G Model
ARTICLE IN PRESS
NSL 30526 1–5
Neuroscience Letters xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Neuroscience Letters journal homepage: www.elsevier.com/locate/neulet
Are there any gender differences in the hippocampus volume after head-size correction? A volumetric and voxel-based morphometric study
1
2
3
4 5 6 7 8 9 10
Q1
Gabor Perlaki a,b,∗ , Gergely Orsi a,b , Eniko Plozer c , Anna Altbacker c , Gergely Darnai c , Szilvia Anett Nagy b , Reka Horvath c , Arnold Toth c,d , Norbert Kovacs a,c , Peter Bogner b , Attila Schwarcz a,d , Jozsef Janszky a,c a
MTA-PTE Clinical Neuroscience MR Research Group, Ret u. 2, 7623 Pecs, Hungary Pecs Diagnostic Centre, Ret u. 2, 7623 Pecs, Hungary c Department of Neurology, University of Pecs, Ret u. 2, 7623 Pecs, Hungary d Department of Neurosurgery, University of Pecs, Ret u. 2, 7623 Pecs, Hungary b
11
12 26
a r t i c l e
i n f o
a b s t r a c t
13 14 15 16 17
Article history: Received 29 January 2014 Received in revised form 24 March 2014 Accepted 9 April 2014
18
25
Keywords: Head-size correction MRI VBM Volumetry Gender difference Hippocampus
27
1. Introduction
19 20 21 22 23 24
28 29 30 31 32 33 34 35 36 37
Previous findings on normal sexual dimorphism in hippocampal volume have not always been consistent. This study investigated gender differences in hippocampal volume using different head-size correction strategies. T1-weighted MR images were collected in 99 healthy, Caucasian, university students (66 female subjects; mean age: 23.1 ± 2.3, range: 19–31 years). Sexual dimorphism in hippocampus was investigated by automated MRI volumetry and voxel-based morphometry (VBM) using both general linear model (GLM) and proportion head-size correction strategies. Absolute hippocampal volumes were larger in men than women. After adjusting for head-size, the proportion method indicated larger hippocampi in women than men, while no gender differences were found using the GLM approach. Investigating absolute hippocampal volumes in 15 head-size matched pairs of males and females indicated no gender differences. We suggest that there is no sexual dimorphism in hippocampal size and the apparent gender differences found by the proportion method may have more to do with head-size than with sex. The GLM and proportion head-size correction strategies are not interchangeable and may yield different results. The importance of the present findings is mostly related to scientific reproducibility across MRI volumetry or VBM studies. © 2014 Published by Elsevier Ireland Ltd.
The hippocampus is a prime focus of MRI-based volumetric analysis in healthy subjects, cognitive disorders and neuropsychiatric conditions. Altered hippocampal volumes were reported in epilepsy, Alzheimer’s disease, schizophrenia and post-traumatic stress disorder [1,2]. Hippocampal volume measurements can assist the diagnosis of Alzheimer’s disease [3] and epilepsy [4]. Hippocampal volume changes were found to differ in some neuropsychiatric conditions as a function of gender [5,6]. Investigating gender difference in hippocampal volumes of normal subjects is essential to understand why male and female hippocampi differ in
∗ Corresponding author at: MTA-PTE Clinical Neuroscience MR Research Group, Ret u. 2, 7623 Pecs, Hungary. Tel.: +36 72535900; fax: +36 72 535911. E-mail address: [email protected] (G. Perlaki).
their predisposition for volumetric changes in certain conditions and to distinguish disease patterns from normal ones. Previous findings on sexual dimorphism in hippocampal volume have not always been consistent. For example, larger hippocampal volumes in females [7,8] and no difference in hippocampal volumes between males and females [9–11] were also reported after correcting for head-size differences. Interestingly, the above studies reporting gender difference used the proportion method [7,8], while the studies describing no gender difference used the general linear model (GLM) approach [9,10] or an extension of that [11] to adjust for head-size. Based on this observation, we hypothesized that part of the discrepancy in results may be attributed to different types of head-size correction used. Volumetric measures tend to be associated with normal variation in head-size; people with larger heads typically have larger brain structures [12]. To account for this irrelevant variance, volumetric measures are often “normalized” with respect to intracranial
http://dx.doi.org/10.1016/j.neulet.2014.04.013 0304-3940/© 2014 Published by Elsevier Ireland Ltd.
Please cite this article in press as: G. Perlaki, et al., Are there any gender differences in the hippocampus volume after head-size correction? A volumetric and voxel-based morphometric study, Neurosci. Lett. (2014), http://dx.doi.org/10.1016/j.neulet.2014.04.013
38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54
G Model NSL 30526 1–5
G. Perlaki et al. / Neuroscience Letters xxx (2014) xxx–xxx
2 55 56 57 58 59
60 61 62 63 64
ARTICLE IN PRESS
volume (ICV) [13]. The strong effect of ICV on voxel-wise volume information was also demonstrated, suggesting that head-size correction should also be considered in voxel-based morphometry (VBM) [10,14]. GLM and proportion approaches are two common methods to correct for ICV [13,15]. (i) The GLM approach corrects for head-size in a linear regression model, including ICV as a covariate of no interest. (ii) The proportion method normalizes for head-size directly by dividing the volumetric measure by the ICV (generating relative measures).
77
Although there are a few critical evaluations about the advantages/disadvantages of the GLM and proportion methods, there is no consensus regarding the most accurate way to adjust for headsize in statistical analyses [12,13,15–17]. The aim of the present study was twofold: (1) to re-examine the gender effect on hippocampal size using GLM and proportion head-size correction strategies. (2) To compare GLM and proportion approaches in hippocampal volumetry and VBM, using the same anatomical images from a set of normal control subjects. Our working hypothesis is that these two normalization methods may yield different results when comparing groups unmatched for head-size or when testing for inter-individual differences in a population with considerable variation in head-size.
78
2. Materials and methods
79
2.1. Subjects
65 66 67 68 69 70 71 72 73 74 75 76
87
Ninety-nine healthy, Caucasian, graduate or postgraduate university students (66 women and 33 men; mean age: 22.9 ± 2.2 and 23.5 ± 2.4, range: 19–31 and 20–30 years respectively) were included. Males and females did not differ in the distribution of education or age. All subjects got detailed information about the investigation and gave written informed consent prior to the examination. The study was approved by the local ethical committee and performed in accordance with the Declaration of Helsinki.
88
2.2. Magnetic resonance imaging
80 81 82 83 84 85 86
95
All measurements were performed on a 3T MRI scanner (MAGNETOM Trio, Siemens AG, Erlangen, Germany) with a 12-channel head coil. T1-weighted three-dimensional MPRAGE images were acquired with the following parameters: TR/TI/TE = 2530/1100/3.37 ms; flip angle = 7◦ ; 176 sagittal slices; slice thickness = 1 mm; FOV = 256 mm × 256 mm; matrix size = 256 × 256; receiver bandwidth = 200 Hz/pixel.
