relation to R2∗ and R2 orientational anisotropies. These two competing anisotropic relaxation pathways exhibit distinct B0 field dependences: susceptibility effects scale with B02, whereas the MAE is independent of B0. Consequently, these mechanisms can, in principle, be distinguished by comparing orientation-dependent R2 across different B0 field strengths. This separation is critical for advancing our understanding of the unique anisotropic relaxation signatures associated with the highly organized microstructures in WM. Given the current ambiguity surrounding the mechanism of orientation-dependent transverse relaxation, this work aims to clarify the long-standing question regarding the origins of anisotropic transverse relaxation in human brain WM by examining orientation-dependent R2 values derived from both spin-echo and gradient-echo measurements at 3 T and 7 T, as previously published in the literature. Compared to our prior investigations [5], [6], this study represents a significant advance by identifying the dominant anisotropic relaxation pathway at clinically widely available B0 field strengths.