Three‐Dimensional Efficient Myelin‐Weighted Imaging Utilizing Direct Visualization of Short Transverse Relaxation Time Component (ViSTa)

ABSTRACT

Measuring myelin concentration in the brain has important implications in basic science and clinical practice. In MRI, myelin water imaging (MWI) has been suggested as a surrogate biomarker that provides high sensitivity and specificity for myelin. However, multi‐exponential fitting is ill‐conditioned, and it is sensitive to noise and artifacts, particularly in vivo. To overcome the ill‐conditioned fitting problem, the two‐dimensional ViSTa myelin‐weighted imaging technique was proposed, and it provides a substantially improved myelin‐weighted image. However, it is based on a two‐dimensional single‐slice acquisition scheme, and it is a limitation. In this study, a whole brain‐covered 3D ViSTa sequence, based on a 3D segmented echo planar imaging (EPI) sequence with a pair of slice selective inversion RF pulses, was proposed. To investigate the 3D ViSTa myelin weighted image, the distribution of myelin content in the white matter of the brain was measured using both conventional MWI and ViSTa MWI. The proposed 3D ViSTa method achieves a whole brain‐covered (FOV = 240 × 240 × 128 mm³) myelin water‐weighted image in less than 8 min (1.5 × 1.5 × 4 mm³) and does not require heavy post‐processing. Pseudo‐quantification (apparent MWF) can be provided by normalizing the ViSTa image with a PD‐weighted image. The voxel‐wise correlation between the conventional MWI and the 3D ViSTa yielded a mean correlation coefficient of 0.74 ± 0.03 (mean ± standard deviation of the five subjects), demonstrating a high spatial similarity in myelin‐weighted contrast between the two maps. The proposed 3D ViSTa with pseudo‐quantification may be useful in clinical applications when absolute quantification is not necessary.

The proposed 3D ViSTa achieved whole brain coverage myelin weighted image in less than 8 min, and it does not require heavy post‐processing to generate a pseudo‐quantitative myelin water fraction (apparent MWF) map. This may be useful in clinical applications when absolute quantification is not necessary.

Summary.

• 3D ViSTa provides high‐quality whole‐brain‐covered myelin water weighted images and pseudo quantitative myelin water fraction map in less than 8 min.

• ViSTa demonstrates myelin water concentration in the white matter of the brain.

• Conventional and ViSTa myelin water maps reveal a high spatial correlation.

1. Introduction

Myelin is an electrical insulator and a key substrate for saltatory conduction in the nerve system. It has been studied that the myelin sheath is a primary determinant for nerve conduction velocity (Waxman 1980). The concentration of myelin varies across the brain (Geyer et al. 2011; Prasloski, Rauscher, et al. 2012; Sprooten et al. 2019), and myelin abnormality affects information processing and cognition (Fields 2008). Thus, measuring myelin concentration in the brain has important applications in both neuroscience and clinical research.

In MRI, imaging myelin concentration in gray matter and white matter has been a topic of interest. Measuring cortical myelin has been suggested as a way to generate an in vivo myeloarchitectural map for brain parcellation (Foit et al. 2022; Geyer et al. 2011). For this purpose, a few methods such as quantitative R₁ imaging (Geyer et al. 2011; Lutti et al. 2014; Marques et al. 2010), $T_2^{*}$ imaging (Cohen‐Adad et al. 2012), and T ₁‐weighted/T ₂‐weighted imaging (Glasser and Van Essen 2011) have been proposed as an indirect (and nonspecific) approach to estimate myelin concentration in cortical gray matter. Imaging myelin in white matter has a longer history. Several imaging methods including diffusion tensor imaging (Song et al. 2002), magnetization transfer imaging (Levesque et al. 2010; Prevost et al. 2017; Sui et al. 2025), myelin water imaging (MacKay et al. 1994), and separating positive and negative susceptibility sources (Lee et al. 2024, 2023) have been explored as surrogate biomarkers for myelin. More recently, imaging methods to target myelin in white matter have been proposed using ultrashort echo time imaging (Seifert et al. 2017), multi‐contrast imaging for myelin volume mapping (Warntjes et al. 2016). These methods showed different levels of sensitivity and specificity to myelin (Alonso‐Ortiz et al. 2015; Duhamel et al. 2019; Laule et al. 2007; Lee et al. 2021).