96
2.3. Data analysis
89 90 91 92 93 94
97 98 99
100 101 102 103 104 105
106 107 108 109
Sexual dimorphism in hippocampal volume was investigated using two different concepts (MR volumetry and VBM) and two head-size correction strategies (GLM and proportion). 2.3.1. Intracranial volume measure Since the ICV cannot be reliably measured on T1-weighted MR images [18], the reciprocal of volumetric scaling factor derived automatically from SIENAX, henceforth referred to as ICVsienax , was used as a surrogate for ICV [19]. The validity of ICVsienax as a predictor of ICV was demonstrated in an earlier study [18]. 2.3.2. MR volumetry Hippocampal volumes were estimated from T1-weighted MPRAGE images using the Freesurfer 4.5 image analysis suite (https://surfer.nmr.mgh.harvard.edu/). The technical details
were described previously [20]. Quality control was performed throughout the automatic processing stream. Error correction was performed when necessary, based on the recommended reconstruction workflow (http://surfer.nmr. mgh.harvard.edu/fswiki/RecommendedReconstruction/). Statistical analyses were performed using SPSS 20. Left and right hippocampal volumes were handled separately. (i) Absolute hippocampal volumes were compared between males and females using unpaired t-test. (ii) To correct for head-size through GLM, a multiple linear regression model was constructed with absolute hippocampal volume as dependent, while gender and ICVsienax as independent variables. (iii) To correct for head-size using the proportion method, the hippocampal volumes were divided by their corresponding ICVsienax . The relative hippocampal volumes were then compared between males and females using unpaired t-test. (iv) A subset of subjects with closely matched head-size was also analyzed. Males and females were paired based on an automated and objective method described earlier [21]. Briefly, the absolute differences in ICVsienax were computed for every possible pairings of males and females and then one male and one female with the minimum absolute difference in ICVsienax were paired and removed from the cohort. This procedure was repeated until there were no remaining pairs of males and females with relative difference in ICVsienax less than 5%. The above method resulted in 15 matched pairs of males and females. This group was then tested for gender differences in absolute hippocampal volumes using non-parametric Wilcoxon signed-rank test. (v) Both women and men were divided into two subgroups based on their median ICVsienax . The hippocampal volumes of women (or men) with smaller ICV were compared to women (men) with larger ICV by using both GLM and proportion head-size correction strategies. When using the GLM method a multiple linear regression model was constructed with absolute hippocampal volume as dependent, while the subgroup membership and ICVsienax as independent variables. In case of the proportion method the relative hippocampal volumes were compared between the subgroups by Mann–Whitney U test. In order to control for the potential confounding effects of age, the above comparisons were repeated using multiple linear regression with age as an independent variable. Results were considered significant at P ≤ 0.05 for all statistical tests. 2.3.3. Voxel-based morphometry Voxel-based morphometry was performed using FSL-VBM (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/fslvbm/). First, all structural images were brain-extracted using BET [22]. Next, tissue-type segmentation was carried out using FAST [23] and the resulting gray matter partial volume images were non-linearly registered to MNI152 standard space [24]. The registered images of 33 men and 33 randomly chosen women were averaged together with their respective mirror images to create a left–right symmetric studyspecific gray matter template (an equal number of subjects from both groups was used to avoid bias). All the 99 native gray matter images were non-linearly registered to this study-specific template and “modulated” to correct for volume changes due to spatial transformation. Two different types of modulations were performed. First, voxel values in gray matter images were corrected for volume changes due to both affine and nonlinear components of the registration (full modulation). Thus, the total amount of gray matter remains
Please cite this article in press as: G. Perlaki, et al., Are there any gender differences in the hippocampus volume after head-size correction? A volumetric and voxel-based morphometric study, Neurosci. Lett. (2014), http://dx.doi.org/10.1016/j.neulet.2014.04.013
110 111 112 113 114 115 116
117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149
150 151 152 153 154
155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172
G Model NSL 30526 1–5
ARTICLE IN PRESS G. Perlaki et al. / Neuroscience Letters xxx (2014) xxx–xxx
3
Fig. 1. Boxplots illustrating the significantly larger left and right relative hippocampal volumes in women compared to men (a and b respectively). Relative hippocampal volume was calculated as the absolute hippocampal volume (mm3 ) dividing by the corresponding ICVsienax . Whiskers are set at minimum and maximum, the horizontal line marks the median, whereas box indicates the interquartile range (25–75%).
173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190
the same as in the original images, so individual differences due to head-size are also preserved. Second, voxel values in gray matter images were corrected only for the nonlinear effects of the spatial transformation (nl modulation). Excluding the affine component from the modulation is an approximate of multiplying the full modulated data with the relative volumetric scaling factor due to affine registration.1 Considering that the reciprocal of this scaling factor is a valid measure of ICV [25], the nl modulated data are inherently adjusted for head-size according to the proportion method. The modulated gray matter images were then smoothed with an isotropic Gaussian kernel (sigma = 3 mm). Finally, to test for gender differences, voxelwise GLM limited to left and right hippocampal regions was applied using permutation-based non-parametric testing (10,000 permutations). The hippocampal regions were defined using the maximum probability Harvard-Oxford Subcortical Structural Atlas thresholded at 25% and masked by the gray matter mask automatically generated as part of the FSL-VBM protocol.
between the head-size based subgroups without considering ICVsienax in the statistical model. In order to control for the potential confounding effects of age, the above analyses were repeated including age as a covariate. Results were considered significant for P ≤ 0.05 (after initial cluster-forming threshold at t = 2.0), fully corrected for multiple comparisons. 3. Results Controlling for the potential confounding effects of age did not change the pattern of significance and age was not proved to be a significant predictor in any of the analyses performed. Final results are presented without controlling for age. There was no interaction between head-size and gender (i.e. ICVsienax *gender) in any of the analyses performed. 3.1. MR volumetry
191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210
(i) Absolute hippocampal size was compared between males and females using the full modulated data without considering ICVsienax in the statistical model. (ii) Head-size was controlled through GLM using the full modulated data and incorporating the gender and ICVsienax as independent variables in the statistical model. (iii) To correct for head-size using the proportion method, the nl modulated data were compared between males and females without considering ICVsienax in the statistical model. (iv) The 15 head-size matched pairs of males and females were tested for gender differences in absolute hippocampal size using the full modulated data without considering ICVsienax in the statistical model. (v) The hippocampal volumes of women (or men) with smaller ICV were compared to women (men) with larger ICV by using both GLM and proportion head-size correction strategies. Headsize was controlled through GLM using the full modulated data and incorporating the subgroup membership and ICVsienax as independent variables in the statistical model. In case of the proportion method, the nl modulated data were compared
1 If there were no skews introduced by the affine transformation, the nl modulated data would be exactly the same as multiplying the full modulated data with the relative volumetric scaling factor due to affine registration. In the presence of skews, the two dataset may slightly differ in certain voxels.
(i) Both left and right absolute hippocampal volumes were larger in men than women (P < 0.001). (ii) GLM analysis (including absolute hippocampal volume as dependent variable and gender and ICVsienax as independent variables) indicated positive effects of ICVsienax on both left and right hippocampal volumes (P < 0.001, see Supplementary Figure 1), while no significant gender differences were observed (see Supplementary Figure 2). (iii) Proportion method indicated that both left and right relative hippocampal volumes (absolute hippocampal volume divided by the ICVsienax ) were larger in women than men (unpaired ttest, P < 0.001; see Fig. 1). (iv) In the cohort of 30 young adults closely matched for head-size, there were no significant effects of gender on either left or right absolute hippocampal volumes. (v) The proportion method indicated significantly increased hippocampal size bilaterally in men with smaller ICV (n = 16) compared to men with larger ICV (n = 17), but not significant differences between women with smaller ICV and women with larger ICV. The GLM indicated no significant differences between subgroups in any sexes. Supplementary Figures 1 and 2 related to this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.neulet.2014.04.013.
Please cite this article in press as: G. Perlaki, et al., Are there any gender differences in the hippocampus volume after head-size correction? A volumetric and voxel-based morphometric study, Neurosci. Lett. (2014), http://dx.doi.org/10.1016/j.neulet.2014.04.013
211 212
213 214 215 216 217
218
219 220 221 222 223 224
225
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246
247 248 249
G Model NSL 30526 1–5
ARTICLE IN PRESS G. Perlaki et al. / Neuroscience Letters xxx (2014) xxx–xxx
4
Q3 Fig. 2. Voxelwise analysis of hippocampi using the nl modulated data without considering ICVsienax as a confounding variable. Red-yellow shows voxels demonstrating significantly increased relative amount of gray matter in women compared to men. Color bar represents t-values. The background image is the MNI152 standard space T1 template. X-, Y- and Z-values indicate the MNI slice coordinates in millimeter. Images are shown in radiological convention (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).
