Seventeen migraine patients (14 females and 3 males, mean age ± standard deviation [SD]: 42.9 ± 10.6 years, age range: 20–65 years) with previously discovered brain WMLs and 15 patients with relapsing–remitting multiple sclerosis (11 females and 4 males, mean age ± SD: 37.5 ± 8.8 years, age range: 23–51 years) were prospectively enrolled in this study between 2010 and 2017 at the Department of Neurology, Medical School, University of Pécs, Hungary. The most important radiological inclusion criterion was the large lesion size (≥ 7 mm, largest diameter) with normal-appearing white matter (NAWM) on the contralateral side in both groups, and a good physical status of MS patients expanded disability status scale (EDSS) score ≤ 3.

Among the migraine patients, 10 met the International Headache Society classification criteria for migraine without aura and seven for migraine with aura. Migraineurs had only supratentorial WMLs on their T2-weighted and FLAIR MRIs without hypointensity on their T1-weighted MRIs. All patients were only on abortive migraine treatment (such as nonsteroidal anti-inflammatory drugs or triptans) with no chronic prophylactic therapy at the time of the MRI study. The MRI studies were taken during a headache-free period.

All MS patients showed supra-and infratentorial WMLs and were diagnosed with relapsing–remitting MS according to the 2005 modified McDonald criteria (EDSS score ± SD: 1.6 ± 1.53; EDSS range: 0–3), and the diagnosis did not change when we revised these patients with the 2017 McDonald criteria (Thompson et al., 2018). WMLs in MS patients showed no hypointensity on T1-weighted images like the migraineurs, and no contrast enhancement was seen, referring to active demyelination. All patients were on chronic immunomodulatory therapy at the time of the MRI study, and the measurement was retaken in the remission phase.

None of the migraine or MS patients suffered from comorbidities that may cause WMLs, e.g., hypertension, diabetes, thyroid gland disease, other cerebrovascular risk factors, systemic autoimmune diseases. Migraine patients with other types of headaches were also excluded from the study. None of the MS patients had a history of migraine headaches.

As a control group, 17 healthy subjects age-and sex-matched to the migraine group (14 females and 3 males, mean age ± SD: 42.8 ± 10.5 years, age range: 20–65 years) without headache and with a normal MRI were also prospectively enrolled in the study. Patients lacked any comorbidities. Studies were performed in accordance with the approval of the Regional Research Ethics Committee of the Clinical Center, Pécs, and written informed consent was obtained from all study participants.

MRI was performed on a 3.0-Tesla clinical MRI scanner (Magnetom TIM Trio, Siemens Medical Solutions, Erlangen, Germany), with a field gradient strength of 40 mT/m and a 12-channel phased array head coil. The following MRI sequences were acquired: T1-and T2-weighted and 3-dimensional (3D) FLAIR images, DWI, PWI, ¹H-MRS, and T1 and T2 relaxation time measurements.

WMLs were defined as hyperintensities on T2-weighted and FLAIR images, without hypointensity on T1-weighted scans, measuring at least 3 mm or larger. Only one WML was investigated in each patient. The investigated WMLs’ largest diameters ranged between 7 and 21 mm and appeared in at least 3 consecutive axial slices on 3D FLAIR images (range between 3 and 10 slices). All investigated WMLs were in the supratentorial deep white matter (Figure 2).

Sagittal T1-weighted images were obtained using a fast low angle shot (FLASH) 2-dimensional (2D) sequence: TR/TE = 300/2.46 milliseconds, flip angle = 88°, 27 slices, slice thickness = 4 mm, 30% interslice gap, FOV = 220 × 220 mm², matrix size = 256 × 320, receiver bandwidth = 330 Hz/pixel.

For T2-weighted images, a turbo spin echo sequence was used: TR/TE = 6000/93 milliseconds, 30 slices, slice thickness = 4 mm, distance factor 20% (0.8 mm gap), FOV = 193 × 220 mm², 280 × 320-pixel matrix, bandwidth = 220 Hz/pixel, number of echo trains = 18. A turbo spin echo sequence was also used for the 3D FLAIR images: TR/TI/TE = 15,710/2750.8/105 milliseconds, 100 slices, slice thickness = 1.5 mm, distance factor 0% (no gap), interleaved slice readout with 2 concatenations, FOV = 220 × 220 mm², 192 × 192-pixel matrix, bandwidth = 400 Hz/pixel, number of echo trains = 14.

Diffusion was determined with a trace-weighted single-shot echo planar 2D imaging sequence: TR/TE = 4000/119 milliseconds, number of slices 20, slice thickness = 3.5 mm, distance factor 0% (no gap), FOV = 188 × 250 mm², 144 × 192 mm² pixel matrix, FOV phase 75%, number of acquisitions 4, b values 0, 500, 1,000 s/mm².

