Application of q-Space Diffusion MRI for the Visualization of White Matter

Abstract

White matter abnormalities in the CNS have been reported recently in various neurological and psychiatric disorders. Quantitation of non-Gaussianity for water diffusion by q-space diffusional MRI (QSI) renders biological diffusion barriers such as myelin sheaths; however, the time-consuming nature of this method hinders its clinical application. In the current study, we aimed to refine QSI protocols to enable their clinical application and to visualize myelin signals in a clinical setting. For this purpose, animal studies were first performed to optimize the acquisition protocol of a non-Gaussian QSI metric. The heat map of standardized kurtosis values derived from optimal QSI (myelin map) was then created. Histological validation of the myelin map was performed in myelin-deficient mice and in a nonhuman primate by monitoring its variation during demyelination and remyelination after chemical spinal cord injury. The results demonstrated that it was sensitive enough to depict dysmyelination, demyelination, and remyelination in animal models. Finally, its utility in clinical practice was assessed by a pilot clinical study in a selected group of patients with multiple sclerosis (MS). The human myelin map could be obtained within 10 min with a 3 T MR scanner. Use of the myelin map was practical for visualizing white matter and it sensitively detected reappearance of myelin signals after demyelination, possibly reflecting remyelination in MS patients. Our results together suggest that the myelin map, a kurtosis-related heat map obtainable with time-saving QSI, may be a novel and clinically useful means of visualizing myelin in the human CNS.

Significance statement: Myelin abnormalities in the CNS have been gaining increasing attention in various neurological and psychiatric diseases. However, appropriate methods with which to monitor CNS myelin in daily clinical practice have been lacking. In the current study, we introduced a novel MRI modality that produces the "myelin map." The myelin map accurately depicted myelin status in mice and nonhuman primates and in a pilot clinical study of multiple sclerosis patients, suggesting that it is useful in detecting possibly remyelinated lesions. A myelin map of the human brain could be obtained in <10 min using a 3 T scanner and it therefore promises to be a powerful tool for researchers and clinicians examining myelin-related diseases.

Keywords: MRI; demyelination; multiple sclerosis; myelin; remyelination.

Figure 1.

Development of the myelin map. The PDF of water diffusion acquired from the cervical spinal cord in adult common marmosets reveals that the diffusion distribution of white matter (anterolateral and posterolateral funiculi) is more leptokurtic than that of gray matter (A). Note that PDF curves obtained with a time-saving protocol (9 b-values; hereafter used for the myelin map; solid lines) had sufficient accuracy to differentiate gray and white matter when compared with the time-consuming full-scale protocol (18 b-values; dotted lines). The histogram of NLD calculated from the PDF is shown in B. The NLD heat map and the myelin map created from the PDF are shown in C. a.u., Arbitrary units.

Figure 2.

Spinal cord myelin map of myelin-deficient mutant mice. T2WI, FA map, various DTI maps (i.e., AD map, RD map, and MD map), and the myelin map, as well as LFB staining, immunohistochemical staining against MBP, and electron microscopic (EM) analysis of postmortem spinal cords of wild-type (wt) and shiverer (shi/shi) mutant mice are shown in A. For quantitative analysis, the sensitivity of various modalities to differentiate white matter from gray matter was discerned using relative contrast values, calculated by first determining the absolute difference in 8-bit grayscale values (GV; 0–255) between ROIs (dorsal funiculus; orange box) and a reference (anterior horn; red box) and then dividing this value by 255. Relative contrast values exceeding 0.05 were considered to be clinically significant. All of the MRI modalities, with the exception of the MD map, were considered to be significant (C). To elucidate the myelin specificity of observed contrast, relative contrast values were compared between wt and shi/shi mice. Results indicated that the myelin map had superior myelin specificity relative to the FA, AD, and RD maps, whereas T2WI and LFB had similar specificity compared with the myelin map (D). Comparison of the FA map, the myelin map, LFB staining, and immunostaining against PLP-1 in jimpy mutant mice also confirmed the high accuracy of the myelin map in detecting myelin-related signals (B). Scale bars: LFB, MBP, and PLP-1, 500 μm; EM, 5 μm. a.u., Arbitrary units. #p < 0.05, Mann–Whitney U test. Error bars indicate SEM.

Figure 3.

Brain myelin map of myelin-deficient Shiverer mutant mice. T2WI, FA map, various DTI maps (i.e., AD map, RD map, and MD map), the myelin map, and histological LFB staining of postmortem wild-type (wt) and shiverer (shi/shi) mouse brains are shown (A). For quantitative analysis, relative contrast values were calculated similarly to the spinal cord analysis (Fig. 2). Myelinated fibers in the corpus callosum and cerebral cortex were selected as ROIs (orange boxes) and the molecular layer of cerebral cortex, in which few myelinated fibers exist, was used as a reference (red box). Relative contrast values exceeding 0.05 were considered to be clinically significant. All of the MRI modalities, with the exception of the AD map, were considered to be significant in the corpus callosum, whereas only the myelin map and MD map were significant in the cerebral cortex (B). Relative contrast ratios were calculated similarly to the spinal cord analysis (Fig. 2). Results indicated that the myelin map was the most myelin-specific modality, with a ratio that was fairly comparable to that of histological LFB staining (C). Scale bar, 1 mm. #p < 0.05, Mann–Whitney U test. Error bars indicate SEM.

