Imaging was performed on 4 healthy participants (4 men, age 37±6 years) and 15 MS cases (10 men, 5 women, age 43±13 years), all of whom gave written, informed consent to participate in an Institutional Review Board-approved protocol. 3D-EPI was acquired using a 3T MRI system (Philips Healthcare, Best, The Netherlands) with the manufacturer's eight-channel receive-only head coil and body transmit coil. An automated whole-brain first-order shimming procedure provided by the manufacturer was performed before acquisition. The 3D-EPI acquisition used 0.55×0.55×0.55 mm3 voxels (0.17 mm3 voxel volume) covering the whole brain (336 slices). This acquisition relied on the manufacturer’s segmented EPI readout using a Cartesian interleaved k-space trajectory and an EPI factor of 15 lines per excitation. Acquisition was three-dimensional (3D) and performed in the sagittal plane with a parallel imaging (Sensitivity Encoding for Fast MRI, also called SENSE) factor of 2 in both left-right (i.e. slice-encoding) and anterior-posterior (i.e. phase-encoding) directions. A fat suppression prepulse removed the artifacts resulting from shifted fat signal (water–fat shift of 9.6 mm here).This was achieved using a fat-selective binomial spatial and spectral saturation pulse from the manufacturer ("Proset 1331"). Other parameters were: field of view = 220 (anterior-posterior) × 220 (foot-head) × 185 (left-right) mm3, flip angle = 10 °, repetition time = 54 ms, echo time = 29 ms, number of slices = 336, bandwidth in the readout direction = 490 Hz/pixel, bandwidth in the phase-encoding direction = 27 Hz/pixel. For improved signal-to-noise ratio (SNR), the sequence was acquired twice (number of excitations = 2), giving a total scan time under 4 min.
To test the effect of GBCA on venous contrast, the 3D-EPI acquisition was repeated for all 15 MS cases at three different times during the scanning: before, during and, approximately 15 min after GBCA injection. The GBCA was a single dose of gadopentetate dimeglumine (Magnevist; Bayer Healthcare, Leverkusen, Germany) or gadobutrol (Gadavist; Bayer Healthcare, Leverkusen, Germany) injected intravenously over 60 seconds. The infusion of GBCA was started simultaneously with the second 3D-EPI acquisition using a power injector (MEDRAD, Warrendale, PA).
Magnitude and phase images were collected as DICOM data produced by the MRI scanner. All analyses done on the magnitude images were performed using Matlab (The MathWorks, Natick, MA) and MIPAV (Medical Image Processing, Analysis & Visualization, NIH). For the phase images, an automatic analytical phase unwrapping was employed as described in (21 (link)). Following phase unwrapping, large background gradients were removed by performing Gaussian filtering with a filter size of 32 pixels and full width at half maximum of 8 pixels (22 ).
To measure the SNR of the magnitude 3D-EPI images, two consecutive acquisitions were performed on the 4 healthy participants. Using the dual-acquisition subtraction method (23 (link)) , the SNR was calculated as:
where SI1 is the mean intensity in the region-of-interest (ROI) on the first magnitude image, and SD1-2 is the standard deviation in the ROI on the subtraction magnitude image. A transverse slice passing though the genu and splenium of the corpus callosum of the healthy subjects was selected and, a 160×160 pixel ROI centered inside the brain parenchyma was used for calculating the SNR.
To measure contrast-to-noise ratio (CNR) between two different types of brain tissue, the mean intensities of the two tissues of interest (SI1 for tissue 1 and SI2 for tissue 2) were used as input in the following equation:
For the CNRlesion-WM, ROIs were drawn manually for the lesions detected on a transverse slice passing though the genu and splenium of the corpus callosum. A nearby ROI with normal appearing white matter was also drawn on this slice. For CNRGP-WM and CNRDN-WM, ROIs were drawn on a sagittal slice where both the globus pallidus (GP) and the dentate nucleus (DN) were visible. A region of normal appearing deep white matter was also drawn on this slice. For the CNRvein-WM, a minimum intensity projection (mIP) was first performed across 15 slices (i.e. 7.5 mm) along the head-foot direction at the level of the corpus callosum for the before, during- and after injection acquisitions. A mIP image of the pre-injection acquisition revealing the deep medullary veins of the white matter (WM) was selected, onto which a ROI primarily containing small veins was drawn. A nearby ROI without any apparent blood vessels was also drawn on the same mIP image. To quantify the effect of GBCA on the conspicuity of small parenchymal veins, the same ROIs were applied to them IP images of the during and after injection acquisitions. All the CNR calculations were performed across the 15 MS cases. Mean values and standard deviation values across the 15 cases are mentioned in the Results section.