Among the white matter myelin biomarkers, myelin water imaging (MWI) has been suggested as a sensitive and specific approach for imaging myelin (MacKay et al. 1994). This method measures T ₂ (or $T_2^{*}$) decay using a multi‐echo spin echo or gradient echo sequence (Du et al. 2007; Hwang et al. 2010; Kolind et al. 2009; MacKay et al. 1994). The measured data are fitted by multiple exponential decay functions to separate short T ₂ (or $T_2^{*}$) myelin water components from long T ₂ (or $T_2^{*}$) intra‐ and extra‐cellular water components. Unfortunately, the exponential decay functions are not orthogonal, and therefore, the fitting is ill‐conditioned. This results in severe artifacts in the estimated myelin water images (Cover 2008; Whittall and MacKay 1989; Reiter et al. 2009).

Previously, we proposed a new approach to acquire a short $T_2^{*}$ signal without using multi‐exponential fitting. This method, which is referred to as direct visualization of short transverse relaxation time component (ViSTa) (Oh et al. 2013), acquires the short $T_2^{*}$ signal in the range of myelin water (~10 ms at 3 T) based on T ₁ relaxation difference among the water pools (Labadie et al. 2014). When the characteristics of ViSTa signals were investigated, the patterns of $T_2^{*}$ decay (Oh et al. 2013), phase evolution (Kim et al. 2015), and magnetization transfer ratio (Lee et al. 2014) were similar to those of myelin water, confirming that the ViSTa signals primarily came from myelin water. Furthermore, ViSTa provides substantially improved image quality in comparison to the conventional methods. Several studies have validated its efficacy in diagnosing diseases in patients with multiple sclerosis, traumatic brain injury, and Alzheimer's disease (Choi, Hart, et al. 2019; Choi, Jeong, et al. 2019; Lim et al. 2022). However, the previous ViSTa sequence was designed only for a 2D single‐slice acquisition, which hindered further applications of this method (Kim et al. 2015; Oh et al. 2013). Hence, it is necessary to expand the method to cover a 3D volume.

In this study, a whole‐brain‐covered 3D ViSTa sequence based on a 3D segmented echo planar imaging (EPI) sequence with a pair of slice selective inversion RF pulses was proposed. With this sequence, we can achieve whole‐brain 3D volume coverage in less than 8 min. B ₁ inhomogeneity effect on ViSTa was verified with additional computer simulation. Then, the distribution of myelin content in the white matter of the brain was measured using both conventional MWI and ViSTa. For the conventional MWI, a modified GRASE sequence as described in Prasloski, Rauscher, et al. (2012) was implemented for a Siemens scanner. The myelin water fraction maps of the two methods were compared.

2. Methods

2.1. Computer Simulation for the Influence of B ₁ Inhomogeneity to ViSTa

A computer simulation was conducted to examine the impact of B ₁ inhomogeneity in ViSTa. The general form of the transverse magnetization as a function of T ₁ in ViSTa is as follows:

$$M_{\mathit{xy}}=M_0\left(\frac{1-\left(1-\text{cosα}\right)e^{-\frac{{\mathit{TI}}_2}{T_1}}-\cos \alpha \left(1-\text{cosα}\right)e^{-\frac{{\mathit{TI}}_1+{\mathit{TI}}_2}{T_1}}-{\cos }^2\alpha \times e^{-\frac{\mathrm{TR}}{T_1}}}{1-{\cos }^2\alpha \times \text{cosβ}\times e^{-\frac{\mathit{TR}}{T_1}}}\right)e^{-\mathrm{TE}/T_2^{*}}$$ (1)

Where α is a flip angle for the first two inversion RF pulses and β is a flip angle for the excitation RF pulse in the ViSTa sequence. TI₁ and TI₂ represent the time between the 1st and 2nd inversion, 2nd inversion, and the excitation RF pulse, respectively. For the simulation, TI₁ = 560 ms, TI₂ = 220 ms, and TR = 1160 ms were used (Oh et al. 2013). In the ideal case of the ViSTa, we assumed α is 180° and β is 90°. For checking the influence of the variation of flip angles due to B ₁ inhomogeneity, α and β were swept from 140° to 220° and from 70° to 110°, respectively.

