ࡱ > b {w jbjb . p ^ ^ ^ n p p p v 8 6 4 j v * " ( , R R" p p p p p ^ ~ ^ p p c p p (l ^ { 0 , " " p v v :
v v :
Supplemental Digital Content 1: Independent Component Analysis Results
Front-parietal Networks Default Network and Executive Control Networks
During wakefulness, independent component analysis (ICA) could identify reproducible connectivity patterns in the default network and the executive control networks (see Supplementary Digital Content 6 and 7) with similar spatial distribution to that found with the region of interest (ROI)-driven approach. Thalamic involvement was identified in all three networks. Areas found to be anticorrelated to default network using the ROI-driven approach were also anticorrelated in results obtained with ICA (see Supplementary Digital Content 6 and 7), though the spatial extent of areas surviving statistical threshold was smaller. The fact that ICA was less able to identify anticorrelations with default network as a whole during wakefulness might reflect a certain heterogeneity in negative connectivity between the different default network nodes. ADDIN EN.CITE Uddin200852835283528317Uddin, L. Q.Clare Kelly, A. M.Biswal, B. B.Xavier Castellanos, F.Milham, M. P.The Phyllis Green and Randolph Cowen Institute for Pediatric Neuroscience, New York University Child Study Center, New York, New York.Functional connectivity of default mode network components: Correlation, anticorrelation, and causalityHum Brain MappHuman brain mappingHum Brain Mapp2008Jan 241065-9471 (Print)18219617http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=18219617 Eng1
During deep sedation, as with the ROI-driven approach, ICA could identify partially preserved residual functional connectivity both in the executive control networks and the default network. For the right executive control network, ICA identified residual connectivity in posterior parietal cortices, and a trend towards significance was also observed in frontal cortices. For the left executive control network, residual connectivity was also identified in posterior parietal cortices, in presupplementary motor area/anterior cingulate cortex, and temporooccipital junction (see Supplementary Digital Content 6 and 7). Though a trend towards significance was observed in frontal areas, their involvement in left executive control network was again under statistical threshold. For the default network, ICA approach identified residual connectivity in posterior cingulated cortex/precuneus, medial prefrontal cortex, bilateral superior frontal sulci, parahippocampal gyrus, and bilateral temporoparietal junctions. No significant anticorrelations with the default network could be identified during deep sedation, though some persisted at a lower threshold (see Supplementary Digital Content 7).
Finally, as with ROI-driven approach, ICA could identify a linear relationship between functional connectivity in key nodes of the three frontoparietal resting state networks and the level of consciousness during propofol-induced sedation (see Supplementary Digital Content 9 and 10). In particular, there was a linear relationship between thalamocortical connectivity in the default network and executive control networks and the subjects level of consciousness across sedation states. For the default network and right executive control network, ICA also identified thalamus as the maximum peak of significance for a relationship between connectivity and consciousness in these networks. Supplementary Digital Content 10 also shows a trend towards significance for a linear relationship between the level of consciousness and the strength of anticorrelations between default network and inferior frontoinsular cortices, posterior parietal cortex, and temporooccipital junction and premotor cortex. These ICA results did however not survive correction for multiple comparisons.
In line with results cited above, the goodness-of-fit scores of best-fit independent components for the bilateral executive control networks and the default network were found to be correlated to the level of consciousness across the four sedation states (see Supplementary Digital Content 13). The goodness of-fit scores can be seen as reflecting the degree to which each network independent component map matched the template. ADDIN EN.CITE Greicius200441104110411017Greicius, M. D.Srivastava, G.Reiss, A. L.Menon, V.Department of Neurology and Neurological Sciences, Stanford University School of Medicine, Stanford, CA 94305-5719, USA. greicius@stanford.eduDefault-mode network activity distinguishes Alzheimer's disease from healthy aging: evidence from functional MRIProc Natl Acad Sci U S AProc Natl Acad Sci U S A4637-4210113AdultAgedAging/*physiologyAlzheimer Disease/*physiopathologyBrain/anatomy & histology/*pathologyHumansImage Processing, Computer-AssistedMagnetic Resonance ImagingNerve Net/pathology/*physiologyReaction TimeReference ValuesReproducibility of ResultsResearch Support, Non-U.S. Gov'tResearch Support, U.S. Gov't, P.H.S.2004Mar 3015070770http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=150707702 This finding is thus in line with a diffuse alteration of connectivity patterns in the executive control networks and default network during propofol-induced anesthesia, in proportion to the extent of subjects loss of consciousness across sedation states.
