Using spatial multiple regression to identify intrinsic connectivity networks involved in working memory performance

• Published in: Human brain mapping Vol. 33; no. 7; pp. 1536 - 1552
• Main Authors: Gordon, Evan M., Stollstorff, Melanie, Vaidya, Chandan J.
• Format: Journal Article
• Language: English
• Published: Hoboken Wiley Subscription Services, Inc., A Wiley Company 01.07.2012; Wiley-Liss; John Wiley & Sons, Inc
• Subjects: Adolescent; Biological and medical sciences; Brain - physiology; default mode; dorsal attention; Female; fMRI; frontoparietal; Fundamental and applied biological sciences. Psychology; Human; Humans; Investigative techniques, diagnostic techniques (general aspects); Learning. Memory; Male; Medical sciences; Membrane Potentials - physiology; Memory; Memory, Short-Term - physiology; Nerve Net - physiology; Nervous system; Photic Stimulation - methods; Psychology. Psychoanalysis. Psychiatry; Psychology. Psychophysiology; Psychomotor Performance - physiology; Radiodiagnosis. Nmr imagery. Nmr spectrometry; reaction time; Reaction Time - physiology; resting state; set maintenance; Young Adult; default mode; Nervous system diseases; Radiodiagnosis; set maintenance; Attention; dorsal attention; Nuclear magnetic resonance imaging; fMRI; resting state; Reaction time; Working memory; Performance; frontoparietal
• ISSN: 1065-9471; 1097-0193; 1097-0193
• DOI: 10.1002/hbm.21306

Many researchers have noted that the functional architecture of the human brain is relatively invariant during task performance and the resting state. Indeed, intrinsic connectivity networks (ICNs) revealed by resting‐state functional connectivity analyses are spatially similar to regions activated during cognitive tasks. This suggests that patterns of task‐related activation in individual subjects may result from the engagement of one or more of these ICNs; however, this has not been tested. We used a novel analysis, spatial multiple regression, to test whether the patterns of activation during an N‐back working memory task could be well described by a linear combination of ICNs delineated using Independent Components Analysis at rest. We found that across subjects, the cingulo‐opercular Set Maintenance ICN, as well as right and left Frontoparietal Control ICNs, were reliably activated during working memory, while Default Mode and Visual ICNs were reliably deactivated. Further, involvement of Set Maintenance, Frontoparietal Control, and Dorsal Attention ICNs was sensitive to varying working memory load. Finally, the degree of left Frontoparietal Control network activation predicted response speed, while activation in both left Frontoparietal Control and Dorsal Attention networks predicted task accuracy. These results suggest that a close relationship between resting‐state networks and task‐evoked activation is functionally relevant for behavior, and that spatial multiple regression analysis is a suitable method for revealing that relationship. Hum Brain Mapp, 2011. © 2011 Wiley‐Liss, Inc.