3590
Non-Stationarity of Resting-State Connectivity in Patients with Brain Tumors in the Awake and Anesthetized State1Dept. of Population and Quantitative Health Sciences, Case Western Reserve University, Cleveland, OH, United States, 2Center for Magnetic Resonance Research, Radiology, University of Minnesota, Minneapolis, MN, United States, 3Dept. of Medicine, Div. of Hematology/Oncology, University of Pittsburgh, Pittsburgh, PA, United States, 4Dept. of Neurosurgery, University of Minnesota, Minneapolis, MN, United States, 5Univ. of New Mexico, Dept. of Neurology, Albuquerque, NM, United States, 6Physics and Astronomy, University of New Mexico, Albuquerque, NM, United States
Synopsis
Keywords: Tumors (Pre-Treatment), fMRI (resting state), connectivity dynamics, temporal autocorrelation, Intra-operative
Motivation: Map temporal fluctuations of functional connectivity (FC) in anesthetized brain tumor patients.
Goal(s): Map static FC (sFC), dynamic inter-region FC (dFC), and test-retest reliability between awake and anesthetized states in patients undergoing resection of brain tumors.
Approach: A sliding-window xDF method was developed to estimate variance of the correlation in spatial-temporal resolution resting-state fMRI, considering nonstationary autocorrelation and cross-correlation.
Results: The largest decrease in sFC during anesthesia was observed across, rather than within, networks. The sliding-window xDF increased sensitivity compared to the static model. Test-retest reliability between cortical areas was higher during anesthesia versus awake state, in contrast to subcortical and cortical-subcortical dFC.
Impact: These results demonstrate the feasibility of performing resting-state functional connectivity studies in intraoperative settings with high spatial-temporal resolution. The higher test-retest reliability within cortical areas during anesthesia versus awake state informs the design of minimum duration intra-operative resting-state fMRI protocols.
INTRODUCTION
Intra-operative resting-state fMRI (rsfMRI) has been used to map functional connectivity (FC) and resting-state networks (RSNs). The feasibility of mapping low-level RSNs (motor, sensory, language) under anesthesia, despite decreased BOLD contrast and brain shifts, has been documented1-4. However, for intra-operative rsfMRI under anesthesia, non-stationarity and dynamic changes in FC, which are prominent in the awake rsfMRI5-7, related to switching between distinct brain states8,9 and differ in clinical populations 10,11, can decrease the sensitivity and specificity of mapping RSNs. In this study, we characterize the difference in static FC (sFC), dynamic inter-region FC (dFC), and test-retest reliability between the awake and the anesthesia in patients undergoing resection of brain tumors using a sliding-window statistics corrected for effective degrees of freedom.METHODS
An IRB-approved HCP-like rsfMRI acquisition protocol (modified for the intra-operative 3T scanner with 20 channel head coil) was used to study eight patients with brain tumors. Four 5-minute runs of MB6-EPI (TR/TE: 900/32.2 ms, 2mm isotropic voxels) were acquired before and after propofol-induced anesthesia.Analyses of sFC were carried out using BrainVoyager and Matlab using band-pass filtering (.01 - .1 Hz), spatial normalization into Talaraich space, and seeds using the Stanford atlas12. After randomly splitting the data into 2 halves, (2 runs per split), standard seed-based FC analyses were carried out to derive 13 FC maps per split and sedation state. sFC was computed by iteratively correlating the maps across splits to produce 2 connectivity matrices (one per sedation state) with a meaningful diagonal. A 95% bootstrap confidence interval and a paired sample t-test were used.
Analyses of dFC in 5 of the patients were carried out with TurboFIRE software13 and Matlab, using functional anatomical segmentation in subject space into 72 left- and 72 right-lateralized seeds using the Talairach Daemon database, extraction of ROI averaged time course data, regression of 6 motion parameters, CSF and white matter signals, and high-pass filtering using sin-cosine detrending up to 5th order.
A sliding-window xDF method14, which considers the nonstationary autocorrelation function (ACF) and the cross-correlation function (XCF), was developed to compute corrected Z-scores based on Pearson’s inter-ROI correlations and variance estimation. The impact of dynamic (sliding window 30 s) and static (using all data within scan) ACF and XCF estimates on the variance of dynamic correlation and inference were assessed. The dFC in awake and anesthetized state were compared using a mixed model. The dFC maps are displayed with zeroed diagonals. Test-retest reliability, the Intra-Class Correlation (ICC), was calculated using nlme, the mixed model package, with the random intercept model in R (www.r-project.org).
RESULTS
The sFC analysis showed that that: (a) sFC was generally larger during the awake state than the anesthetized state (p<.05) (Fig. 1a,b); (b) the largest difference in sFC before and after anesthesia was observed across, rather than within, networks (Fig. 1c,d).Dynamic estimates of ACF and XCF show temporal fluctuation patterns that were consistent across different lags (Fig. 2a). The dynamic XCF estimates were highly correlated with the dynamic correlation, which is the XCF at lag 0. Fig. 2b shows the impact of ACF and XCF estimates on variance and inference. The variance considering the dynamic ACF and XCF estimates was smaller than the static one, leading to a narrower confidence interval. The dynamic Z-score were less prone to false positives compared to the uncorrected Z-scores and more sensitive than the static Z-score for detecting dFC changes.