2.2. 3D ViSTa MWI Sequence

To cover a 3D volume, a new ViSTa sequence was developed as shown in Figure 1. The readout was implemented using a 3D segmented (or multi‐shot) EPI acquisition. The EPI ghosting artifact was corrected by acquiring navigator echoes at the start of the sequence (Bruder et al. 1992; Reeder et al. 1999). Since the myelin water signal has a short $T_2^{*}$ (~10 ms), reducing TE improves the signal‐to‐noise ratio in ViSTa. To reduce TE, a partial k‐space acquisition (= 6/8) was applied in phase encoding. For the same reason, an excitation RF was designed with minimum phase using the Shinnar‐Le Roux (SLR) algorithm (Pauly et al. 1991). The designed parameters for the excitation RF pulse were flip angle = 90°, duration = 2.56 ms, number of steps = 512, time‐bandwidth product = 13.31, effective pass‐band ripple = 1%, and effective rejection‐band ripple = 0.001%. The time between the peak and the end of the RF was reduced from 1.28 ms (linear phase RF) to 0.19 ms (minimum phase RF) saving 1.09 ms. For long T ₁ suppression, a scheme of double inversion was employed as in 2D ViSTa (Oh et al. 2013). The inversion pulse was a hyperbolic‐secant pulse for adiabatic inversion (duration = 10.24 ms and bandwidth = 1 kHz). Since spins are sequentially excited or inverted in time with a frequency‐modulated RF pulse like the hyperbolic‐secant RF pulse, the inversion time of the pulse was defined to be the instant when 85% of the spins are inverted during the pulse duration, that is, 2 ms after the center of the pulse in our case (Park and Garwood 2009). Hence, TI₁ was defined as the duration between 2 ms after the center of the first inversion RF pulse and 2 ms after the center of the second inversion RF pulse; TI₂ was the duration between 2 ms after the center of the second inversion RF pulse and the center of the excitation RF pulse; TD was the duration from the peak of the excitation RF pulse to 2 ms after the center of the first inversion RF pulse (Figure 1). These three time intervals were set to be as follows: TI₁ = 560 ms, TI₂ = 220 ms, and TD = 380 ms (Oh et al. 2013). Since arteries and fats appeared to be bright in our previous study with 2D ViSTa images due to the effect of the blood inflow and the short T ₁ of fats, respectively (Oh et al. 2013), saturation pulses were introduced to suppress both effects. The flow saturation pulse (flip angle = 90°, duration = 3.84 ms, and TBW = 8) was located between the 1st and 2nd inversion RF pulses. The time between the peak of 1st inversion RF and the peak of the flow saturation RF was defined as saturation delay time (Sat_delay). The delay time of the flow saturation pulse (Sat_delay = 300 ms) was empirically decided. The flow saturation band covered 11 cm thickness in the lower head and neck areas and was positioned below an imaging volume with a 5 mm gap. The fat saturation pulse (flip angle = 110°, duration = 5.12 ms, and TBW = 2) was applied 7.18 ms before the excitation RF pulse. These additional RF pulses may induce the MT effects thereby reducing myelin water signals, but the effects were shown to be limited (Oh et al. 2013).

FIGURE 1.

A schematic diagram of the 3D ViSTa pulse sequence and timing. Data acquisition utilizes a 3D segmented EPI readout. To reduce TE, a SLR pulse with minimum phase was used for excitation.

