Fig. 1 shows the procedure for producing a MWF image. On a 3.0 Tesla MR scanner system (Discovery MR750, GE Healthcare, Waukesha, WI, USA) and a standard head coil receiver, a healthy volunteer (24-year-old male) study was performed using three-dimensional spoiled gradient-echo (3D-SPGR) with and without FS. The imaging parameters were as follows: TE, 0.032, 2.2, 4.6, 6.9, and 10 ms; repetition time (TR), 265.1 ms; field of view, 200 mm; matrix size, 256×256; band width, 976.6 kHz; slices, 26; slice thickness, 5 mm; flip angle (FA), 24 degrees. Then, the TEs were set coherently at an in-phase time taking into consideration the signal attenuation effect (i.e., Dixon effect) due to fat phase variation. The CHESS method was used for fat suppression. Next, B₁ inhomogeneity correction was performed to the acquired UTE-SPGR dataset. B₁ correction was derived from the double angle method. This method uses the MR images of two FAs using a gradient echo (GRE) sequence. The dataset for B₁ correction was acquired with 2000 ms TR, 5.8 ms TE, 40 degrees FA α, and 80 degrees FA 2α. The following formula is used for B₁ correction:

Fig. 2: B1 correction.

where S_α and S_2α are the signal intensities from the α and 2α FA data, respectively. Then, calculated B₁ was applied to pixel-by-pixel of MR imaging data. Next, T₂^* values of two-components (T_2,s^* and T_2,l^*) were calculated from each UTE-SPGR dataset with and without FS using the bi-exponential fitting model, and then each T₂^* map was created. Then, the Levenberg-Marquardt algorism was applied. The following bi-exponential fitting model is used [5]: 

Fig. 3: Bi-exponential fitting model.

Finally, MWF image was generated. The formula for calculating MWF is as follows: 

Fig. 4: MWF.

  Fig. 5 shows a magnitude image for the Region of Interest (ROI) setting. ROI analysis was performed on different WM areas (bilateral frontal and opposite) of each T₂^* map and the mean and standard deviation (SD) were measured. The ROI was set to 5 × 5 pixels. The ROI placement is shown in Fig. 5. All imaging analysis was applied using an in-house MATLAB program (Mathworks, Natick, MA, USA).

  The results of this experiment are as follows: Fig. 6 shows magnitude images without FS and acquired using multi-echo SPGR. 

  Table 1 shows each measurement T₂^* values of frontal WM and opposite WM areas obtained by ROI analysis. T_2,s^* and T_2,l^* values with FS showed a lower value than those without FS. 

Table 1 Measurement T₂^*values at each ROI.

Notes: Data are presented as mean ± SD.

  Fig. 7 shows T_2,s^* and T_2,l^* maps derived from UTE-SPGR dataset. T_2,s^* map shows a high signal in the WM area (Fig. 7a, b). T_2,s^* map without FS showed a slightly higher value than with FS (Fig. 7b, black arrows). On the contrary, T_2,l^* map showed a low signal in the WM area and there was too great of a difference between the T_2,s^* maps with and without FS (Fig. 7c, d). 

  Fig. 8 shows MWF images with and without FS. MWF image without FS shows a higher signal in the WM containing myelin components than with FS (Fig. 8a). As a result, it is possible that axon-derived signals were extracted by applying FS.