Fig. 3 shows the mean inter-regional dFC of 4 scans during awake state and 4 scans during anesthesia in a single patient. The magnitude of dFC was overall larger during the awake than the anesthetized state, with considerable variability between scans. Group- and scan-averaged inter-regional dFC was larger during the awake state compared to the anesthetized state (Fig. 4a), consistent with sFC (Fig. 1) and the largest difference was observed within BA01-47 (Fig. 4b). However, test-retest reliability within BA01-47 was on average higher during anesthesia (0.51+/-0.35) compared to awake state (0.46+/-0.32) (Fig. 5). Within the subcortex and between the subcortex and BA01-47, test-retest reliability was higher during awake compared to anesthetized state.
DISCUSSION
The enhanced sensitivity of this sliding-window xDF approach enabled detection of surprisingly high test-retest reliability between cortical areas during anesthesia. Further analyses are in progress to assess peri-tumoral dFC where impaired neurovascular coupling may reduce rsfMRI sensitivity. The results of this study are expected to inform the design of intra-operative resting-state fMRI protocols that minimize scan time while maintaining sensitivity for mapping peri-tumoral dFC.CONCLUSIONS
These preliminary results demonstrate the feasibility of performing resting-state functional connectivity studies in intraoperative settings with high spatial-temporal resolution.Acknowledgements
Supported in part by NINDS (R42NS134505) and NSF (DRL 1561716).References
[1] H. Metwali, M. Raemaekers, K. Kniese, and A. Samii, "Intraoperative Resting-State Functional Connectivity and Resting-State Networks in Patients with Intracerebral Lesions: Detectability and Variations Between Sessions," World Neurosurg, vol. 133, pp. e197-e204, Jan 2020.
[2] T. M. Qiu, F. Y. Gong, X. Gong, J. S. Wu, C. P. Lin, B. B. Biswal, D. X. Zhuang, C. J. Yao, X. L. Zhang, J. F. Lu, F. P. Zhu, Y. Mao, and L. F. Zhou, "Real-Time Motor Cortex Mapping for the Safe Resection of Glioma: An Intraoperative Resting-State fMRI Study," AJNR Am J Neuroradiol, vol. 38, pp. 2146-2152, Nov 2017.
[3] C. Roder, E. Charyasz-Leks, M. Breitkopf, K. Decker, U. Ernemann, U. Klose, M. Tatagiba, and S. Bisdas, "Resting-state functional MRI in an intraoperative MRI setting: proof of feasibility and correlation to clinical outcome of patients," J Neurosurg, vol. 125, pp. 401-409, Aug 2016.
[4] A. K. Yamamoto, J. Magerkurth, L. Mancini, M. J. White, A. Miserocchi, A. W. McEvoy, I. Appleby, C. Micallef, J. S. Thornton, C. J. Price, N. Weiskopf, and T. A. Yousry, "Acquisition of sensorimotor fMRI under general anaesthesia: Assessment of feasibility, the BOLD response and clinical utility," Neuroimage Clin, vol. 23, p. 101923, 2019, PMC6699415.
[5] C. Chang and G. H. Glover, "Time-frequency dynamics of resting-state brain connectivity measured with fMRI," Neuroimage, vol. 50, pp. 81-98, Mar 2010, 2827259.
[6] V. Kiviniemi, T. Vire, J. Remes, A. Abou elseoud, T. Starck, O. Tervonen, and J. Nikkinen, "A Sliding Time-Window ICA Reveals Spatial Variability of the Default Mode Network in Time," Brain Connectivity, vol. 1, pp. 339-347, 2011.
[7] U. Sakoglu, G. D. Pearlson, K. A. Kiehl, Y. Wang, A. Michael, and V. D. Calhoun, "A Method for Evaluating Dynamic Functional Network Connectivity and Task-Modulation: Application to Schizophrenia," MAGMA, vol. 23, pp. 351-366, 2010, PMC pending #180300.
[8] E. A. Allen, E. Damaraju, S. M. Plis, E. B. Erhardt, T. Eichele, and V. D. Calhoun, "Tracking whole-brain connectivity dynamics in the resting state," Cereb Cortex, vol. 24, pp. 663-676, Mar 2014, 3920766.
[9] X. Liu and J. H. Duyn, "Time-varying functional network information extracted from brief instances of spontaneous brain activity," Proc Natl Acad Sci U S A, vol. 110, pp. 4392-4397, Mar 12 2013, 3600481.
[10] Y. I. Sheline and M. E. Raichle, "Resting state functional connectivity in preclinical Alzheimer's disease," Biol Psychiatry, vol. 74, pp. 340-347, Sep 1 2013, 3537262.
[11] A. Faivre, A. Rico, W. Zaaraoui, L. Crespy, F. Reuter, D. Wybrecht, E. Soulier, I. Malikova, S. Confort-Gouny, P. J. Cozzone, J. Pelletier, J. P. Ranjeva, and B. Audoin, "Assessing brain connectivity at rest is clinically relevant in early multiple sclerosis," Multiple Sclerosis, vol. 18, pp. 1251-1258, Sep 2012.
[12] W. R. Shirer, S. Ryali, E. Rykhlevskaia, V. Menon, and M. D. Greicius, "Decoding subject-driven cognitive states with whole-brain connectivity patterns," Cereb Cortex, vol. 22, pp. 158-165, Jan 2012, PMC3236795.
[13] K. Vakamudi, C. Trapp, K. Talaat, K. Gao, B. Sa De La Rocque Guimaraes, and S. Posse, "Real-Time Resting-State Functional Magnetic Resonance Imaging Using Averaged Sliding Windows with Partial Correlations and Regression of Confounding Signals," Brain Connect, vol. 10, pp. 448-463, Oct 2020, PMC7580636.
[14] S. Afyouni, S. M. Smith, and T. E. Nichols, "Effective degrees of freedom of the Pearson's correlation coefficient under autocorrelation," Neuroimage, vol. 199, pp. 609-625, May 31 2019.
Figures