2.3. 3D Conventional MWI Sequence

A 3D multi‐echo spin echo sequence for conventional MWI was implemented as described in Prasloski, Rauscher, et al. (2012) This sequence acquires multiple phase encoding lines (three echoes) per refocusing pulse to reduce total scan time. To correct for the EPI ghosting artifact, navigator echoes were acquired at the beginning of the sequence (Bruder et al. 1992; Reeder et al. 1999). Truncated sinc‐shaped RF pulses were used for both slab‐selective excitation (flip angle = 90°, duration = 2.56 ms, and time‐bandwidth product = 5.2) and slab non‐selective refocusing (flip angle = 180°, duration = 2.56 ms, and time‐bandwidth product = 2) RF pulses. To suppress unwanted fat signals, a fat saturation pulse (flip angle = 110°, duration = 5.12 ms, and TBW = 2) was applied before the excitation RF pulse.

2.4. MRI Scans

Five healthy volunteers (mean age = 34.8 ± 2.3) who provided written consent (IRB approved) were scanned at a 3 T MRI (Trio, Siemens, Erlangen, Germany). For data acquisition, a 32‐channel phased array coil was used. After a localizer scan, shimming was performed over the whole brain. Then, the following data were obtained.

2.4.1. ViSTa

Scan parameters were FOV = 240 × 240 × 128 mm³, matrix size = 160 × 160 × 32, resolution = 1.5 × 1.5 × 4 mm³, TR = 1160 ms, TE = 6.5 ms, TI₁ = 560 ms, TI₂ = 220 ms, TD = 380 ms, readout bandwidth = 801 Hz/pixel, partial k‐space in phase encoding direction = 6/8, phase encoding lines per excitation = 11, number of EPI segments = 11, and scan time = 6 min 53 s. For the comparison with the conventional MWI, the ViSTa data were acquired twice and averaged in order to match the scan time. To generate a myelin‐weighted image, a proton density (PD) weighted GRE scan that had the same readout as the 3D ViSTa sequence was obtained with TR = 75 ms, TE = 6.5 ms, and flip angle = 5° (scan time = 30 s). The total scan time of the ViSTa (2 averages) and the PD‐weighted scan was 14 min 16 s.

2.4.2. Conventional MWI

Scan parameters were FOV = 240 × 240 × 80 mm³, matrix size = 160 × 160 × 20, resolution = 1.5 × 1.5 × 4 mm³, TR = 1000 ms, TE = 10 ms to 320 ms with echo spacing of 10 ms, number of echoes = 32 echoes, readout bandwidth = 1002 Hz/pixel, phase encoding lines per each segment = 3, partial k‐space in slice = 6/8, and scan time = 13.33 min. The image center and orientation were matched with the ViSTa scan.

2.5. Data Processing

Because several factors, such as the $T_1$‐weighting, cross‐relaxation, and chemical exchange effects (Oh et al. 2013), ViSTa may underestimate myelin water fraction and, therefore, have limited accuracy in myelin water quantification. However, pseudo‐quantification, apparent myelin water fraction (aMWF), can be provided by dividing a ViSTa image with a proton density (PD)‐weighted image and scaling the result. For a PD‐weighted image, additional 3D GRE data using a small flip angle (5°) and short TR (75 ms) was obtained. Due to the small flip angle, it has PD‐weighted image contrast. The scan time was 30 s and the amount of distortion was the same as 3D ViSTa. The normalization of 3D ViSTa was performed first with a multiply magnitude‐threshold mask generated from 3D GRE data to avoid noise amplification. After that, 3D ViSTa data was divided by 3D GRE. To compensate $T_1$ and $T_2^{*}$ weighting in ViSTa and GRE scaling factors were multiplied. The scaling factor was calculated as follows:

$$\left(\text{scaling factor}\right)=k·\frac{\left(T_1-\text{weighting of}\mathrm{GRE}\right)\times \left(T_2^{*}-\text{weighting of}\mathrm{GRE}\right)}{\left(T_2^{*}-\text{weighting of ViSTa}\right)}$$

