Universität zu Köln

Jain, Shraddha (2022). Sex Differences in the Correlation between Empirical and Simulated Brain Connectomes. Masters thesis, Universität zu Köln.

PDF (Master Thesis Report) Master_thesis_report_shraddha_jain.pdf - Accepted Version Download (14MB)

Abstract

Investigation of sex differences in the brain connectomes, that is, comprehensive maps of the underlying structural and functional connections across different regions of the human brain defined by a specific parcellation scheme, has been an active field of research in neuroscience. There are largely two main approaches to study the complex dynamics of the brain - empirical neuroimaging techniques and whole brain dynamical models. The studies in the past, however, have utilized only the empirical brain data extracted from the neuroimaging techniques in their investigation of sex differences. The aim of this project, therefore, is to also include the simulated data generated from the whole brain dynamical models and examine the sex differences in its correlation with the empirical data, within a given brain parcellation scheme. The analysis involves 272 subjects from the Human Connectome Project (144 females). For each individual under 11 brain parcellation schemes, we calculated an empirical structural connectivity (eSC), an empirical functional connectivity (eFC) using the empirical neuroimaging data and two simulated functional connectivity (sFC) matrices based on the ensembles of coupled phase- (PO) and limit-cycle (LC) oscillators. The sex difference was then investigated in the goodness-of-fit - the maximal Pearson’s correlation coefficient between the sFC and the eFC matrices (corr(sFC, eFC)max). We observed a significantly higher correlation for males within each of the 11 parcellation schemes. Since the models utilize the empirical information, we regressed out the brain size and empirical structure-function relationship, to check if the sex difference still persists. After the regression, this difference remains significant for 10 and 8 parcellation schemes for PO and LC model, respectively. We speculated that a potential reason for this could be the differences in the ‘complexity’ of the eFC matrix between the two sexes, which may in turn negatively influence the quality of their model fitting, i.e. a higher ‘complexity’ implying a lower fit. We then calculated three potential ‘complexity’ measures - the Shannon entropy H(eFC), the standard deviation σ(|eFC|) and the area under the eigen value curve A(λeFC) of the eFC matrix, to not only examine the sex differences in them, but to also investigate their ability to account for the sex differences in the goodness-of-fit. We found that the first two measures are significantly higher for males and are, therefore, positively correlated with the goodness-of-fit. However, the third measure is found to be significantly higher for females, resulting a negative correlation with the goodness-of-fit. The study was, therefore, successful in establishing the statistical differences in the goodness-of-fit and other properties of the eFC matrix between males and females. However, a precise interpretation of the term ‘complexity’ of a connectome and the validity of our hypothesis about its negative correlation with the goodness-of-fit demands further investigation.