250
251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270
3.2. Voxel-based morphometry (i) Using the full modulated data without considering ICVsienax as a confounding variable indicated that absolute hippocampal size was larger in men than women. (ii) No statistically significant gender difference was found using the full modulated data and controlling for head-size through GLM. The voxelwise analysis of hippocampi indicated significant positive effects of ICVsienax bilaterally (see Supplementary Figure 3). (iii) The nl modulated data (inherently corrected for head-size according to the proportion method) indicated significantly larger hippocampi in women than men (see Fig. 2). (iv) In case of the matched pairs of male and female subjects, no significant gender differences were found in absolute hippocampal size using the full modulated data without considering ICVsienax as a confounding variable. (v) The proportion method indicated significantly increased left hippocampal size in women (or men) with smaller ICV compared to women (men) with larger ICV. The GLM method indicated no significant differences between subgroups in any sexes.
273
Supplementary Figure 3 related to this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.neulet.2014.04.013.
274
4. Discussion
271 272
275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294
In many studies the volumetric measures of hippocampus are controlled for ICV, in order to allow for comparisons between subjects who differ in head-size. GLM and proportion methods are the two most common approaches for making this adjustment. Although both are used pervasively in neuroimaging research, there is little direct comparison between them. In practice, they are used rather interchangeably, not really considering the difference between them. In the present study, sexual dimorphism in hippocampus was investigated by using both general linear model (GLM) and proportion head-size correction strategies. Both MRI volumetry and VBM indicated significant positive effects of ICVsienax on absolute hippocampal size bilaterally after the effects of gender were removed (Supplementary Figures 1 and 3), which is consistent with previous data suggesting that increasing head-size is associated with larger volumes [14]. As expected based on the above, absolute hippocampal size was larger in men than women. Both MRI volumetry and VBM indicated significantly larger hippocampal volume in women than men, when the effect of headsize was “normalized” by the proportion method (Figs. 1 and 2),
while this apparent gender effect was absent, when head-size was controlled through GLM (Supplementary Figure 2). This is good evidence that the two methods are not interchangeable and may yield different results. The discrepancy can be resolved by recognizing the difference between proportion and GLM approaches and the different questions they address. Our results are consistent with earlier studies using the proportion method [7,8] or GLM approach [9,10]. While these studies may provide support of our findings, they are not directly comparable to our investigation, due to methodological differences. In case of GLM the phrase “correcting for” refers to statistical control. Statistical control of head-size serves to remove the portion of the variance in hippocampal volume due to head-size. Unlike GLM, the proportion method calculates ratios (relative sizes), which was shown to be ineffective for statistical control in most cases [26]. Calculation of proportions with the intention to control for head-size is based on the assumption that hippocampal volume varies with head-size proportionally, so their relationship is linear with zero y-intercept for the regression line. The zero y-intercept assumption is seldom satisfied for real data. If this assumption fails, simple ratio is expected to be associated with its denominator (i.e. head-size), thus not resulting in a “head-size-free” measure [12]. To address this issue, we checked whether the volumetric measures normalized according to the proportion method correlate with the ICVsienax . Using Spearman’s correlation, we found that both the left and right relative hippocampal volumes as well as the mean hippocampal gray matter values obtained by averaging the smoothed nl modulated data in left and right hippocampal regions were inversely correlated with ICVsienax (P < 0.002), which remained significant (P < 0.05) after statistically controlling for gender through multiple linear regression. These correlations demonstrate the inability of proportion method to perform statistical control. Based on our present results, we suggest that before using either GLM or proportion method, the investigators should clarify whether they are attempting to statistically control for head-size or to examine relative size with respect to head-size and choose accordingly. We need to emphasize this differentiation, because it is still often overlooked and considered secondary in scientific studies. For instance, in a recent VBM study the two concepts are mixed even within a single evaluation [27]. In the literature there is clear distinction between VBM analyses with and without a modulation step. The analysis of modulated data allows testing for regional differences in the amount (volume) of gray matter, whereas the unmodulated data reflect regional gray matter concentration [28]. However, there is no such clear differentiation between nl modulated and full modulated datasets. Modulation of non-linear warping only (nl modulation) have gained popularity in VBM analyses, suggesting that this type of modulation accounts for head-size and no further corrections are needed [29,30]. If the effect of head-size was completely removed by not modulating for volume changes due to the affine part of the registration, any significant relationship between nl modulated data and ICVsienax would not be found. However, our data indicate a negative correlation between nl modulated data in the hippocampus and ICVsienax as a remaining effect of head-size. The idea that we do not need to estimate ICV and correct for it in subsequent statistical analyses is undoubtedly attractive, but if we are interested in “head-size-free” hippocampal size, the full modulated data should be used including ICV measure as a confounding variable in the statistical model. Results from the subset of subjects closely matched for headsize support the findings of the GLM approach. Based on these results and the observation that smaller brains exhibit relatively larger hippocampi regardless of gender, we suggest that there is no sexual dimorphism in hippocampal size and the apparent gender differences found by the proportion method may be not
Please cite this article in press as: G. Perlaki, et al., Are there any gender differences in the hippocampus volume after head-size correction? A volumetric and voxel-based morphometric study, Neurosci. Lett. (2014), http://dx.doi.org/10.1016/j.neulet.2014.04.013
295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360
G Model NSL 30526 1–5
ARTICLE IN PRESS G. Perlaki et al. / Neuroscience Letters xxx (2014) xxx–xxx
361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394
395
396 397 398
399
gender specific but rather a consequence of the well-known sexual dimorphism in head-size. This impression is further strengthened by the increased hippocampal size of women (or men) with smaller ICV compared to women (men) with larger ICV when using the proportion method. Although, the difference was not significant in all of the examined comparisons (probably due to the reduced variability of hippocampal volumes and ICVsienax when sexes are examined separately), these results provide more straightforward and convincing evidence that the proportion method is unable to control for head-size even within very homogeneous samples. To our knowledge this is one of the very first studies exploring differences between proportion and GLM methods in practice using both MRI volumetry and VBM. Earlier studies comparing head-size “normalization” methods focused mainly on theoretical issues such as reliability [12,16] and how the introduction of systematic and random errors may affect the final outcome of normalized measures [13] or used pediatric samples [17]. None of these studies was concerned about the potential impact of head-size normalization on voxelwise techniques (e.g. VBM), where head-size normalization and modulation step are highly related issues. A potential limitation of our study is the unequal number of males and females. However, when the 33 men were compared to 33 randomly selected women the overall pattern of results remained unchanged, suggesting that our findings are not driven by unequal sample sizes. Conclusively, we suggest that there is no sexual dimorphism in hippocampal size and the apparent differences found by the proportion method may have more to do with head-size than with sex. The importance of the present findings about the fundamental differences between GLM and proportion approaches is mostly related to scientific reproducibility across MRI volumetry or VBM studies. Our results may have clinical importance, when adjusted hippocampal volumes are considered as diagnostic tools in epilepsy or Alzheimer’s disease. Conflict of interest We are not aware of any actual or potential conflicts of interest regarding this work. The authors alone are responsible for the content and writing of the paper. Acknowledgements
405
This work was supported by the Hungarian Research Fund (OTKA K105357; OTKA PD103964), PTE AOK-KA-2013/12, PTE AOK-KA No: 2011/34039/KA-OTKA/11-10, KTIA-NAP-13-1-20130001 and the grant SROP-4.2.2/A-11/1/KONV-2012-0017. AS and NK were supported by the Bolyai Scholarship. NK was also supported by the Romhanyi Scholarship and Magyary Scholarship.