Perfusion images were acquired with a single-shot echo planar 2D sequence: TR/TE = 1400/33 milliseconds, flip angle = 68°, 20 slices, slice thickness = 3 mm, distance factor 33%, FOV = 210 × 210 mm², 176 × 176-pixel matrix, bandwidth = 1,150 Hz/pixel, fat saturation switched on. One hundred volumes were consecutively acquired, and contrast agents were administered after acquiring the 20th volume. A Medrad power injector was used for contrast agent and saline administration. A 0.1 mL/body weight amount of gadopentetate dimeglumine (Magnevist, Bayer Schering Pharma AG, Berlin, Germany) was given at a 5 mL/s flow rate, and a 30-ml saline flush was also used for washing at a 5 ml/s flow rate.

Spectroscopy was performed before contrast agent administration to avoid any confounding effects on T1 and T2 relaxations. Before single-voxel ¹H-MRS acquisition, a voxel of 12 × 12 × 12 mm³ was positioned on a preselected WML. Two voxels were placed in every migraine and MS patient: one in the selected WML showing the radiological characteristics of the corresponding disease and one in the contralateral, homotopic, NAWM area without MRI signal abnormality. Two voxels were defined in each healthy subject according to the locations of the voxels of the age-matched migraine patient. Voxels were placed using T2 and 3D FLAIR images to position each voxel. After localized manual shimming and water suppression adjustment, fully relaxed short-echo time proton magnetic resonance spectra (point resolved spectroscopy sequence [PRESS], TR/TE = 6000/30 milliseconds, 128 accumulations, bandwidth = 1,200 Hz, vector size 1,024 points) were acquired. Water suppression was accomplished with a chemical shift-selective sequence pulse. At the end of the ¹H-MRS acquisition, a reference water signal for the calibration of metabolite concentration was also acquired by turning off the water suppression. After acquiring the metabolite spectra, the water signal T1 and T2 parameters were also determined. T1 was measured using the saturation-recovery method. Six data points were collected, and only the TR was changed between each data point acquisition using the PRESS sequence: TR = 490, 900, 1,400, 2000, 3,000, and 4,000 milliseconds; TE = 30 milliseconds, 1 accumulation, water signal suppression turned off, bandwidth = 2,500 Hz, vector size = 1,024, 4 preparation scans. T2 was obtained by measuring six data points with parameters differing only in echo times using the PRESS sequence: TR = 3,000 milliseconds, TE = 30, 60, 90, 120, 180, and 240 ms, one accumulation, water signal suppression turned off, bandwidth: 2500 Hz, vector size = 1,024, four preparation scans. The total experimental protocol lasted for 45.5 min in the following order: (1) ~28 min for quantitative proton spectroscopy (with manual adjustments); (2) ~2.5 min for T1 measurement; (3) ~3 min for T2 measurement; (4) ~6.5 min for apparent diffusion coefficient (ADC) quantification; (5) ~1 min for T2-weighted images; (6) ~3.5 min for 3D FLAIR; (7) ~2 min for perfusion.

The T1 relaxation time for each voxel was calculated from the acquired water signals with different repetition times by applying a standard exponential fit:

where M is the actual signal intensity, M0 is the signal intensity at thermal equilibrium, and TR is the repetition time.

The T2 relaxation time for each voxel was calculated from the acquired water signals with different echo times, assuming a standard exponential signal decay:

where M is the actual signal intensity, M0 is the signal intensity at thermal equilibrium, and TE is the echo time. Curve fittings were carried out on a Siemens Leonardo workstation using Siemens spectroscopy software. Only Fourier transformation and phase correction on the measured signals were applied; no filters or any other corrections were used. The integral of the fitted water signal was used for the T1 and T2 fittings.

To calculate the ADC values, freehand regions of interest (ROIs) were drawn on b0 images on the preselected WMLs. ROIs covered the WMHs selectively. Just like for the spectroscopy, a control area was also measured in the contralateral, homotopic NAWM. Within each ROI, the mean intensities for the b0, b500, and b1000 images were monoexponentially fitted using the following equation:

where M is the measured signal intensity in the presence of diffusion sensitization, M0 is the signal intensity in the absence of diffusion sensitization, b is the b-value, and ADC is the ADC value. Curve fitting and data analysis were performed using Matlab software (The MathWorks, Inc., Natick, MA, USA) for T1, T2, and ADC fittings.

