A New Analysis of Resting State Connectivity and Graph Theory Reveals Distinctive Short-Term Modulations due to Whisker Stimulation in Rats

Abstract

Resting state (RS) connectivity has been increasingly studied in healthy and diseased brains in humans and animals. This paper presents a new method to analyze RS data from fMRI that combines multiple seed correlation analysis with graph-theory (MSRA). We characterize and evaluate this new method in relation to two other graph-theoretical methods and ICA. The graph-theoretical methods calculate cross-correlations of regional average time-courses, one using seed regions of the same size (SRCC) and the other using whole brain structure regions (RCCA). We evaluated the reproducibility, power, and capacity of these methods to characterize short-term RS modulation to unilateral physiological whisker stimulation in rats. Graph-theoretical networks found with the MSRA approach were highly reproducible, and their communities showed large overlaps with ICA components. Additionally, MSRA was the only one of all tested methods that had the power to detect significant RS modulations induced by whisker stimulation that are controlled by family-wise error rate (FWE). Compared to the reduced resting state network connectivity during task performance, these modulations implied decreased connectivity strength in the bilateral sensorimotor and entorhinal cortex. Additionally, the contralateral ventromedial thalamus (part of the barrel field related lemniscal pathway) and the hypothalamus showed reduced connectivity. Enhanced connectivity was observed in the amygdala, especially the contralateral basolateral amygdala (involved in emotional learning processes). In conclusion, MSRA is a powerful analytical approach that can reliably detect tiny modulations of RS connectivity. It shows a great promise as a method for studying RS dynamics in healthy and pathological conditions.

Capturing early human memory consolidation utilizing the higher functional specificity of 7T compared to 3T-fMRI

“…The average time course of each seed region was correlated with every voxel time course within the brain and the resulting correlation maps were thresholded using BY-FDR (q=0.05, n=298 time points) to determine significantly correlating voxels. In opposite to the predefined seed regions the location of those target voxels per brain region was determined purely data driven, thereby enhancing sensitivity of the connectivity between pairs of nodes 27 . The average Pearson’s correlation r of all target voxels per brain region was used to define the connectivity strength to the respective seed region.…”

“…7T resting-state graphs showed higher specificity of the underlying functional connectivity Compared to the ICA-derived RSNs, graph-theoretical FC analysis generates deeper insights into the brain-wide information flow. For this purpose, we used the multi-seed-region approach (MSRA 27 ) to create subject specific correlation matrices for rs1 using 158 brain regions (n=18, paired design).…”

“…Graph-theoretical resting-state analysis describes brain regions as nodes and focusses on the connections between each pair of brain region, called edges in network terminology. To assess the functional connectivity between the brain regions we used a multi-seed-region approach (MSRA) introduced by Kreitz et al ( 2018) 27 . Briefly, a predefined seed region was placed automatically in the center of mass of each brain region as determined by atlas registration and region segmentation described above.…”

“…NBS exploits the interconnected extent of univariate significant different edges by permutation of subject specific networks between experimental groups. In a paired design, we introduced an additional paired control group in order to control for general effects of repeated measurements (pNBS) 27 . Here, we used an unpublished control data set of 11 healthy subjects ( 7females, age 25 to 63) who were measured twice with an interval of two days on the same Siemens TRIO Magnetom Scanner (GE-EPI, TR: 3 sec, TE: 30 ms, flip angle: 90°, matrix size 128 x 128 pixel, pixel resolution: 1.5 mm x 1.5 mm, 36 slices, slice thickness: 3 mm, slice gap: 0.75 mm, 200 volumes).…”

“…paired rs1 scans with 7T and 3T), and all remaining connected components equal or smaller to the control component were eliminated. Finally, the family-wise-error (p FWE ) was controlled by 10,000 randomized permutations of pairs between experimental and control group (for details see 27 ). The α-value ensures that the observed differences correspond dominantly to the experimental conditions, in this case the field strength, whereas the p FWE value indicates the probability that the observed component is not random.…”