A count of the lesions with and without a central vein was also performed for the 15 MS cases using the same transverse slice as the one previously selected for CNRlesion-WM calculations. Finally, lesions with a hypointense rim on magnitude image were counted when examining all the brain slices of the 15 MS patients.
To test the effect of GBCA on venous contrast, the 3D-EPI acquisition was repeated for all 15 MS cases at three different times during the scanning: before, during and, approximately 15 min after GBCA injection. The GBCA was a single dose of gadopentetate dimeglumine (Magnevist; Bayer Healthcare, Leverkusen, Germany) or gadobutrol (Gadavist; Bayer Healthcare, Leverkusen, Germany) injected intravenously over 60 seconds. The infusion of GBCA was started simultaneously with the second 3D-EPI acquisition using a power injector (MEDRAD, Warrendale, PA).
Magnitude and phase images were collected as DICOM data produced by the MRI scanner. All analyses done on the magnitude images were performed using Matlab (The MathWorks, Natick, MA) and MIPAV (Medical Image Processing, Analysis & Visualization, NIH). For the phase images, an automatic analytical phase unwrapping was employed as described in (21 (link)). Following phase unwrapping, large background gradients were removed by performing Gaussian filtering with a filter size of 32 pixels and full width at half maximum of 8 pixels (22 ).
To measure the SNR of the magnitude 3D-EPI images, two consecutive acquisitions were performed on the 4 healthy participants. Using the dual-acquisition subtraction method (23 (link)) , the SNR was calculated as:
where SI1 is the mean intensity in the region-of-interest (ROI) on the first magnitude image, and SD1-2 is the standard deviation in the ROI on the subtraction magnitude image. A transverse slice passing though the genu and splenium of the corpus callosum of the healthy subjects was selected and, a 160×160 pixel ROI centered inside the brain parenchyma was used for calculating the SNR.
To measure contrast-to-noise ratio (CNR) between two different types of brain tissue, the mean intensities of the two tissues of interest (SI1 for tissue 1 and SI2 for tissue 2) were used as input in the following equation:
For the CNRlesion-WM, ROIs were drawn manually for the lesions detected on a transverse slice passing though the genu and splenium of the corpus callosum. A nearby ROI with normal appearing white matter was also drawn on this slice. For CNRGP-WM and CNRDN-WM, ROIs were drawn on a sagittal slice where both the globus pallidus (GP) and the dentate nucleus (DN) were visible. A region of normal appearing deep white matter was also drawn on this slice. For the CNRvein-WM, a minimum intensity projection (mIP) was first performed across 15 slices (i.e. 7.5 mm) along the head-foot direction at the level of the corpus callosum for the before, during- and after injection acquisitions. A mIP image of the pre-injection acquisition revealing the deep medullary veins of the white matter (WM) was selected, onto which a ROI primarily containing small veins was drawn. A nearby ROI without any apparent blood vessels was also drawn on the same mIP image. To quantify the effect of GBCA on the conspicuity of small parenchymal veins, the same ROIs were applied to them IP images of the during and after injection acquisitions. All the CNR calculations were performed across the 15 MS cases. Mean values and standard deviation values across the 15 cases are mentioned in the Results section.
A count of the lesions with and without a central vein was also performed for the 15 MS cases using the same transverse slice as the one previously selected for CNRlesion-WM calculations. Finally, lesions with a hypointense rim on magnitude image were counted when examining all the brain slices of the 15 MS patients.
