Multimodal Fingerprints of Resting State Networks as assessed by Simultaneous Trimodal MR-PET-EEG Imaging
by Shah NJ1,2,3,4,5, Rajkumar R1,3,6, Neuner I1,3,6*
1Institute of Neuroscience and Medicine 4, INM-4, Forschungszentrum Jülich, Germany
2Institute of Neuroscience and Medicine 11, INM-11, Forschungszentrum Jülich, Germany
3JARA – BRAIN – Translational Medicine, Germany
4Department of Neurology, RWTH Aachen University, Germany
5Monash Biomedical Imaging, School of Psychological Sciences, Monash University, Melbourne, Victoria, Australia.
6Department of Psychiatry, Psychotherapy and Psychosomatics,RWTH Aachen University, Germany
*corresponding author: Professor Dr. Irene Neuner, Institute for Neuroscience and Medicine 4, INM4, Forschungszentrum Jülich GmbH, 52425 Jülich, Germany (i.neuner@fz-juelich.de, phone +49 2461 616356, fax +49 2461 611919)
Short abstract
Recent efforts have seen advances in hybrid imaging, i.e. simultaneous acquisition of data from magnetic resonance imaging – electroencephalography (MRI-EEG) (Mullinger et al. 2011; Neuner et al. 2013) or MRI-positron emission tomography (PET) (Wehrl et al. 2013; Shah et al. 2013). In this work, we show the implementation of simultaneous trimodal imaging by employing the benefits of EEG, to acquire the electrophysiology of the brain, simultaneously with PET, to ascertain metabolic details, and MRI, to integrate brain function and structure. Trimodal imaging methodology is presented here for the first time, and we have carried out a pilot study to highlight its advantages. This article is based on recently published work, “Shah NJ, Arrubla J, Rajkumar R, Farrher E, Mauler J, Kops ER, et al. Multimodal Fingerprints of Resting State Networks as assessed by Simultaneous Trimodal MR-PET-EEG Imaging. Scientific Reports 2017;7:6452. doi:10.1038/s41598-017-05484-w”.
Introduction
The human brain is one of the most complex and efficient organs in the body. It can be functionally segregated into various functional networks (Shirer et al. 2012). One such functional network is the resting state network (RSN), which organises the brain in a large-scale cerebral network, in the absence of any external stimulation (Biswal et al. 1995). Among RSNs, the so called default mode network (DMN) is widely studied and its hubs are found to be most vulnerable to neurological disorders (Chételat & Marine 2013). It has also been found that the functional connectivity within DMN has a high impact on task performance (Berkovich-Ohana et al. 2016). The energy metabolism of the DMN and its relationship with the concentration of the neurotransmitters, as well as its electrophysiological signatures, could be potential biomarkers in the early detection of neuro-disorders. To date, such parameters have been studied using neuroimaging techniques via sequential measurements. However, sequential measurement has the major confounding factor that the data are recorded at different time points and the physiological condition of the brain might have altered between the different time points. Also, from a clinical routine point of view, sequential measurement is time consuming and requires more human resources. Thus, in order to simultaneously measure structural and functional information via magnetic resonance imaging (MRI), metabolic information via positron emission tomography (PET) and electrophysiological information via electroencephalography (EEG), the modalities of MRI, PET and EEG have been combined into a single trimodal imaging facility in our Institute. The simultaneous trimodal facility provides high spatial resolution MR images, highly molecular specific PET images (depending on the radiolabelled tracer used) and high temporal resolution EEG signals. By utilising the complementary information provided by each modality in this novel imaging facility, we investigated the DMN region of the brain and characterised DMN in healthy male subjects using multimodal fingerprints, such as functional connectivity via fMRI, energy metabolism via 2- [18F]fluoro-2-desoxy-D-glucose PET (FDG-PET), mean diffusivity (MD) via diffusion weighted imaging (DWI), the inhibition – excitation balance of neuronal activation via MR spectroscopy (MRS), and electrophysiological signature via EEG.
Methods
Data Acquisition
In a single imaging session per subject, MRI, FDG-PET and EEG data were recorded simultaneously from 11 healthy male volunteers (age 28.6 ± 3.4) using a 3T hybrid MR-BrainPET system (Siemens, Erlangen, Germany). A 32-channel, MR-compatible EEG system from Brain Products GmbH (Gilching, Germany) was used for EEG data acquisition. The trimodal data acquisition is shown in Fig. 1.