Visual and Auditory Networks
ICA analyses could identify reproducible visual and auditory networks during normal wakefulness (see Supplementary Digital Content 11 and 12). As with ROI-driven approach, ICA identified a visual network of cortical areas encompassing primary visual, lingual, fusiform, cuneal, middle, and inferior occipital cortices as well as additional areas including the posterior parietal, superior temporal and precuneal cortices, and the inferior parietal lobule. For the auditory network, ICA identified a set of areas encompassing Heschls gyrus, inferior parietal lobule, and superior temporal cortices. As previously described, ADDIN EN.CITE Damoiseaux200650695069506917Damoiseaux, J. S.Rombouts, S. A.Barkhof, F.Scheltens, P.Stam, C. J.Smith, S. M.Beckmann, C. F.Department of Neurology, VU University Medical Center, De Boelelaan 1117, 1081 HV Amsterdam, The Netherlands. j.damoiseaux@vumc.nlConsistent resting-state networks across healthy subjectsProc Natl Acad Sci U S AProceedings of the National Academy of Sciences of the United States of AmericaProc Natl Acad Sci U S A13848-5310337Brain/*physiology*Brain MappingHealthHumansMagnetic Resonance ImagingRest/*physiology2006Sep 120027-8424 (Print)16945915http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=16945915 eng3 ICA was less efficient in the identification of the auditory network than of the other networks, in particular for subcortical areas like the thalamus.
During deep sedation, ICA identified a global preservation of functional connectivity within early visual and auditory cortices (see Supplementary Digital Content 11 and 12). In the visual network, the ICA approach identified preserved connectivity in a set of areas encompassing the thalamus, primary visual, lingual, fusiform, cuneal, middle, and inferior occipital cortices. In the auditory network, ICA identified residual connectivity in primary auditory cortices, though these connectivity values were just under statistical threshold after correction for multiple comparisons.
No significant relationship could be identified between the goodness-of-fit scores of visual and auditory components and the level of consciousness across sedation states see Supplementary Digital Content 13), suggesting a global preservation of connectivity patterns in these networks, despite the subjects loss of consciousness during anesthesia. No significant relationship could be identified between the level of consciousness and connectivity in early visual and auditory cortices. In the same line, no relationship could be identified between thalamocortical connectivity in visual and auditory networks and consciousness across sedation states (see Supplementary Digital Content 12).
Finally, as with the ROI-driven approach, ICA identified during wakefulness a significant temporal correlation between the activity of primary visual and primary auditory cortices. Indeed, during wakefulness, we could identify with both approaches a contribution of primary auditory cortex to the visual network maps. This cross-modal interaction between auditory and visual networks was lost during deep sedation. As with the ROI-driven approach, the strength of cross-modal functional connectivity between primary and auditory cortices as assessed by ICA showed a linear relationship with level of consciousness of our subjects during sedation (see Supplementary Digital Content 14 and 15). However, ICA results did not survive correction for multiple comparisons.
Materials and Methods
Propofol Administration and Blood Sample Data Acquisition
Arterial blood samples were taken immediately before and after scan in each clinical state for subsequent determination of the concentration of propofol and for blood gas analysis. Each blood sample for determination of the propofol plasma concentration was collected in a heparinized tube and centrifuged at 5,800 rpm for 5 min. The plasma was separated and stored at -80C. The arterial plasma propofol concentration was measured using high-pressure liquid chromatography. Each blood sample for blood gas analysis was collected in a heparinized syringe and stored in ice for 3 h before being analyzed.
Data Analysis
Preprocessing
Before functional data analysis, the first two scans were discarded from each scanning session in all subjects, allowing for T2* signal equilibration. Functional and structural images were preprocessed using Statistical Parametric Mapping software, version 5.* Functional and structural images were manually reoriented, realigned then spatially normalized. To avoid computational burden, normalized functional scans were resampled using a 4 x 4 x 4 mm voxel size.