Figure 4.

Myelin map of common marmosets with chemically induced focal demyelination. Axial images of spinal cord obtained 1 week after chemically induced demyelination, showing T2WI and myelin map results in a live animal, and postmortem LFB, semithin sections, and electron microscopic (EM) analysis (A). The same set of axial images was obtained 6 weeks after demyelination (B). The g-ratio, calculated by dividing axonal diameter by myelinated fiber diameter (e.g., the g-ratio for a completely demyelinated axon is 1), shows decreased values at 6 weeks after injury (0.74 ± 0.070; mean ± SE) compared with 1 week postinjury (0.90 ± 0.095), suggestive of remyelination (C). Quantitative comparative analysis of demyelinated areas in the dorsal funiculus using a myelin map in a live animal and postmortem LFB (n = 4 each for 1 and 6 weeks after injury) suggests the high accuracy of the myelin map in detecting demyelinated areas (D). Note that T2WI failed to detect demyelinating lesions clearly in this model. a.u., Arbitrary units. Scale bars: LFB, 1 μm; EM, 10 μm. Error bars indicate SEM.

Figure 5.

Remyelination activity detected in chronological myelin maps of common marmosets after chemically induced focal demyelination. Chronological, 3D, color-coded myelin maps sequentially obtained from the same animal as in Figure 4 reveal signal loss due to chemical demyelination in the dorsal funiculus at 1 week (arrowheads), as well as signal recovery at 6 weeks. a.u., Arbitrary units.

Figure 6.

Myelin map of normal human brain. The whole-brain myelin map of a 40-year-old healthy male volunteer is shown (A). Note that the high sensitivity of the myelin map allows detection of myelin signals in the cerebral peduncles of the midbrain and in the pyramidal tracts in the medulla oblongata (B and C, respectively). The corresponding T2-weighted images are shown in D–F.

Figure 7.

Comparison of myelin map, FA map, and DTI maps in normal human brain. Axial T2WI, FA map, DTI maps (i.e., AD map, RD map, and MD map), and the myelin map of a normal human brain (28-year-old healthy female volunteer) are shown (A). For quantitative analysis, the sensitivity of various modalities to differentiate myelinated areas from nonmyelinated areas was evaluated by relative contrast values, calculated by first determining the absolute difference in 8-bit grayscale values (GV; 0–255) between ROIs [orange boxes; deep white matter (WM) or cortical WM] and a reference (red box; CSF in the lateral ventricle), and then dividing this value by 255 (B). Note that the myelin map shows superior contrast (i.e., higher relative contrast values) to the T2WI, FA map, and other DTI maps, not only in deep WM but also in cortical WM (C). #p < 0.01 between the myelin map compared with all other modalities by Mann–Whitney U test.

Figure 8.

Myelin maps of MS patients. Axial T2WI and myelin map are shown for the brains of 2 patients: a relapsing-remitting MS patient (patient #1; 43-year-old male; A and B, respectively) and a primary progressive MS patient (patient #2; 40-year-old male; C and D, respectively). Detailed clinical characteristics are summarized in Table 1. Red arrows indicate T2-hyperintense lesions with positive myelin signals in the myelin map (suggestive of possibly remyelinated lesions).

Figure 9.

Myelin signals are virtually absent in T1 black holes. Axial T1WI, T2WI, and myelin maps are shown for the brains (deep cerebral white matter) of 2 patients: a relapsing-remitting MS patient (patient #1; 43-year-old male; A and B) and a primary progressive MS patient (patient #2; 40-year-old male; C). Red asterisks indicate T1-hypointense lesions (T1 black holes) where myelin signals are always absent in the myelin map. These data are consistent with those of a previous radiopathological correlation study (Barkhof et al., 2003).

Figure 10.

Sequential myelin maps during recovery after an attack in an MS patient. The figure shows a sagittal T2WI of the cervical spinal cord at onset of acute relapse in a relapsing-remitting MS patient (patient #1; 43-year-old male) (A) and sequential axial T1-weighted images enhanced with gadolinium-based contrast agents (Gd-T1WI), T2WI, and myelin maps of the spinal cord at C4/5 at onset (pretreatment), 2 weeks after IVMP, and 10 weeks after IVMP (B, C, D, respectively). EDSS scores were 2.0, 1.5, and 0 at onset, 2 weeks after IVMP, and 10 weeks after IVMP, respectively. Abnormal gadolinium enhancement suggesting a breakdown of the blood–brain barrier is evident during the first 2 weeks after IVMP (white arrowheads). Although T2WI shows no significant differences throughout the study period (asterisks), the gadolinium enhancement disappears and the myelin map depicts the reappearance of myelin signals (black arrowheads, possibly reflecting remyelination), which correlated well with the clinical recovery.