M69. Changes in Neural Measures of Emotion Processes Following Targeted Social Cognition Training

Abhishek Saxena, Erin Guty, David Dodell-Feder, Hong Yin, Kristen Haut, Mor Nahum, Christine Hooker
2017 Schizophrenia Bulletin  
This analysis aims to validate the use of graph analysis (GA) metrics in an Ultra High Risk of Psychosis (UHR) sample in order to inform future research in this area. Methods: A 6 minute resting state fMRI scan (154 volumes) was administered to 21 UHR and 5 controls recruited from the South London area through the MTOP study (fellowship of Dr. Matthew Kempton). Parameters were: EPI BOLD, TR = 2000 ms, TE = 30 ms, flip angle = 77, slice thickness = 4 mm, slice gap = 5mm, FoV = 220 mm, matrix
more » ... = 64 × 64. Data was preprocessed within FSL FEAT, which included deletion of the first 10 volumes, spatial smoothing (FWHM: 5mm), high pass filtering (0.01 Hz), registration to a T1 image, transformation into MNI space and motion correction. MELODIC ICA and FSL FIX was used to automatically remove components suggestive of noise (cardiac activity, motion, CSF, white matter). Time series were extracted by averaging intensities within 48 cortical ROIs (Harvard-Oxford Cortical Structural Probabilistic Atlas, thresholded at 25%). Correlation matrices were formed then thresholded across a wide range of sparsity thresholds (15%-30% in steps of 1%). Whole brain GA metrics (CC: clustering coefficient, SPL: characteristic shortest path length, LE: local efficiency) were then calculated at each threshold. The area under the curve (AUC) was taken for each threshold range of GA metrics giving singular results for each measure. Results: Comparison with random networks of the same node number suggested network small worldness. Bootstrapped Yuen's tests showed no significant differences in GA metrics between groups. Following this only the UHR sample was explored. Nonparametric bootstrapped multiple regressions were performed to test for associations between AUC GA metrics and PANSS scores, and WAIS symbol coding, symbol search and digit span, while controlling for age and gender. SPL was negatively associated with PANSS scores (95% CI: −26.93 to −3.80), positively with digit symbol coding (CI: 1.61 to 21.73) and digit symbol search (CI: 0.77 to 9.53). No other significant associations were found. Conclusion: This analysis was underpowered to fully explore group differences generalizable to UHR populations. However, lower characteristic shortest path length is reflective of a more efficient network, so associations with PANSS scores and cognitive tests were as expected. A relation to PANSS scores is in line with limited evidence suggesting aberrant functional connectivity in UHR participants (Lord et al.
doi:10.1093/schbul/sbx022.064 fatcat:ma67j5x5erc6rdrh26r5r7tpw4