406
References
Q2 400 401 402 403 404
407 408 409 410 411 412 413 414 415 416 417 418 419 420
[1] E. Geuze, E. Vermetten, J.D. Bremner, MR-based in vivo hippocampal volumetrics: 2. Findings in neuropsychiatric disorders, Mol. Psychiatry 10 (2005) 160–184. [2] M.E. Smith, Bilateral hippocampal volume reduction in adults with posttraumatic stress disorder: a meta-analysis of structural MRI studies, Hippocampus 15 (2005) 798–807. [3] K. Kantarci, Y. Xu, M.M. Shiung, P.C. O’Brien, R.H. Cha, G.E. Smith, R.J. Ivnik, B.F. Boeve, S.D. Edland, E. Kokmen, E.G. Tangalos, R.C. Petersen, C.R. Jack Jr., Comparative diagnostic utility of different MR modalities in mild cognitive impairment and Alzheimer’s disease, Dement. Geriatr. Cogn. Disord. 14 (2002) 198–207. [4] C.R. Jack Jr., F.W. Sharbrough, G.D. Cascino, K.A. Hirschorn, P.C. O’Brien, W.R. Marsh, Magnetic resonance image-based hippocampal volumetry: correlation with outcome after temporal lobectomy, Ann. Neurol. 31 (1992) 138–146. [5] R.S. Briellmann, S.F. Berkovic, G.D. Jackson, Men may be more vulnerable to seizure-associated brain damage, Neurology 55 (2000) 1479–1485.
5
[6] C. Exner, B. Nehrkorn, V. Martin, M. Huber, K. Shiratori, W. Rief, Sex-dependent hippocampal volume reductions in schizophrenia relate to episodic memory deficits, J. Neuropsychiatry Clin. Neurosci. 20 (2008) 227–230. [7] C.A. Szabo, J.L. Lancaster, J. Xiong, C. Cook, P. Fox, MR imaging volumetry of subcortical structures and cerebellar hemispheres in normal persons, AJNR Am. J. Neuroradiol. 24 (2003) 644–647. [8] P.A. Filipek, C. Richelme, D.N. Kennedy, V.S. Caviness Jr., The young adult human brain: an MRI-based morphometric analysis, Cereb. Cortex 4 (1994) 344–360. [9] C.R. Jack Jr., C.K. Twomey, A.R. Zinsmeister, F.W. Sharbrough, R.C. Petersen, G.D. Cascino, Anterior temporal lobes and hippocampal formations: normative volumetric measurements from MR images in young adults, Radiology 172 (1989) 549–554. [10] G.S. Pell, R.S. Briellmann, C.H. Chan, H. Pardoe, D.F. Abbott, G.D. Jackson, Selection of the control group for VBM analysis: influence of covariates, matching and sample size, Neuroimage 41 (2008) 1324–1335. [11] R.C. Gur, F. Gunning-Dixon, W.B. Bilker, R.E. Gur, Sex differences in temporolimbic and frontal brain volumes of healthy adults, Cereb. Cortex 12 (2002) 998–1003. [12] D.H. Mathalon, E.V. Sullivan, J.M. Rawles, A. Pfefferbaum, Correction for head size in brain-imaging measurements, Psychiatry Res. 50 (1993) 121–139. [13] M.P. Sanfilipo, R.H. Benedict, R. Zivadinov, R. Bakshi, Correction for intracranial volume in analysis of whole brain atrophy in multiple sclerosis: the proportion vs. residual method, Neuroimage 22 (2004) 1732–1743. [14] J. Barnes, G.R. Ridgway, J. Bartlett, S.M. Henley, M. Lehmann, N. Hobbs, M.J. Clarkson, D.G. MacManus, S. Ourselin, N.C. Fox, Head size, age and gender adjustment in MRI studies: a necessary nuisance? Neuroimage 53 (2010) 1244–1255. [15] L.M. O’Brien, D.A. Ziegler, C.K. Deutsch, D.N. Kennedy, J.M. Goldstein, L.J. Seidman, S. Hodge, N. Makris, V. Caviness, J.A. Frazier, M.R. Herbert, Adjustment for whole brain and cranial size in volumetric brain studies: a review of common adjustment factors and statistical methods, Harv Rev Psychiatry 14 (2006) 141–151. [16] S. Arndt, G. Cohen, R.J. Alliger, V.W. Swayze 2nd, N.C. Andreasen, Problems with ratio and proportion measures of imaged cerebral structures, Psychiatry Res. 40 (1991) 79–89. [17] L.M. O‘Brien, D.A. Ziegler, C.K. Deutsch, J.A. Frazier, M.R. Herbert, J.J. Locascio, Statistical adjustments for brain size in volumetric neuroimaging studies: some practical implications in methods, Psychiatry Res. 193 (2011) 113–122. [18] G. Fein, V. Di Sclafani, C. Taylor, K. Moon, J. Barakos, H. Tran, B. Landman, R. Shumway, Controlling for premorbid brain size in imaging studies: T1-derived cranium scaling factor vs. T2-derived intracranial vault volume, Psychiatry Res. 131 (2004) 169–176. [19] S.M. Smith, Y. Zhang, M. Jenkinson, J. Chen, P.M. Matthews, A. Federico, N. De Stefano, Accurate, robust, and automated longitudinal and cross-sectional brain change analysis, Neuroimage 17 (2002) 479–489. [20] B. Fischl, D.H. Salat, E. Busa, M. Albert, M. Dieterich, C. Haselgrove, A. van der Kouwe, R. Killiany, D. Kennedy, S. Klaveness, A. Montillo, N. Makris, B. Rosen, A.M. Dale, Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain, Neuron 33 (2002) 341–355. [21] B.A. Ardekani, K. Figarsky, J.J. Sidtis, Sexual dimorphism in the human corpus callosum: an MRI study using the OASIS brain database, Cereb. Cortex 23 (2013) 2514–2520. [22] S.M. Smith, Fast robust automated brain extraction, Hum. Brain Mapp. 17 (2002) 143–155. [23] Y. Zhang, M. Brady, S. Smith, Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm, IEEE Trans. Med. Imaging 20 (2001) 45–57. [24] J.L. Andersson, M. Jenkinson, S. Smith, Non-Linear Registration, aka Spatial Normalisation FMRIB Technical Report TR07JA2, FMRIB Analysis Group of the University of Oxford, 2007. [25] R.L. Buckner, D. Head, J. Parker, A.F. Fotenos, D. Marcus, J.C. Morris, A.Z. Snyder, A unified approach for morphometric and functional data analysis in young, old, and demented adults using automated atlas-based head size normalization: reliability and validation against manual measurement of total intracranial volume, Neuroimage 23 (2004) 724–738. [26] R.J. Smith, Relative size versus controlling for size: interpretation of ratios in research on sexual dimorphism in the human corpus callosum, Curr. Anthropol. 46 (2005) 249–273. [27] K. Mevel, B. Desgranges, J.C. Baron, B. Landeau, V. de La Sayette, F. Viader, F. Eustache, G. Chetelat, Which SPM method should be used to extract hippocampal measures in early Alzheimer’s disease? J. Neuroimaging 21 (2011) 310–316. [28] C.D. Good, I.S. Johnsrude, J. Ashburner, R.N. Henson, K.J. Friston, R.S. Frackowiak, A voxel-based morphometric study of ageing in 465 normal adult human brains, Neuroimage 14 (2001) 21–36. [29] C.E. Sexton, C.L. Allan, M. Le Masurier, L.M. McDermott, U.G. Kalu, L.L. Herrmann, M. Maurer, K.M. Bradley, C.E. Mackay, K.P. Ebmeier, Magnetic resonance imaging in late-life depression: multimodal examination of network disruption, Arch. Gen. Psychiatry 69 (2012) 680–689. [30] A. Barros-Loscertales, H. Garavan, J.C. Bustamante, N. Ventura-Coampos, J.J. Llopis, V. Belloch, M.A. Parcet, C. Avila, Reduced striatal volume in cocainedependent patients, Neuroimage 56 (2011) 1021–1026.