Figure 1: Trimodal set-up. The diagram displays the connections between the various components of the simultaneous MR-PET-EEG setup. The components inside the red dotted, rectangular box are inside the RF-shielded MR room. (Shah NJ, et. al., 2017)
The subjects were instructed to fast overnight and to skip breakfast on the day of measurement. First, the subjects were prepared for EEG recording. An intravenous (IV) line was inserted in the right arm of the subject to facilitate injection of the FDG tracer. FDG tracer (~200 MBq) was injected as a single bolus into the subject, while lying in supine position in the scanner. Dynamic PET data recording in list mode started immediately after the injection of the tracer and lasted 60 minutes. Simultaneously, MR structural data and MRS data (in posterior cingulate cortex (PCC), medial prefrontal cortex (MPfC), and precuneus) were recorded. Exactly 50 minutes after the FDG tracer injection, eyes closed resting state fMRI (T2*-weighted echo planar imaging) was measured for about 6 minutes. Simultaneously during resting state fMRI, the EEG data were also recorded using BrainVision Recorder (Brain Products GmbH, Gilching, Germany).
Data Analysis
MRS: The GABA ratio (to Cr+PCr) and the glutamate-glutamine ratio (to Cr+PCr) were extracted for each of the three investigated voxels (PCC, MPfC and precuneus) and used for posterior analyses.
fMRI: Data was pre-processed for motion correction, brain extraction, spatial smoothing (6 mm FWHM) and high pass filtering (100 s). DMN regions were identified for every subject via temporal concatenation ICA. A group DMN mask was created using the mean of the identified DMN regions in each subject. Similarly, a non-DMN mask was created by subtracting the DMN mask from the whole brain mask. Additionally, for functional network level comparison, a mask of the dorsal DMN (dDMN) and the sensorimotor network (SMN) was obtained from the 90 fROI atlas (Shirer et al. 2012). All four masks (DMN, non-DMN, dDMN and SMN) were corrected for grey matter. After pre-processing and intensity normalisation (grand mean scaling), the mean BOLD signal intensity was extracted individually from the created masks for each subject.
DWI: Diffusion data were corrected for eddy-current and motion distortions. The MD maps (Basser & Pierpaoli 1996) were calculated for further analysis.
PET: PET data reconstructed between 30 and 60 minutes after the injection of the FDG tracer was used for the analysis. The PET images were converted to standard uptake value (SUV) maps, accounting for the body weight and injected dose of every subject.
EEG: EEG data were partly pre-processed using a BrainVision Analyzer 2 (Brain Products, Germany). The EEG raw data were corrected for artefacts (gradient, ocular, cardioballistic). The de-noised data was exported to the LORETA-KEY software. The distribution of the neuro-electrical generators was computed using eLORETA (Pascual-Marqui et al. 1994) at different EEG frequency bands (δ ,θ, α and β).
The mean MD, SUV and neuro-electrical generators (from eLORETA) values were extracted from the created masks. A Wilcoxon-Mann-Whitney (WMW) exact test was performed using the SAS software (version 9.4) to compare the mean value of each parameter in the DMN and the non-DMN, as well as in the dDMN and the SMN. Additionally, bivariate correlation tests were performed between calculated parameters using the Spearman rank-order correlation coefficient.
Results
fMRI: The WMW test showed that the mean BOLD signal in the DMN mask was higher than outside the DMN mask (Z = 3.94, two sided p < 0.001). Similarly the mean BOLD signal in the dDMN mask was higher than in the SMN mask (Z= 3.94, two sided p < 0.0001).
DWI: No significant differences were found between the mean MD in the DMN and in non DMN (Z = 1.44, two sided p= 0.15) as well as in dDMN mask and in SMN mask (Z = 0.06, two sided p = 0.95)
PET: The WMW test showed that the SUV in the DMN mask was higher than outside the DMN mask (Z = 3.94, p < 0.0001). Similarly, the SUV in the dDMN mask and in SMN mask (Z = 3.81, two sided p < 0.0001) showed a significant difference.
