All MR examinations were obtained with a 32-channel phased array head coil on a 3T scanner (Magnetom Trio Tim,
Siemens,
Erlangen,
Germany).
High resolution anatomical T1-weighted images were acquired with a magnetization-prepared rapid acquisition gradient echo (MP-RAGE) sequence in the axial plane (176 slices,
1 mm isotropic resolution,
field-of-view: 256 mm x 192 mm; repetition time TR: 1910 ms; echo time TE: 3.37 ms; flip angle: 15°).
A total of 75 healthy young women,
with no history of neurological of psychiatric illness,
were enrolled.
The study was approved by the institutional review board,
and all subjects signed a written informed consent form.
Of our subjects, 38 women had a natural menstrual cycle (NC-group),
while 27 women were taking second and third generation monophasic hormonal contraceptives (HC-A group),
and 10 women were taking fourth generation monophasic hormonal contraceptives (HC-B group).
In the NC-group,
we obtained MR examinations of the brain during early follicular phase (days 2 to 7,
depending on mean cycle length).
In both HC-groups,
MR examinations were performed at two time-points during the cycle: (1) during the last days of the inactive pill phase (days 4 to 7 of the pill-free week); and (2) approximately 2 weeks in the active pill phase (days 13 to 16 of the pill-intake).
For voxel-based-morphometry (VBM) preprocessing we used the SPM8 software [15].
Initial approximate manual alignment was performed by adapting the headers from the NifTi-files,
using the view-tool.
The New Segment tool was used to segment images into gray matter (GM),
white matter (WM) and cerebrospinal fluid (CSF).
This feature uses unified tissue segmentation,
warping of tissue probability maps to match the image [16].
Intensities are modeled by a Gaussian Mixture Model.
Resulting images are bias-corrected for the smoothly varying intensity inhomogeneity caused by magnetic field imperfections,
using a linear combination of low frequency Discrete Cosine Transformation basis functions as the intensity inhomogeneity field model,
and are registered into standard space.
Next,
Diffeomorphic anatomical registration through exponentiated lie algebra (DARTEL) was used to account for more detailed shape variability in the population [17]. Images were spatially normalized to standard Montreal Neurological Institute (MNI) space,
followed by a smoothing procedure with a Gaussian Kernel of 6 mm.
Inter-subject alignment is of capital importance in VBM.
For every separate phase we calculated the average GM image and assessed quality of the registration algorithm by subtraction of each individual GM image from this average.
Standard deviations of subtraction image histograms were calculated,
and outliers were manually checked.
The inter-subject alignment consistently underperformed in one subject of the HC-A group,
and in one subject of the NC group,
for all phases.
In order to avoid false positive results,
these two individuals were discarded.
Any voxel in the GM images with value <0.2 was excluded from statistical analysis.
Age was inserted in the analysis as a covariate of no interest,
and T-tests or paired T-tests were used appropriately with a threshold of p < 0.001 (uncorrected).
We chose to exclude inter-subject variability to a maximum extent in our longitudinal study,
so we only included subjects which completed all examinations for further analysis.
In the HC groups,
coincidental abnormalities in the brain on the T1-weighted images were detected by a senior radiologist in 2 subjects.
These were consequently excluded from further analysis,
together with another subject that was accidentally scanned twice during the inactive pill-phase.
After data quality control,
a total of 105 scans remain; 37 scans in the NC group,
48 (2x24) in the HC-A group,
and 20 (2x10) in the HC-B group.
The mean age,
with standard deviation,
of these groups respectively were: 24.3 ± 3.9,
22.8 ± 2.5 and 22.3 ± 2.3 years.
The mean age of the groups showed no significant differences using a One way ANOVA (p-value = 0.138; F-value = 2.049).