Please cite this article in press as: G. Perlaki, et al., Are there any gender differences in the hippocampus volume after head-size correction? A volumetric and voxel-based morphometric study, Neurosci. Lett. (2014), http://dx.doi.org/10.1016/j.neulet.2014.04.013
421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502
ARTICLE IN PRESS
NSL 30526 1–5
Neuroscience Letters xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Neuroscience Letters journal homepage: www.elsevier.com/locate/neulet
Are there any gender differences in the hippocampus volume after head-size correction? A volumetric and voxel-based morphometric study
1
2
3
4 5 6 7 8 9 10
Q1
Gabor Perlaki a,b,∗ , Gergely Orsi a,b , Eniko Plozer c , Anna Altbacker c , Gergely Darnai c , Szilvia Anett Nagy b , Reka Horvath c , Arnold Toth c,d , Norbert Kovacs a,c , Peter Bogner b , Attila Schwarcz a,d , Jozsef Janszky a,c a
MTA-PTE Clinical Neuroscience MR Research Group, Ret u. 2, 7623 Pecs, Hungary Pecs Diagnostic Centre, Ret u. 2, 7623 Pecs, Hungary c Department of Neurology, University of Pecs, Ret u. 2, 7623 Pecs, Hungary d Department of Neurosurgery, University of Pecs, Ret u. 2, 7623 Pecs, Hungary b
11
12 26
a r t i c l e
i n f o
a b s t r a c t
13 14 15 16 17
Article history: Received 29 January 2014 Received in revised form 24 March 2014 Accepted 9 April 2014
18
25
Keywords: Head-size correction MRI VBM Volumetry Gender difference Hippocampus
27
1. Introduction
19 20 21 22 23 24
28 29 30 31 32 33 34 35 36 37
Previous findings on normal sexual dimorphism in hippocampal volume have not always been consistent. This study investigated gender differences in hippocampal volume using different head-size correction strategies. T1-weighted MR images were collected in 99 healthy, Caucasian, university students (66 female subjects; mean age: 23.1 ± 2.3, range: 19–31 years). Sexual dimorphism in hippocampus was investigated by automated MRI volumetry and voxel-based morphometry (VBM) using both general linear model (GLM) and proportion head-size correction strategies. Absolute hippocampal volumes were larger in men than women. After adjusting for head-size, the proportion method indicated larger hippocampi in women than men, while no gender differences were found using the GLM approach. Investigating absolute hippocampal volumes in 15 head-size matched pairs of males and females indicated no gender differences. We suggest that there is no sexual dimorphism in hippocampal size and the apparent gender differences found by the proportion method may have more to do with head-size than with sex. The GLM and proportion head-size correction strategies are not interchangeable and may yield different results. The importance of the present findings is mostly related to scientific reproducibility across MRI volumetry or VBM studies. © 2014 Published by Elsevier Ireland Ltd.
The hippocampus is a prime focus of MRI-based volumetric analysis in healthy subjects, cognitive disorders and neuropsychiatric conditions. Altered hippocampal volumes were reported in epilepsy, Alzheimer’s disease, schizophrenia and post-traumatic stress disorder [1,2]. Hippocampal volume measurements can assist the diagnosis of Alzheimer’s disease [3] and epilepsy [4]. Hippocampal volume changes were found to differ in some neuropsychiatric conditions as a function of gender [5,6]. Investigating gender difference in hippocampal volumes of normal subjects is essential to understand why male and female hippocampi differ in
∗ Corresponding author at: MTA-PTE Clinical Neuroscience MR Research Group, Ret u. 2, 7623 Pecs, Hungary. Tel.: +36 72535900; fax: +36 72 535911. E-mail address: [email protected] (G. Perlaki).
their predisposition for volumetric changes in certain conditions and to distinguish disease patterns from normal ones. Previous findings on sexual dimorphism in hippocampal volume have not always been consistent. For example, larger hippocampal volumes in females [7,8] and no difference in hippocampal volumes between males and females [9–11] were also reported after correcting for head-size differences. Interestingly, the above studies reporting gender difference used the proportion method [7,8], while the studies describing no gender difference used the general linear model (GLM) approach [9,10] or an extension of that [11] to adjust for head-size. Based on this observation, we hypothesized that part of the discrepancy in results may be attributed to different types of head-size correction used. Volumetric measures tend to be associated with normal variation in head-size; people with larger heads typically have larger brain structures [12]. To account for this irrelevant variance, volumetric measures are often “normalized” with respect to intracranial
http://dx.doi.org/10.1016/j.neulet.2014.04.013 0304-3940/© 2014 Published by Elsevier Ireland Ltd.
Please cite this article in press as: G. Perlaki, et al., Are there any gender differences in the hippocampus volume after head-size correction? A volumetric and voxel-based morphometric study, Neurosci. Lett. (2014), http://dx.doi.org/10.1016/j.neulet.2014.04.013
38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54
G Model NSL 30526 1–5
G. Perlaki et al. / Neuroscience Letters xxx (2014) xxx–xxx
2 55 56 57 58 59
60 61 62 63 64
ARTICLE IN PRESS
volume (ICV) [13]. The strong effect of ICV on voxel-wise volume information was also demonstrated, suggesting that head-size correction should also be considered in voxel-based morphometry (VBM) [10,14]. GLM and proportion approaches are two common methods to correct for ICV [13,15]. (i) The GLM approach corrects for head-size in a linear regression model, including ICV as a covariate of no interest. (ii) The proportion method normalizes for head-size directly by dividing the volumetric measure by the ICV (generating relative measures).
77
Although there are a few critical evaluations about the advantages/disadvantages of the GLM and proportion methods, there is no consensus regarding the most accurate way to adjust for headsize in statistical analyses [12,13,15–17]. The aim of the present study was twofold: (1) to re-examine the gender effect on hippocampal size using GLM and proportion head-size correction strategies. (2) To compare GLM and proportion approaches in hippocampal volumetry and VBM, using the same anatomical images from a set of normal control subjects. Our working hypothesis is that these two normalization methods may yield different results when comparing groups unmatched for head-size or when testing for inter-individual differences in a population with considerable variation in head-size.
78
2. Materials and methods
79
2.1. Subjects
65 66 67 68 69 70 71 72 73 74 75 76
87
Ninety-nine healthy, Caucasian, graduate or postgraduate university students (66 women and 33 men; mean age: 22.9 ± 2.2 and 23.5 ± 2.4, range: 19–31 and 20–30 years respectively) were included. Males and females did not differ in the distribution of education or age. All subjects got detailed information about the investigation and gave written informed consent prior to the examination. The study was approved by the local ethical committee and performed in accordance with the Declaration of Helsinki.
88
2.2. Magnetic resonance imaging
80 81 82 83 84 85 86
95
All measurements were performed on a 3T MRI scanner (MAGNETOM Trio, Siemens AG, Erlangen, Germany) with a 12-channel head coil. T1-weighted three-dimensional MPRAGE images were acquired with the following parameters: TR/TI/TE = 2530/1100/3.37 ms; flip angle = 7◦ ; 176 sagittal slices; slice thickness = 1 mm; FOV = 256 mm × 256 mm; matrix size = 256 × 256; receiver bandwidth = 200 Hz/pixel.