$$=k·\left\{\left[\frac{\left(1-\exp \left(-{\mathrm{TR}}_{\mathrm{G}}/T_{1,\mathrm{G}}\right)\right)·\sin {\mathrm{\theta }}_{\mathrm{G}}}{\left(1-\cos {\mathrm{\theta }}_{\mathrm{G}}·\exp \left(-{\mathrm{TR}}_{\mathrm{G}}/T_{1,\mathrm{G}}\right)\right)}\right]·\left(\exp \left(-{\mathrm{TE}}_{\mathrm{G}}/T_{2,\mathrm{G}}^{*}\right)\right)\right\}/\left(\exp \left(-{\mathrm{TE}}_{\mathrm{V}}/T_{2,\mathrm{V}}^{*}\right)\right)$$ (2)

where the subscript G and V represent GRE and ViSTa, respectively, and the $\theta$ is the flip angle. The constant k includes factors such as myelin water T ₁‐weighting, cross‐relaxation, water exchange, and MT effects in ViSTa. In this study, the effects of myelin water T ₁‐weighting were only considered for determining k. The nominal T ₁ of ViSTa was assumed to be 118 ms (the median value of all measurements in Labadie et al. (2014)). With this T ₁, the signal intensity of the short $T_2^{*}$ component in ViSTa is reduced to 69% of the fully relaxed signal intensity (Equation (1) in Oh et al. (2013)). The nominal T ₁ of GRE was set to 800 ms and the nominal $T_2^{*}$ of GRE and ViSTa were assumed to be 50 and 10 ms, respectively.

To generate a MWF map in the conventional MWI, the multi‐echo data were corrected for stimulated echoes and fitted to multiple exponential decay functions (200 logarithmically spaced T ₂ from 0.015 to 2 s) using a regularized nonnegative least‐squares fitting method proposed by Prasloski et al. (Prasloski, Madler, et al. 2012).

Once MWF map and aMWF map were obtained from the conventional MWI and the ViSTa, respectively, the two maps were co‐registered. The co‐registration parameters (six degrees of freedom) were estimated using the 1st echo conventional 3D MWI image and the ViSTa aMWF map (SPM5; (Ashburner and Friston 2005)). After co‐registration, a voxel‐wise correlation between the two maps was calculated within a white matter mask that was determined from the 1st echo conventional 3D MWI images using the automatic segmentation function of SPM5 (Ashburner and Friston 2005). Because of the artifacts caused by B ₁ inhomogeneity in the conventional MWI (white circle in Figure 3), a few inferior slices were excluded from calculating the correlation.

FIGURE 3.

(A) A MWF map from the conventional MWI and (C) a zoomed image of a red rectangular box in (A). (B) An aMWF map from the ViSTa MWI and (D) a zoomed image of a blue rectangular box in (B).

3. Results

3.1. Transverse Magnetization as a Function of T ₁

Table 1 shows the effects of transmit B ₁ inhomogeneity simulation. The level of B ₁ inhomogeneity in 180° inversion RF and 90° excitation RF pulses is shifted from −20% to +20%. These results show sensitivity to the B ₁ field in the inversion pulses. When the B ₁ variation of the inversion pulse is −20%, the long T ₁ signal is increased up to 12.6 times. For 5% B ₁ variation, it is 1.4 times that of no variation. On the other hand, the variation in the excitation pulse has little effect. The long T ₁ signal shows less than 10% change when B ₁ variation is ±20%.

TABLE 1.

The effects of transmit B ₁ inhomogeneity simulation. (Each value represents ratio of long T ₁ signal value with and without B ₁ variation).

		B 1 inhomogeneity level of inversion RF (%)
		−20	−15	−10	−5	0	5	10	15	20
B 1 inhomogeneity level of excitation RF (%)	−20	13.1	7.5	3.9	1.8	1.1	1.5	3.3	6.6	11.7
	−15	13.0	7.4	3.8	1.8	1.1	1.5	3.3	6.5	11.6
	−10	12.9	7.4	3.8	1.8	1.0	1.5	3.2	6.4	11.4
	−5	12.7	7.2	3.7	1.7	1.0	1.5	3.2	6.3	11.3
	0	12.6	7.2	3.7	1.7	1.0	1.4	3.1	6.2	11.2
	5	12.4	7.1	3.6	1.7	1.0	1.4	3.1	6.1	11.0
	10	12.3	7.0	3.6	1.7	1.0	1.4	3.0	6.1	10.9
	15	12.2	6.9	3.5	1.6	1.0	1.4	3.0	6.0	10.8
	20	12.1	6.8	3.5	1.6	0.9	1.4	2.9	5.9	10.7