For metabolite quantitative analysis, spectroscopic raw data were postprocessed using the LCModel (Stephen Provencher Inc., Oakville, Ontario, Canada). The concentrations of NAA, glutamate + glutamine + GABA, (Glx), Cho, creatine + phosphocreatine (Cr), and mI were determined (Figure 3).

Perfusion analysis was performed on a Siemens Leonardo Workstation using Siemens Perfusion software (Siemens Medical Solutions, Erlangen, Germany). The relative cerebral blood flow (rCBF) and relative cerebral blood volume (rCBV) were calculated. The arterial input function was determined by an experienced radiologist. Freehand region of interest (ROI) analysis was performed. Lesions with a moderate T2 signal abnormality verified on T2 and FLAIR images could also be identified on the native, slightly T2-and T2*-weighted raw ep2d images measured for perfusion analysis as high-intensity areas. ROIs were drawn on these raw ep2d images; afterwards, these ROIs were copied and used at the same slice of the calculated perfusion maps. The lesions and control areas to be measured as ROIs were selected and verified on T2 and FLAIR images to have a control contralateral white matter area without signal abnormality.

To test normality, the Kolmogorov–Smirnov and Shapiro–Wilk tests were applied. Because the variables did not follow the normal distribution, non-parametric tests were used in subsequent analyses. The acquired MRI variables were compared between the lesional side and the contralateral NAWM of the subjects in each group separately using Mann–Whitney U-tests. The same variables for the lesional and then the contralateral sides were compared among the three subject groups using Kruskal-Wallis tests with post hoc analyses. Due to multiple comparisons, p values were Bonferroni corrected. Fisher’s exact test was used to examine the difference in the distribution of lobar locations of WMLs between the migraine and MS groups. The ability of intralesional mI to discriminate MS lesions from migraine lesions was assessed by fitting the receiver operating characteristic (ROC) curve and calculating sensitivity, specificity, and area under the ROC curve (AUC). All statistical tests were performed using SPSS Statistics 25.0 software (IBM Corp., Armonk, NY). Results were considered significant (with Bonferroni correction, where applicable) at p ≤ 0.05.

In the migraine group, decreased NAA and creatinine (p = 0.016 and p < 0.001, respectively), and elevated T2 relaxation time and ADC values (p ≤ 0.001) were detected in the WMLs compared to the contralateral NAWM (see data in Table 1).

Similarly, in the MS group, decreased NAA and elevated T2 relaxation time and ADC values (p < 0.001 in all cases) were found in the WMLs compared to the contralateral side. In addition, elevated T1 values were also observed in MS WMLs (p < 0.001), while Cr values did not differ significantly between the two sides (see details in Table 1).

No difference was observed between the two hemispheres of the control subjects in any MRI variables. Perfusion variables (rCBF and rCBV) did not show any difference between the intralesional and the contralateral sides in any groups.

Fisher’s exact test indicated that the distribution of lobar location of the large WMLs were not different between our two patient groups, regardless of whether the lobes were considered bilateral structures or whether the left and right hemispheric lobes were considered separately (p = 0.697 and p = 0.954, respectively; see distribution in Table 2). On the lesional side, significant differences were shown in NAA, Cr, mI, T1, T2 relaxation times, and ADC values among the 3 patient groups (p ≤ 0.001 in all cases). T1, T2 relaxation times, and ADC values followed a similar trend, with the lowest values observed in controls and the highest in MS patients (Table 3). In post hoc tests, migraine patients showed significantly higher NAA and lower mI and T1 values than MS patients (p = 0.033, p = 0.001, and p = 0.029, respectively, see details in Table 3). Moreover, migraineurs had decreased creatine and increased T2 relaxation time and ADC values compared to the control subjects (p ≤ 0.001 in all cases, Table 3). The MS group showed elevated mI (p = 0.035), T1, T2 relaxation time, and ADC (p ≤ 0.001), and reduced NAA (p < 0.001) values compared to the control group (Table 3). The ROC analysis for intralesional mI indicated that MS lesions can be differentiated from migraine lesions with a sensitivity of 86.7% and a specificity of 82.4% when using an optimal cutoff point of 5.224 (i.e., the cutoff point nearest to the upper left corner of the ROC space); AUC = 0.878.

In the contralateral NAWM, lower mI values were observed in migraine patients compared to MS patients, while choline values showed only a trend (p = 0.001 and p = 0.05, respectively, Table 3). MS patients showed higher mI values than control subjects in the NAWM (p = 0.018). No significant difference was found in any values of the normal-appearing side contralateral to the lesion between the migraine and control groups (Table 3). None of the PWI variables were different in any comparisons among the three groups.