96
2.3. Data analysis
89 90 91 92 93 94
97 98 99
100 101 102 103 104 105
106 107 108 109
Sexual dimorphism in hippocampal volume was investigated using two different concepts (MR volumetry and VBM) and two head-size correction strategies (GLM and proportion). 2.3.1. Intracranial volume measure Since the ICV cannot be reliably measured on T1-weighted MR images [18], the reciprocal of volumetric scaling factor derived automatically from SIENAX, henceforth referred to as ICVsienax , was used as a surrogate for ICV [19]. The validity of ICVsienax as a predictor of ICV was demonstrated in an earlier study [18]. 2.3.2. MR volumetry Hippocampal volumes were estimated from T1-weighted MPRAGE images using the Freesurfer 4.5 image analysis suite (https://surfer.nmr.mgh.harvard.edu/). The technical details
were described previously [20]. Quality control was performed throughout the automatic processing stream. Error correction was performed when necessary, based on the recommended reconstruction workflow (http://surfer.nmr. mgh.harvard.edu/fswiki/RecommendedReconstruction/). Statistical analyses were performed using SPSS 20. Left and right hippocampal volumes were handled separately. (i) Absolute hippocampal volumes were compared between males and females using unpaired t-test. (ii) To correct for head-size through GLM, a multiple linear regression model was constructed with absolute hippocampal volume as dependent, while gender and ICVsienax as independent variables. (iii) To correct for head-size using the proportion method, the hippocampal volumes were divided by their corresponding ICVsienax . The relative hippocampal volumes were then compared between males and females using unpaired t-test. (iv) A subset of subjects with closely matched head-size was also analyzed. Males and females were paired based on an automated and objective method described earlier [21]. Briefly, the absolute differences in ICVsienax were computed for every possible pairings of males and females and then one male and one female with the minimum absolute difference in ICVsienax were paired and removed from the cohort. This procedure was repeated until there were no remaining pairs of males and females with relative difference in ICVsienax less than 5%. The above method resulted in 15 matched pairs of males and females. This group was then tested for gender differences in absolute hippocampal volumes using non-parametric Wilcoxon signed-rank test. (v) Both women and men were divided into two subgroups based on their median ICVsienax . The hippocampal volumes of women (or men) with smaller ICV were compared to women (men) with larger ICV by using both GLM and proportion head-size correction strategies. When using the GLM method a multiple linear regression model was constructed with absolute hippocampal volume as dependent, while the subgroup membership and ICVsienax as independent variables. In case of the proportion method the relative hippocampal volumes were compared between the subgroups by Mann–Whitney U test. In order to control for the potential confounding effects of age, the above comparisons were repeated using multiple linear regression with age as an independent variable. Results were considered significant at P ≤ 0.05 for all statistical tests. 2.3.3. Voxel-based morphometry Voxel-based morphometry was performed using FSL-VBM (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/fslvbm/). First, all structural images were brain-extracted using BET [22]. Next, tissue-type segmentation was carried out using FAST [23] and the resulting gray matter partial volume images were non-linearly registered to MNI152 standard space [24]. The registered images of 33 men and 33 randomly chosen women were averaged together with their respective mirror images to create a left–right symmetric studyspecific gray matter template (an equal number of subjects from both groups was used to avoid bias). All the 99 native gray matter images were non-linearly registered to this study-specific template and “modulated” to correct for volume changes due to spatial transformation. Two different types of modulations were performed. First, voxel values in gray matter images were corrected for volume changes due to both affine and nonlinear components of the registration (full modulation). Thus, the total amount of gray matter remains
Please cite this article in press as: G. Perlaki, et al., Are there any gender differences in the hippocampus volume after head-size correction? A volumetric and voxel-based morphometric study, Neurosci. Lett. (2014), http://dx.doi.org/10.1016/j.neulet.2014.04.013
110 111 112 113 114 115 116
117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149
150 151 152 153 154
155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172
G Model NSL 30526 1–5
ARTICLE IN PRESS G. Perlaki et al. / Neuroscience Letters xxx (2014) xxx–xxx
3
Fig. 1. Boxplots illustrating the significantly larger left and right relative hippocampal volumes in women compared to men (a and b respectively). Relative hippocampal volume was calculated as the absolute hippocampal volume (mm3 ) dividing by the corresponding ICVsienax . Whiskers are set at minimum and maximum, the horizontal line marks the median, whereas box indicates the interquartile range (25–75%).
173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190
the same as in the original images, so individual differences due to head-size are also preserved. Second, voxel values in gray matter images were corrected only for the nonlinear effects of the spatial transformation (nl modulation). Excluding the affine component from the modulation is an approximate of multiplying the full modulated data with the relative volumetric scaling factor due to affine registration.1 Considering that the reciprocal of this scaling factor is a valid measure of ICV [25], the nl modulated data are inherently adjusted for head-size according to the proportion method. The modulated gray matter images were then smoothed with an isotropic Gaussian kernel (sigma = 3 mm). Finally, to test for gender differences, voxelwise GLM limited to left and right hippocampal regions was applied using permutation-based non-parametric testing (10,000 permutations). The hippocampal regions were defined using the maximum probability Harvard-Oxford Subcortical Structural Atlas thresholded at 25% and masked by the gray matter mask automatically generated as part of the FSL-VBM protocol.
between the head-size based subgroups without considering ICVsienax in the statistical model. In order to control for the potential confounding effects of age, the above analyses were repeated including age as a covariate. Results were considered significant for P ≤ 0.05 (after initial cluster-forming threshold at t = 2.0), fully corrected for multiple comparisons. 3. Results Controlling for the potential confounding effects of age did not change the pattern of significance and age was not proved to be a significant predictor in any of the analyses performed. Final results are presented without controlling for age. There was no interaction between head-size and gender (i.e. ICVsienax *gender) in any of the analyses performed. 3.1. MR volumetry
191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210
(i) Absolute hippocampal size was compared between males and females using the full modulated data without considering ICVsienax in the statistical model. (ii) Head-size was controlled through GLM using the full modulated data and incorporating the gender and ICVsienax as independent variables in the statistical model. (iii) To correct for head-size using the proportion method, the nl modulated data were compared between males and females without considering ICVsienax in the statistical model. (iv) The 15 head-size matched pairs of males and females were tested for gender differences in absolute hippocampal size using the full modulated data without considering ICVsienax in the statistical model. (v) The hippocampal volumes of women (or men) with smaller ICV were compared to women (men) with larger ICV by using both GLM and proportion head-size correction strategies. Headsize was controlled through GLM using the full modulated data and incorporating the subgroup membership and ICVsienax as independent variables in the statistical model. In case of the proportion method, the nl modulated data were compared
1 If there were no skews introduced by the affine transformation, the nl modulated data would be exactly the same as multiplying the full modulated data with the relative volumetric scaling factor due to affine registration. In the presence of skews, the two dataset may slightly differ in certain voxels.
(i) Both left and right absolute hippocampal volumes were larger in men than women (P < 0.001). (ii) GLM analysis (including absolute hippocampal volume as dependent variable and gender and ICVsienax as independent variables) indicated positive effects of ICVsienax on both left and right hippocampal volumes (P < 0.001, see Supplementary Figure 1), while no significant gender differences were observed (see Supplementary Figure 2). (iii) Proportion method indicated that both left and right relative hippocampal volumes (absolute hippocampal volume divided by the ICVsienax ) were larger in women than men (unpaired ttest, P < 0.001; see Fig. 1). (iv) In the cohort of 30 young adults closely matched for head-size, there were no significant effects of gender on either left or right absolute hippocampal volumes. (v) The proportion method indicated significantly increased hippocampal size bilaterally in men with smaller ICV (n = 16) compared to men with larger ICV (n = 17), but not significant differences between women with smaller ICV and women with larger ICV. The GLM indicated no significant differences between subgroups in any sexes. Supplementary Figures 1 and 2 related to this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.neulet.2014.04.013.
Please cite this article in press as: G. Perlaki, et al., Are there any gender differences in the hippocampus volume after head-size correction? A volumetric and voxel-based morphometric study, Neurosci. Lett. (2014), http://dx.doi.org/10.1016/j.neulet.2014.04.013
211 212
213 214 215 216 217
218
219 220 221 222 223 224
225
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246
247 248 249
G Model NSL 30526 1–5
ARTICLE IN PRESS G. Perlaki et al. / Neuroscience Letters xxx (2014) xxx–xxx
4
Q3 Fig. 2. Voxelwise analysis of hippocampi using the nl modulated data without considering ICVsienax as a confounding variable. Red-yellow shows voxels demonstrating significantly increased relative amount of gray matter in women compared to men. Color bar represents t-values. The background image is the MNI152 standard space T1 template. X-, Y- and Z-values indicate the MNI slice coordinates in millimeter. Images are shown in radiological convention (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).