Figure 2 represents a 3D MWF map from the conventional MWI (left two columns) and a 3D aMWF map from the 3D ViSTa (right two columns). The distribution of MWF (or aMWF) shows a large variation across the brain, suggesting a non‐uniform distribution of myelin in the brain. Once the two maps are compared, they reveal overall similar spatial distributions although the display range is different (0.0% to 30.0% in the conventional MWI and 0.0% to 9.0% in the ViSTa; see Discussion). The voxel‐wise correlation yielded a mean correlation coefficient of 0.74 ± 0.03 (mean ± standard deviation of the five subjects), demonstrating a high similarity between the two maps. In both maps, high MWFs are observed in optic radiation, corpus callosum, internal capsule, longitudinal fasciculus, and white matter near the motor/sensory cortex. In the conventional 3D MWF map, lower slices show artifacts (white circle) that are most likely caused by B ₁ inhomogeneity. When the image quality of the two maps is compared, the ViSTa aMWF map exhibits better image quality with reduced speckle‐like noise.

FIGURE 2.

A MWF map from the conventional MWI (left) and an aMWF map from the 3D ViSTa MWI (right).

The image quality of each method is better appreciated in grayscale, as shown in Figure 3. A MWF map from the conventional MWI (Figure 3A) shows a substantially noisier image than an aMWF map from the ViSTa (Figure 3B). The difference becomes more distinguishable in the areas of low MWFs (Figure 3C,D). This may be related to the unreliable estimation of MWFs in low SNR of conventional MWI (Cover 2008). The overall myelin water distribution produced a similar pattern with both methods.

The normalized 3D ViSTa images are shown in Figure 4. In Figure 4, 16 slices out of 32 slices are displayed with a color scale bar. As shown, with the 3D ViSTa sequence we can achieve whole brain coverage (240 × 240 × 128 mm³) in less than 8 min (6.53 min for 3D ViSTa image and 30 s for 3D GRE image). The mean SNR after Rician noise bias correction over 5 subjects is 26.3 ± 1.6. The 3D ViSTa aMWF map reveals higher signal intensity distribution (> 8.0%) in crus cerebri (CC; S11), optic radiation (OR; S11–S13), splenium of corpus callosum (SCC; S13–S16), genu of corpus callosum (GCC; S14–S16), internal capsule (or cortico spinal tract, IC (or CST); S13–S16), SLF (S18–S19) and motor (S23–S24) areas than neighboring white and gray matter areas. These areas contain long‐range fibers. Hence, this result might suggest that fibers connecting long distances tend to have higher myelination.

FIGURE 4.

A whole brain (240 × 240 × 128 mm³) quantitative ViSTa aMWF map at 1.5 × 1.5 × 4 mm³ in less than 8 min (no average). Areas of high myelin concentration include optic radiation, corpus callosum, internal capsule, longitudinal fasciculus, and white matter near motor/sensory cortex. A total of 16 slices out of 32 slices are shown.

Compared with a study of conventional 3D myelin water imaging using a 3D GRASE sequence (Figure 4 in Prasloski, Rauscher, et al. (2012)), the result of an MWF map from 3D ViSTa has similar distribution patterns except for the genu of the corpus callosum region (0.00 to 0.08 and the scale in Prasloski, Rauscher, et al. (2012) is 0.0 to 0.3). In conventional MWF, the genu region shows low MWF due to B ₁ inhomogeneity.