ICA approach extraction of individual resting state networks
The second part of our study used Probabilistic ICA as implemented in MELODIC 3.0, part of Oxford Centre for Functional Magnetic Resonance Imaging of the Brain Laboratory Software Library. ICA is a statistical technique that separates a set of signals into independent (uncorrelated and non-Gaussian) spatiotemporal components. ADDIN EN.CITE Beckmann200450905090509017Beckmann, C. F.Smith, S. M.Medical Vision Laboratory, Department of Engineering Science and the Oxford Centre for Functional Magnetic Resonance Imaging of the Brain, University of Oxford, Oxford OX3 9DU, UK. beckmann@fmrib.ox.ac.ukProbabilistic independent component analysis for functional magnetic resonance imagingIEEE Trans Med ImagingIEEE transactions on medical imagingIEEE Trans Med Imaging137-52232*AlgorithmsBrain/*physiologyCerebral Cortex/physiologyHumansImage Enhancement/*methodsImage Interpretation, Computer-Assisted/*methodsMagnetic Resonance Imaging/*methods*Models, Neurological*Models, StatisticalNeurons/*physiologyPhantoms, ImagingPrincipal Component AnalysisReproducibility of ResultsSensitivity and SpecificityStochastic ProcessesVision/physiology2004Feb0278-0062 (Print)14964560http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=14964560 eng4 When applied to the T2* signal of functional magnetic resonance imaging, ICA allows not only for the removal of artifacts, ADDIN EN.CITE ADDIN EN.CITE.DATA 5-6 but also for the isolation of task-activated neural networks, ADDIN EN.CITE ADDIN EN.CITE.DATA 5 7 or of low-frequency neural networks during task-free or cognitively undemanding functional magnetic resonance imaging scans. ADDIN EN.CITE ADDIN EN.CITE.DATA 2 8-9 Before ICA, scans with excessive movement (> 1 voxel size, i.e., 3.45 mm) were removed from each session. The number of scans per session was then matched in each subject, in order to have a similar number of scans in all four clinical states (mean 250 SD 78 scans/session). Functional images were here smoothed using a 4 mm full width at half maximum Gaussian kernel, in order to allow optimal ICA decomposition without excessive smoothing. Bandpass filtering, helpful in removing high- and low-frequency noise before running ROI analyses, is probably less critical in ICA, which isolates these noise sources as independent components. ADDIN EN.CITE ADDIN EN.CITE.DATA 9-10 Given the potential risk of removing signal in addition to noise, low-pass filtering was not applied to the data used in the ICA experiments. ICA analysis was performed separately for each individual scanning session, after removal of low frequency drifts (150 s high pass filter). We allowed the software to automatically calculate the number of non-Gaussian sources (independent components) present in each session. The best-fit components for each network were selected in an automated three-step process described as the goodness-of-fit approach in. ADDIN EN.CITE ADDIN EN.CITE.DATA 2 9 11 This method allowed us to pinpoint the component for each subject that best corresponded to the default network, a left and a right executive control networks, and two purely sensory networks encompassing respectively the primary visual and primary auditory cortices. Templates used for this goodness-of-fit based component selection consisted of resting state networks identified from an independent functional magnetic resonance imaging resting state study using ICA group analysis and are displayed in Supplementary Digital Content 3. First, because intrinsic connectivity is detected in the very low frequency range, ADDIN EN.CITE Cordes200143444344434417Cordes, D.Haughton, V. M.Arfanakis, K.Carew, J. D.Turski, P. A.Moritz, C. H.Quigley, M. A.Meyerand, M. E.Department of Medical Physics, University of Wisconsin at Madison, 1300 University Ave., 1530 MSC, Madison, WI 53706, USA.Frequencies contributing to functional connectivity in the cerebral cortex in "resting-state" dataAJNR Am J NeuroradiolAJNR Am J Neuroradiol1326-33227AdultArousal/*physiologyArtifactsBrain MappingCerebral Cortex/*physiologyDominance, Cerebral/physiologyFemaleHumans*Magnetic Resonance ImagingMaleNerve