250
251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270
3.2. Voxel-based morphometry (i) Using the full modulated data without considering ICVsienax as a confounding variable indicated that absolute hippocampal size was larger in men than women. (ii) No statistically significant gender difference was found using the full modulated data and controlling for head-size through GLM. The voxelwise analysis of hippocampi indicated significant positive effects of ICVsienax bilaterally (see Supplementary Figure 3). (iii) The nl modulated data (inherently corrected for head-size according to the proportion method) indicated significantly larger hippocampi in women than men (see Fig. 2). (iv) In case of the matched pairs of male and female subjects, no significant gender differences were found in absolute hippocampal size using the full modulated data without considering ICVsienax as a confounding variable. (v) The proportion method indicated significantly increased left hippocampal size in women (or men) with smaller ICV compared to women (men) with larger ICV. The GLM method indicated no significant differences between subgroups in any sexes.
273
Supplementary Figure 3 related to this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.neulet.2014.04.013.
274
4. Discussion
271 272
275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294
In many studies the volumetric measures of hippocampus are controlled for ICV, in order to allow for comparisons between subjects who differ in head-size. GLM and proportion methods are the two most common approaches for making this adjustment. Although both are used pervasively in neuroimaging research, there is little direct comparison between them. In practice, they are used rather interchangeably, not really considering the difference between them. In the present study, sexual dimorphism in hippocampus was investigated by using both general linear model (GLM) and proportion head-size correction strategies. Both MRI volumetry and VBM indicated significant positive effects of ICVsienax on absolute hippocampal size bilaterally after the effects of gender were removed (Supplementary Figures 1 and 3), which is consistent with previous data suggesting that increasing head-size is associated with larger volumes [14]. As expected based on the above, absolute hippocampal size was larger in men than women. Both MRI volumetry and VBM indicated significantly larger hippocampal volume in women than men, when the effect of headsize was “normalized” by the proportion method (Figs. 1 and 2),
while this apparent gender effect was absent, when head-size was controlled through GLM (Supplementary Figure 2). This is good evidence that the two methods are not interchangeable and may yield different results. The discrepancy can be resolved by recognizing the difference between proportion and GLM approaches and the different questions they address. Our results are consistent with earlier studies using the proportion method [7,8] or GLM approach [9,10]. While these studies may provide support of our findings, they are not directly comparable to our investigation, due to methodological differences. In case of GLM the phrase “correcting for” refers to statistical control. Statistical control of head-size serves to remove the portion of the variance in hippocampal volume due to head-size. Unlike GLM, the proportion method calculates ratios (relative sizes), which was shown to be ineffective for statistical control in most cases [26]. Calculation of proportions with the intention to control for head-size is based on the assumption that hippocampal volume varies with head-size proportionally, so their relationship is linear with zero y-intercept for the regression line. The zero y-intercept assumption is seldom satisfied for real data. If this assumption fails, simple ratio is expected to be associated with its denominator (i.e. head-size), thus not resulting in a “head-size-free” measure [12]. To address this issue, we checked whether the volumetric measures normalized according to the proportion method correlate with the ICVsienax . Using Spearman’s correlation, we found that both the left and right relative hippocampal volumes as well as the mean hippocampal gray matter values obtained by averaging the smoothed nl modulated data in left and right hippocampal regions were inversely correlated with ICVsienax (P < 0.002), which remained significant (P < 0.05) after statistically controlling for gender through multiple linear regression. These correlations demonstrate the inability of proportion method to perform statistical control. Based on our present results, we suggest that before using either GLM or proportion method, the investigators should clarify whether they are attempting to statistically control for head-size or to examine relative size with respect to head-size and choose accordingly. We need to emphasize this differentiation, because it is still often overlooked and considered secondary in scientific studies. For instance, in a recent VBM study the two concepts are mixed even within a single evaluation [27]. In the literature there is clear distinction between VBM analyses with and without a modulation step. The analysis of modulated data allows testing for regional differences in the amount (volume) of gray matter, whereas the unmodulated data reflect regional gray matter concentration [28]. However, there is no such clear differentiation between nl modulated and full modulated datasets. Modulation of non-linear warping only (nl modulation) have gained popularity in VBM analyses, suggesting that this type of modulation accounts for head-size and no further corrections are needed [29,30]. If the effect of head-size was completely removed by not modulating for volume changes due to the affine part of the registration, any significant relationship between nl modulated data and ICVsienax would not be found. However, our data indicate a negative correlation between nl modulated data in the hippocampus and ICVsienax as a remaining effect of head-size. The idea that we do not need to estimate ICV and correct for it in subsequent statistical analyses is undoubtedly attractive, but if we are interested in “head-size-free” hippocampal size, the full modulated data should be used including ICV measure as a confounding variable in the statistical model. Results from the subset of subjects closely matched for headsize support the findings of the GLM approach. Based on these results and the observation that smaller brains exhibit relatively larger hippocampi regardless of gender, we suggest that there is no sexual dimorphism in hippocampal size and the apparent gender differences found by the proportion method may be not
Please cite this article in press as: G. Perlaki, et al., Are there any gender differences in the hippocampus volume after head-size correction? A volumetric and voxel-based morphometric study, Neurosci. Lett. (2014), http://dx.doi.org/10.1016/j.neulet.2014.04.013
295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360
G Model NSL 30526 1–5
ARTICLE IN PRESS G. Perlaki et al. / Neuroscience Letters xxx (2014) xxx–xxx
361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394
395
396 397 398
399
gender specific but rather a consequence of the well-known sexual dimorphism in head-size. This impression is further strengthened by the increased hippocampal size of women (or men) with smaller ICV compared to women (men) with larger ICV when using the proportion method. Although, the difference was not significant in all of the examined comparisons (probably due to the reduced variability of hippocampal volumes and ICVsienax when sexes are examined separately), these results provide more straightforward and convincing evidence that the proportion method is unable to control for head-size even within very homogeneous samples. To our knowledge this is one of the very first studies exploring differences between proportion and GLM methods in practice using both MRI volumetry and VBM. Earlier studies comparing head-size “normalization” methods focused mainly on theoretical issues such as reliability [12,16] and how the introduction of systematic and random errors may affect the final outcome of normalized measures [13] or used pediatric samples [17]. None of these studies was concerned about the potential impact of head-size normalization on voxelwise techniques (e.g. VBM), where head-size normalization and modulation step are highly related issues. A potential limitation of our study is the unequal number of males and females. However, when the 33 men were compared to 33 randomly selected women the overall pattern of results remained unchanged, suggesting that our findings are not driven by unequal sample sizes. Conclusively, we suggest that there is no sexual dimorphism in hippocampal size and the apparent differences found by the proportion method may have more to do with head-size than with sex. The importance of the present findings about the fundamental differences between GLM and proportion approaches is mostly related to scientific reproducibility across MRI volumetry or VBM studies. Our results may have clinical importance, when adjusted hippocampal volumes are considered as diagnostic tools in epilepsy or Alzheimer’s disease. Conflict of interest We are not aware of any actual or potential conflicts of interest regarding this work. The authors alone are responsible for the content and writing of the paper. Acknowledgements
405
This work was supported by the Hungarian Research Fund (OTKA K105357; OTKA PD103964), PTE AOK-KA-2013/12, PTE AOK-KA No: 2011/34039/KA-OTKA/11-10, KTIA-NAP-13-1-20130001 and the grant SROP-4.2.2/A-11/1/KONV-2012-0017. AS and NK were supported by the Bolyai Scholarship. NK was also supported by the Romhanyi Scholarship and Magyary Scholarship.