4. Discussion

In this study, we proposed a new method to acquire a whole brain‐covered 3D ViSTa myelin‐weighted image. In addition, we explored the 3D spatial distribution of myelin water in the white matter of the brain using two different MWI methods: a conventional MWI using a modified GRASE sequence and a newly developed 3D ViSTa. The two methods revealed a high correlation in their spatial distribution of myelin water fraction that varied across the brain. Compared to the conventional approach, the 3D ViSTa aMWF provided high‐quality myelin water‐weighted images generating a whole brain (FOV = 240 × 240 × 128 mm³) map of myelin water at 1.5 × 1.5 × 4 mm³ resolution in less than 8 min. The resulting myelin water map illustrated a high concentration of myelin water in the areas of long‐ranging fibers.

The anisotropic 1.5 × 1.5 × 4 mm³ matrix was chosen to mirror clinical routine protocols; however, fully isotropic imaging is technically feasible with 3D ViSTa. A 2 mm³ isotropic dataset with whole‐brain coverage (80 slices) can be obtained in 8 min 47 s, including the PD‐weighted GRE required for normalization. This flexibility allows investigators to trade spatial resolution against scan time or SNR according to study goals.

When compared to the conventional MWF, aMWF obtained from ViSTa has smaller values (Figure 2). This is because of the incomplete knowledge of the scaling factor (k) in Equation (2). Additional $T_1$‐weighted signal loss from the double inversion (Oh et al. 2013) can be one of the explanations. In addition to this, cross‐relaxation and chemical exchange between myelin water and macromolecules, magnetization transfer effects of fat saturation and flow saturation RF pulses also can support this observation (Bjarnason et al. 2005; Deoni et al. 2008; Dortch et al. 2013; Harkins et al. 2012; Kalantari et al. 2011; Stanisz et al. 1999; Vavasour et al. 2000). For this reason, the ViSTa MWF is referred to as “apparent” MWF (or aMWF). However, the method can be still used for group comparisons or longitudinal studies as long as the same scaling factor is used.

In this study, the GRASE‐based MWF maps were reconstructed with the non‐negative least‐squares (NNLS) algorithm (Prasloski, Madler, et al. 2012). Fitting using a nonlinear least squares (NLLS) approach under a multi‐exponential signal decay model may be another option, which might improve the output as demonstrated in a previous study (Faulkner et al. 2024; Li et al. 2021).

In this study, normalization employed a short‐TR PD‐weighted GRE to minimize T ₁ weighting, yet residual T ₁ effects may still influence aMWF, particularly in tissues where T ₁ deviates due to age‐related or pathological variation. Future work may explore multi‐angle normalization or integrated T ₁ mapping to mitigate this variability.

One of the primary goals of this paper was to evaluate consistency in the spatial distributions of the myelin‐weighted contrast between 3D ViSTa and conventional MWI rather than claiming quantitative accuracy against conventional MWF. Without additional studies including numerical simulations or histology, quantification accuracy is difficult to evaluate. Nevertheless, ViSTa's robust correspondence with conventional MWF, together with prior evidence that ViSTa signal obeys $T_2^{*}$ decay, phase evolution, and MT behavior characteristic of myelin water (Kim et al. 2015; Lee et al. 2014; Oh et al. 2013), supports its practical utility despite the quantification issue.

In terms of the magnetization characteristic of ViSTa, the accuracy of a flip angle is important for the analysis of ViSTa. In our study, we confirmed the influence of the flip angle variation due to the B ₁ inhomogeneity effect in ViSTa. As shown in Table 1, the suppression level of the long T ₁ component is sensitive to the B ₁ variation in the inversion RF pulse. This potential compound is carefully avoided in our ViSTa sequence as described in Method.

In ViSTa, a PD‐weighted GRE (TR = 75 ms) with a small flip angle (= 5°) was acquired to generate the aMWF. In our previous study, a larger flip angle (28°; Ernst angle) was used.²⁸ The difference between them is compensated by the scaling factor in Equation (2) when generating the aMWF map. The use of small or large flip angles has different advantages and disadvantages. A small flip angle (~5°) in the GRE induces a large CSF signal, thereby helping to suppress CSF regions when the ViSTa image is divided by the GRE images (Figure 4). On the other hand, when an Ernst angle is employed, the aMWF map contains unwanted CSF signal due to a reduced CSF signal in the GRE (see figure 4f in Oh et al. 2013). Compared to the small flip angle, the Ernst angle increases SNR and is less sensitive to B ₁ inhomogeneity since the GRE signal around the Ernst angle has a smaller change than in the small flip angle. Hence, one may need to consider these factors in acquiring the GRE for the ViSTa myelin‐weighted image. For quantitative comparison, the Ernst angle may be preferred due to a higher SNR and a better B ₁ characteristic.