Net/*physiologyReference ValuesRest2001Aug11498421http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=1149842112 a frequency filter was applied to remove any components in which high-frequency signal (>0.1 Hertz) constituted 50% or more of the power in the Fourier spectrum. In one subject, the template-matching procedure was unable to detect any low-frequency component. The deep sedation scan of this subject was therefore not included in further analyses. Next, we obtained goodness-of-fit scores for each template in the remaining low-frequency components of each subject. To do this, we used a template-matching procedure that calculates the average z-score of voxels falling within the chosen template minus the average z-score of voxels outside the template and selects the component in which this difference (the goodness-of-fit) is the greatest. Z-scores here reflect the degree to which the time series of a given voxel correlates with the time series corresponding to the specific ICA component, scaled by the standard deviation of the error term. The z-score is therefore a measure of how many standard deviations the signal is from the background noise. Finally, the component with the highest goodness-of-fit score is selected as the best-fit component and used in the subsequent group analysis. This template-matching procedure was performed separately for each network and each scanning session. It is important to note that this approach does not alter the components to fit the template in any way, but merely scores the predetermined components on how well they match the template. ADDIN EN.CITE Seeley200752715271527117Seeley, W. W.Menon, V.Schatzberg, A. F.Keller, J.Glover, G. H.Kenna, H.Reiss, A. L.Greicius, M. D.Department of Neurology, School of Medicine, University of California, San Francisco, San Francisco, California 94143, USa.Dissociable intrinsic connectivity networks for salience processing and executive controlJ NeurosciJ Neurosci2349-56279Adaptation, PhysiologicalAdultAgedFemaleGyrus Cinguli/*physiologyHumansMagnetic Resonance ImagingMaleMemory/physiologyMiddle AgedNerve Net/physiologyPrefrontal Cortex/physiologyReference ValuesThinking/*physiology2007Feb 281529-2401 (Electronic)17329432http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=17329432 eng9
ICA approach Statistical analysis
All group analyses were performed on the subjects best-fit component z-score images. We used a random-effects model that estimates the error variance across subjects, rather than across scans ADDIN EN.CITE Holmes199812331233123317Holmes, A.Friston, K.Generalisability, random effects and population inference.NeuroimageNeuroimage7547199813 and therefore provides generalization to the population from which data are acquired. It should be noted that although the best-fit components were selected with a standard template, the images have z-scores assigned to every voxel in the brain so that the group analyses were not constrained by the standard template used to select the components. ADDIN EN.CITE Seeley200752715271527117Seeley, W. W.Menon, V.Schatzberg, A. F.Keller, J.Glover, G. H.Kenna, H.Reiss, A. L.Greicius, M. D.Department of Neurology, School of Medicine, University of California, San Francisco, San Francisco, California 94143, USa.Dissociable intrinsic connectivity networks for salience processing and executive controlJ NeurosciJ Neurosci2349-56279Adaptation, PhysiologicalAdultAgedFemaleGyrus Cinguli/*physiologyHumansMagnetic Resonance ImagingMaleMemory/physiologyMiddle AgedNerve Net/physiologyPrefrontal Cortex/physiologyReference ValuesThinking/*physiology2007Feb 281529-2401 (Electronic)17329432http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=17329432 eng9 The Statistical Parametric Mapping random effects group analysis consisted in a repeated measures ANOVA with the four states of consciousness as the factor. The error covariance was not assumed to be independent between regressors and a correction for nonsphericity was applied. After model estimation, contrasts images were computed in a similar manner to the ROI-driven approach, and similar thresholds were used on the outputs of the analyses.