406
References
Q2 400 401 402 403 404
407 408 409 410 411 412 413 414 415 416 417 418 419 420
[1] E. Geuze, E. Vermetten, J.D. Bremner, MR-based in vivo hippocampal volumetrics: 2. Findings in neuropsychiatric disorders, Mol. Psychiatry 10 (2005) 160–184. [2] M.E. Smith, Bilateral hippocampal volume reduction in adults with posttraumatic stress disorder: a meta-analysis of structural MRI studies, Hippocampus 15 (2005) 798–807. [3] K. Kantarci, Y. Xu, M.M. Shiung, P.C. O’Brien, R.H. Cha, G.E. Smith, R.J. Ivnik, B.F. Boeve, S.D. Edland, E. Kokmen, E.G. Tangalos, R.C. Petersen, C.R. Jack Jr., Comparative diagnostic utility of different MR modalities in mild cognitive impairment and Alzheimer’s disease, Dement. Geriatr. Cogn. Disord. 14 (2002) 198–207. [4] C.R. Jack Jr., F.W. Sharbrough, G.D. Cascino, K.A. Hirschorn, P.C. O’Brien, W.R. Marsh, Magnetic resonance image-based hippocampal volumetry: correlation with outcome after temporal lobectomy, Ann. Neurol. 31 (1992) 138–146. [5] R.S. Briellmann, S.F. Berkovic, G.D. Jackson, Men may be more vulnerable to seizure-associated brain damage, Neurology 55 (2000) 1479–1485.
5
[6] C. Exner, B. Nehrkorn, V. Martin, M. Huber, K. Shiratori, W. Rief, Sex-dependent hippocampal volume reductions in schizophrenia relate to episodic memory deficits, J. Neuropsychiatry Clin. Neurosci. 20 (2008) 227–230. [7] C.A. Szabo, J.L. Lancaster, J. Xiong, C. Cook, P. Fox, MR imaging volumetry of subcortical structures and cerebellar hemispheres in normal persons, AJNR Am. J. Neuroradiol. 24 (2003) 644–647. [8] P.A. Filipek, C. Richelme, D.N. Kennedy, V.S. Caviness Jr., The young adult human brain: an MRI-based morphometric analysis, Cereb. Cortex 4 (1994) 344–360. [9] C.R. Jack Jr., C.K. Twomey, A.R. Zinsmeister, F.W. Sharbrough, R.C. Petersen, G.D. Cascino, Anterior temporal lobes and hippocampal formations: normative volumetric measurements from MR images in young adults, Radiology 172 (1989) 549–554. [10] G.S. Pell, R.S. Briellmann, C.H. Chan, H. Pardoe, D.F. Abbott, G.D. Jackson, Selection of the control group for VBM analysis: influence of covariates, matching and sample size, Neuroimage 41 (2008) 1324–1335. [11] R.C. Gur, F. Gunning-Dixon, W.B. Bilker, R.E. Gur, Sex differences in temporolimbic and frontal brain volumes of healthy adults, Cereb. Cortex 12 (2002) 998–1003. [12] D.H. Mathalon, E.V. Sullivan, J.M. Rawles, A. Pfefferbaum, Correction for head size in brain-imaging measurements, Psychiatry Res. 50 (1993) 121–139. [13] M.P. Sanfilipo, R.H. Benedict, R. Zivadinov, R. Bakshi, Correction for intracranial volume in analysis of whole brain atrophy in multiple sclerosis: the proportion vs. residual method, Neuroimage 22 (2004) 1732–1743. [14] J. Barnes, G.R. Ridgway, J. Bartlett, S.M. Henley, M. Lehmann, N. Hobbs, M.J. Clarkson, D.G. MacManus, S. Ourselin, N.C. Fox, Head size, age and gender adjustment in MRI studies: a necessary nuisance? Neuroimage 53 (2010) 1244–1255. [15] L.M. O’Brien, D.A. Ziegler, C.K. Deutsch, D.N. Kennedy, J.M. Goldstein, L.J. Seidman, S. Hodge, N. Makris, V. Caviness, J.A. Frazier, M.R. Herbert, Adjustment for whole brain and cranial size in volumetric brain studies: a review of common adjustment factors and statistical methods, Harv Rev Psychiatry 14 (2006) 141–151. [16] S. Arndt, G. Cohen, R.J. Alliger, V.W. Swayze 2nd, N.C. Andreasen, Problems with ratio and proportion measures of imaged cerebral structures, Psychiatry Res. 40 (1991) 79–89. [17] L.M. O‘Brien, D.A. Ziegler, C.K. Deutsch, J.A. Frazier, M.R. Herbert, J.J. Locascio, Statistical adjustments for brain size in volumetric neuroimaging studies: some practical implications in methods, Psychiatry Res. 193 (2011) 113–122. [18] G. Fein, V. Di Sclafani, C. Taylor, K. Moon, J. Barakos, H. Tran, B. Landman, R. Shumway, Controlling for premorbid brain size in imaging studies: T1-derived cranium scaling factor vs. T2-derived intracranial vault volume, Psychiatry Res. 131 (2004) 169–176. [19] S.M. Smith, Y. Zhang, M. Jenkinson, J. Chen, P.M. Matthews, A. Federico, N. De Stefano, Accurate, robust, and automated longitudinal and cross-sectional brain change analysis, Neuroimage 17 (2002) 479–489. [20] B. Fischl, D.H. Salat, E. Busa, M. Albert, M. Dieterich, C. Haselgrove, A. van der Kouwe, R. Killiany, D. Kennedy, S. Klaveness, A. Montillo, N. Makris, B. Rosen, A.M. Dale, Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain, Neuron 33 (2002) 341–355. [21] B.A. Ardekani, K. Figarsky, J.J. Sidtis, Sexual dimorphism in the human corpus callosum: an MRI study using the OASIS brain database, Cereb. Cortex 23 (2013) 2514–2520. [22] S.M. Smith, Fast robust automated brain extraction, Hum. Brain Mapp. 17 (2002) 143–155. [23] Y. Zhang, M. Brady, S. Smith, Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm, IEEE Trans. Med. Imaging 20 (2001) 45–57. [24] J.L. Andersson, M. Jenkinson, S. Smith, Non-Linear Registration, aka Spatial Normalisation FMRIB Technical Report TR07JA2, FMRIB Analysis Group of the University of Oxford, 2007. [25] R.L. Buckner, D. Head, J. Parker, A.F. Fotenos, D. Marcus, J.C. Morris, A.Z. Snyder, A unified approach for morphometric and functional data analysis in young, old, and demented adults using automated atlas-based head size normalization: reliability and validation against manual measurement of total intracranial volume, Neuroimage 23 (2004) 724–738. [26] R.J. Smith, Relative size versus controlling for size: interpretation of ratios in research on sexual dimorphism in the human corpus callosum, Curr. Anthropol. 46 (2005) 249–273. [27] K. Mevel, B. Desgranges, J.C. Baron, B. Landeau, V. de La Sayette, F. Viader, F. Eustache, G. Chetelat, Which SPM method should be used to extract hippocampal measures in early Alzheimer’s disease? J. Neuroimaging 21 (2011) 310–316. [28] C.D. Good, I.S. Johnsrude, J. Ashburner, R.N. Henson, K.J. Friston, R.S. Frackowiak, A voxel-based morphometric study of ageing in 465 normal adult human brains, Neuroimage 14 (2001) 21–36. [29] C.E. Sexton, C.L. Allan, M. Le Masurier, L.M. McDermott, U.G. Kalu, L.L. Herrmann, M. Maurer, K.M. Bradley, C.E. Mackay, K.P. Ebmeier, Magnetic resonance imaging in late-life depression: multimodal examination of network disruption, Arch. Gen. Psychiatry 69 (2012) 680–689. [30] A. Barros-Loscertales, H. Garavan, J.C. Bustamante, N. Ventura-Coampos, J.J. Llopis, V. Belloch, M.A. Parcet, C. Avila, Reduced striatal volume in cocainedependent patients, Neuroimage 56 (2011) 1021–1026.
Please cite this article in press as: G. Perlaki, et al., Are there any gender differences in the hippocampus volume after head-size correction? A volumetric and voxel-based morphometric study, Neurosci. Lett. (2014), http://dx.doi.org/10.1016/j.neulet.2014.04.013
421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502
