BOLD activity related functional connectivity in patients with schizophrenia and controls during working memory
Grega Repovš, Deanna M Barch
Department of Psychology, Washington University in Saint Louis, 63130 MO, USA

Brain imaging studies have been successful in promoting functional specialization as an important principle of organization in the human brain. However, recently more and more studies are focusing on exploring functional and effective connectivity in normal and dysfunctional brain. While similar in their goals, published studies vary widely in their operationalization of connectivity, as well as the methods and techniques for data preprocessing and analysis. The aims of the current study were to: a) compare results obtained by different methods of extracting timeseries that can be related to either sustained or transient changes in BOLD activity, b) use seed-region correlation images as means of identifying functional networks submitted to further effective connectivity analysis, c) identify functional and effective connectivity changes in patients with schizophrenia during working memory performance.

The analysis was conducted on functional MRI imaging data of 38 patients with schizophrenia and their matched controls presented in a previous study (1). The data for each subject consisted of two block-design runs (TR = 2.5 sec; 102 frames), of a 2-back working memory task (4x16 trials) and fixation (4, 3x10, 4 frames). Word stimuli were used in one and faces in the other run. After initial preprocessing steps and transformation to atlas space, functional runs were spatially smoothed (6mm FWHM Gaussian kernel) and high-pass filtered (0.009 Hz). Nuisance variables were removed by means of multiple regression. An initial set of regions of interest (ROI) was defined based on previous activation analysis. In addition, a set of seed-region correlation images were computed based on whole time series, task related time series and mean block activation levels. Average group correlation images were used to identify the networks related to individual timeseries. Differences in connectivity due to diagnostic status and task modality - words vs. faces - were explored. ROI representing identified network nodes were used as a basis for subsequent structural equation modeling (SEM).

Results & Discussion
Initial results suggest that the patterns of functional connectivity within groups differ between timeseries. While stronger and more extensive in timeseries that include both sustained and transient responses, important changes in functional connectivity between patients and controls remain to be present in timeseries based on just transient activity as well. Specifically, individuals with schizophrenia demonstrate consistent patterns of reduced connectivity between inferior frontal, posterior parietal and cerebellar regions. We also found evidence of reduced connectivity between anterior cingulate and basal ganglia regions. In addition, individuals with schizophrenia show evidence of enhanced intraregional connectivity, particular among posterior parietal and cingulate regions.

The study of functional and effective connectivity entails a number of processing decisions that can importantly affect the results obtained. Stepwise approach based on initial seed-region correlation analysis can be helpful in identifying key ROI and progressing from the study of functional to effective connectivity. Consistent with prior research, the current results provide evidence of reduced prefrontal to parietal connectivity in schizophrenia, but also suggest reduced connectivity with cerebellar regions and enhanced intraregional connectivity.

1. D. Barch, et al., J Abnorm Psychol 111, 478-94 (2002)