References
Uddin LQ, Clare Kelly AM, Biswal BB, Xavier Castellanos F, Milham MP: Functional connectivity of default mode network components: Correlation, anticorrelation, and causality. Hum Brain Mapp 2009; 30:625-37
Greicius MD, Srivastava G, Reiss AL, Menon V: Default-mode network activity distinguishes Alzheimer's disease from healthy aging: Evidence from functional MRI. Proc Natl Acad Sci U S A 2004; 101:4637-42
Damoiseaux JS, Rombouts SA, Barkhof F, Scheltens P, Stam CJ, Smith SM, Beckmann CF: Consistent resting-state networks across healthy subjects. Proc Natl Acad Sci U S A 2006; 103:13848-53
Beckmann CF, Smith SM: Probabilistic independent component analysis for functional magnetic resonance imaging. IEEE Trans Med Imaging 2004; 23:137-52
McKeown MJ, Makeig S, Brown GG, Jung TP, Kindermann SS, Bell AJ, Sejnowski TJ: Analysis of fMRI data by blind separation into independent spatial components. Hum Brain Mapp 1998; 6:160-88
Quigley MA, Haughton VM, Carew J, Cordes D, Moritz CH, Meyerand ME: Comparison of independent component analysis and conventional hypothesis-driven analysis for clinical functional MR image processing. AJNR Am J Neuroradiol 2002; 23:49-58
Calhoun VD, Pekar JJ, McGinty VB, Adali T, Watson TD, Pearlson GD: Different activation dynamics in multiple neural systems during simulated driving. Hum Brain Mapp 2002; 16:158-67
Beckmann CF, DeLuca M, Devlin JT, Smith SM: Investigations into resting-state connectivity using independent component analysis. Philos Trans R Soc Lond B Biol Sci 2005; 360:1001-13
Seeley WW, Menon V, Schatzberg AF, Keller J, Glover GH, Kenna H, Reiss AL, Greicius MD: Dissociable intrinsic connectivity networks for salience processing and executive control. J Neurosci 2007; 27:2349-56
De Luca M, Beckmann CF, De Stefano N, Matthews PM, Smith SM: fMRI resting state networks define distinct modes of long-distance interactions in the human brain. Neuroimage 2006; 29:1359-67
Greicius MD, Kiviniemi V, Tervonen O, Vainionpaa V, Alahuhta S, Reiss AL, Menon V: Persistent default-mode network connectivity during light sedation. Hum Brain Mapp 2008; 29:839-47
Cordes D, Haughton VM, Arfanakis K, Carew JD, Turski PA, Moritz CH, Quigley MA, Meyerand ME: Frequencies contributing to functional connectivity in the cerebral cortex in "resting-state" data. AJNR Am J Neuroradiol 2001; 22:1326-33
Holmes A, Friston K: Generalisability, random effects and population inference. Neuroimage 1998; 7:754
* HYPERLINK "http://www.fil.ion.ucl.ac.uk" http://www.fil.ion.ucl.ac.uk. Last accessed on June 16, 2010.
HYPERLINK "http://www.fmrib.ox.ac.uk/fsl" http://www.fmrib.ox.ac.uk/fsl. Last accessed on June 16, 2010.
PAGE
PAGE 9
H I < a 4 9 ^ _ T t ȽȲȣ}}g}ȣȲȲȲȣX hm hvb B*mH ph sH *hm hvb B*H*]mH nH ph sH u )j hm hvb B*U]mH ph sH hm hvb B*]mH ph sH hm hvb B*mH ph sH h hvb mH sH hvb hvb mH sH hm hvb mH sH hm hvb 5hm hvb 5\ hm hvb 5\mH sH hm hvb 5mH sH ! H I ! ! " + - 0 3 3 3 3 3 :6 ;6 I6 W6 )8 $a$ $a$ $d `a$ $a$ $
p# d a$ $d a$ v aw zw * ! ! " " h" " r$ s$ * * * * , ;, . . i0 0 3 ?3 3 3 3 3 ;6 ̺謢{o^ hm hvb 56\]mH sH hm hvb 5mH sH hm hvb H*mH nH sH u j hm hvb UmH sH hm hvb 5hm hvb 5\ hm hvb 5\mH sH "hm hvb H*aJ mH nH sH u !j hm hvb UaJ mH sH h hvb mH sH hm hvb mH sH hm hvb aJ mH sH $;6 I6 [7 \7 )8 *8 i8 8 8 8 %9 &9 9 9 A A A A B B *B +B ?B @B BB EB FB B B B B B B B B B -C .C =C >C RC SC UC 籡珱}k#j hm hvb UmH sH #j hm hvb UmH sH #j hm hvb UmH sH hm hvb H*mH nH sH u j hm hvb UmH sH hvb hvb mH sH hm hvb B*mH ph sH hm hvb 0J mH sH hm hvb mH sH hm hvb 6]mH sH *)8 *8 i8 ^ ^ _ l l l l m n Bo o p q 8r r s ! $
&