In this study, the 3D ViSTa sequence was implemented using a 3D segmented (or multi‐shot) EPI readout. Using minimum phase SLR excitation pulse (Pauly et al. 1991), TE was reduced by 1.09 ms. The minimum TE of the current scan parameter was 6.51 ms, and if we can reduce the minimum TE it may reduce $T_2^{*}$ decay and provide higher SNR and improved quantification. Using a 3D spiral readout with a double inversion RF pulse sequence, the time between excitation and center of k‐space can be minimized, resulting in further improved SNR.

The double‐inversion RF pulses in ViSTa are also used in a double‐inversion recovery (DIR) sequence, which is utilized for cortical lesion detection in multiple sclerosis patients (Geurts et al. 2005). The primary difference is the targets for suppression. In the DIR sequence, the timing for the double inversion RF pulses is optimized to suppress CSF and white matter. In the ViSTa sequence, however, the timing is calibrated to suppress any T ₁ signal longer than T ₁ of 750 ms. Hence, the sequence suppresses the signals from CSF, gray matter, and axonal/extracellular water in white matter, leaving myelin water signals. Another difference between the two sequences is the readout schemes. The DIR is implemented with a fast spin echo (or turbo spin echo) readout, providing high‐quality images. This readout, however, affects T ₁ recovery when B ₁ inhomogeneity exists. In ViSTa, this effect may hamper the proper suppression of long T ₁ signals.

The 3D ViSTa method provides information of myelin concentration in the brain and may potentially be used as a measure of nerve conductivity when combined with tractography from DTI (Horowitz et al. 2015). Another potential application is to study neuroplasticity in white matter where myelination has been suspected as a source for change in FA values of DTI (Scholz et al. 2009).

T₂based MWI (MacKay et al. 1994) is considered the conventional approach but it requires a long scan time and is sensitive to B₁ inhomogeneity. $T_2^{*}$‐based MWI (Du et al. 2007) offers faster acquisition but suffers from susceptibility and eddy current artefacts (Shin et al. 2019). mcDESPOT (Deoni et al. 2008) and BMC‐mcDESPOT (Bouhrara and Spencer 2017) achieve 1.7 mm isotropic whole‐brain mapping in 15 min via multi‐flip‐angle steady‐state acquisitions at the cost of sensitive nonlinear model fitting that may suffer from motion and registration issues. ViSTa, in contrast, requires only one myelin‐weighted and one PD‐weighted volume, avoiding extensive fitting and reducing sensitivity to motion or scanner instability. Although ViSTa does not furnish absolute MWF, its simplicity and efficiency make it a clinically practical alternative for imaging the myelin‐related contrast in vivo (Faulkner et al. 2024; Lee et al. 2021).

5. Conclusions

In this study, we proposed a new method to acquire whole brain‐covered 3D ViSTa. The influence of B ₁ inhomogeneity in ViSTa was confirmed. The proposed 3D ViSTa method achieved whole brain coverage ViSTa imaging in less than 8 min, and it does not require heavy post‐processing to generate an aMWF map. Pseudo‐quantification (apparent MWF) can be provided by normalizing the ViSTa image with a PD‐weighted image. This may be useful in clinical applications when absolute quantification is not necessary.

Oh, S.‐H. , Lee G., and Lee J.. 2025. “Three‐Dimensional Efficient Myelin‐Weighted Imaging Utilizing Direct Visualization of Short Transverse Relaxation Time Component (ViSTa).” Human Brain Mapping 46, no. 12: e70307. 10.1002/hbm.70307.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.