EP 33. On the relationship between gray matter and behavioral data: lessons learned. Issue 9 (September 2016)

• Record Type: Journal Article
• Authors: Wensing, T.; Genon, S.; Hoffstaedter, F.; Reid, A.; Caspers, S.; Grefkes, C.; Nickl-Jockschat, T.; Eickhoff, S.B.
• Is Part Of: Clinical neurophysiology. Volume 127:Issue 9(2016:Sep.)
• Page Start: e249
• Page End: e250
• Publication Date: 2016-09
• Subjects: Neurophysiology -- Periodicals; Electroencephalography -- Periodicals; Electromyography -- Periodicals; Neurology -- Periodicals; 612.8
• DOI: 10.1016/j.clinph.2016.05.088
• Languages: English
• ISSNs: 1388-2457

Abstract

Abstract : Recently, we functionally characterized five right dorsal premotor (PMdr) subregions (identified by coactivation based parcellation) by quantitative forward and reverse inference based on a wide range of task-based activation studies (Genon et al, in prep.). We found that the rostral subregion was associated to high-level cognitive tasks, the caudal one to motor functions, the dorsal one to hand-movements, the ventral one to visual functions while the central one showed a heterogeneous profile. Here, we aimed to reinforce and complement this characterization by using correlations between brain morphology and standard neuropsychological tests in two independent large datasets, which were matched for age, gender, education, depression and handedness (Forschungszentrum Jülich [FZJ, n = 87, age range: 21–71] and Nathan Kline Institute [NKI; 135: 20–75]). T1 images were processed with the VBM8 toolbox (Friston & Ashburner, 2000; using linear and non-linear modulation). The five PMdr subregions were used as regions of interest (ROIs), from which regional gray matter volume (GMV) was extracted. We performed partial correlations to examine correlation between each ROI GMV and neuropsychological test performance when controlling for confounding effects of age, education and handedness. To further examine the reliability of correlation, we also performed the analyses in 1000 subsamples of 30 randomly selected subjects within each sample individually. Our analyses yielded generally low correlation coefficients of which only few were statistically significant even at a lenient threshold of p < .05, uncorrected for multiple testing. Notably, a considerable number of correlations showed an inverse relationship between test performance and brain structure, i.e., higher performance was associated with lower GMV. In addition, the patterns of correlations were inconsistent across samples in magnitude, significance and direction. Finally, subsampling revealed that the pattern of correlations that could be observed varied widely when analyses were based on randomly drawn small samples. Our correlational approach, relating GMV and performance across a wide range of standard neuropsychological tests in two large population-based samples did not corroborate the profiles previously revealed using functional decoding based on fMRI-activations. On the one hand, these findings suggest that the functional specialization of some cortical regions, as highlighted by fMRI studies, does not necessarily imply a significant covariance of their actual structure with related standard task performance across individuals in the healthy population. That is, a cognitive-morphologic approach based on healthy interindividual differences may not mirror functional characterization revealed by activation studies in some brain regions. On the other hand, the fact that we found (all kinds of) significant associations in the smaller, randomly assembled subsamples also highlights the high likelihood of spurious findings when performing such correlation analyses in small samples as prevalent in current neuroimaging studies.