EEG: The WMW test showed a significant difference in the electrical sources between the DMN and SMN in δ (Z= 3.35, two sided p = 0.0003) , θ (Z = 3.22, two sided p = 0.0006), α (Z = 3.09, two sided p = 0.001) and β–1 (Z = 2.76, two sided p ≤ 0.0041) frequency bands. Such differences were not found in the β–2 (Z = 0.2, two sided p = 0.85) and β–3 (Z = 0.98, two sided p = 0.33) frequency bands.
The Spearman rank-order correlation did not show any statistically significant correlation between the MRS parameters (the GABA ratio and the glutamate-glutamine ratio) and any of the other measured imaging parameters, such as BOLD intensity, MD, SUV and electrical generators.
The Spearman rank-order correlation shows a statistically significant positive correlation between BOLD signal intensity and SUV of FDG in DMN (rs = 0.77, n = 11, p = 0.0053) and dDMN (rs = 0.71, n = 11, p = 0.0146) (Fig. 2).
Figure 2: Correlation plot between normalised SUV uptake of FDG PET and BOLD signal intensity in DMN (right side) and dDMN (left side). (Shah NJ, et. al., 2017)
Discussion and Conclusions
Our aim of simultaneously measuring MRI, PET and EEG was successfully developed and implemented for the first time. This exploratory study, using FDG as PET tracer for the trimodal study, has demonstrated the feasibility of measuring three modalities simultaneously. Also, the significant differences observed in calculated parameters (BOLD intensity, SUV, electrical generators) between DMN and non-DMN regions during resting state confirm the reliability of the results.
Higher correlation between resting state BOLD intensity and metabolic activity of glucose (accessed via FDG-PET) in the DMN shows that higher neuronal activation is coupled to a higher energy consumption, which is in agreement with a study by Riedl et al. (Riedl et al. 2014). A significant difference was found between the DMN and SMN when comparing the electrical sources for δ ,θ, α and β-1. These frequency ranges play an important role in the long-range synchronisation for the effective coupling between more remote brain regions (Uhlhaas & Singer 2015).
This explorative pilot study in humans has the limitations of being based only on a small sample size and on the limited number of parameters assessed. However, the successful implementation of the trimodal simultaneous approach is demonstrated here in a general context.
In addition to providing insight into basic neuroscience questions addressing neurovascular-metabolic coupling, this trimodal methodology lays the foundation for investigating individual physiological and pathological fingerprints/biomarkers. These then have the potential for supporting a wide research field addressing, for example, healthy aging, gender effects, plasticity and various diseases. In particular, studies addressing pharmacological challenges will profit from this approach which paves the way for the possibility to predict individual treatment response in the framework of individually tailored medication. This would have significant benefits for the treatment of psychiatric and neurological disorders ranging from major depression and schizophrenia to epilepsy and neurodegenerative diseases.
Acknowledgements
This project is funded in part through the EU FP7 project TRIMAGE (Grant number 602621). We thank Mrs. Claire Rick for proofreading the article.
References
[1] Basser, P.J. & Pierpaoli, C., 1996.
Microstructural and Physiological Features of Tissues Elucidated by Quantitative-Diffusion-Tensor MRI.
J. Magn. Reson., Ser. B, 111, pp.209–219.[2] Berkovich-Ohana, A. et al., 2016.
Data for default network reduced functional connectivity in meditators, negatively correlated with meditation expertise.
Data in Brief, 8, pp.910–914.[3] Biswal, B. et al., 1995.
Functional connectivity in the motor cortex of resting human brain using echo-planar MRI.
Magnetic resonance in medicine, 34(4), pp.537–41.[4] Chételat, G. & Marine, 2013.
Neuroimaging biomarkers for Alzheimer’s disease in asymptomatic APOE4 carriers.
Revue Neurologique, 169(10), pp.729–36.[5] Mullinger, K.J., Yan, W.X. & Bowtell, R., 2011.
Reducing the gradient artefact in simultaneous EEG-fMRI by adjusting the subject’s axial position.
NeuroImage.[6] Neuner, I. et al., 2013.
Simultaneous EEG-fMRI acquisition at low, high and ultra-high magnetic fields up to 9.4T: Perspectives and challenges.
NeuroImage.[7] Pascual-Marqui, R.D. et al., 1994.
Low resolution electromagnetic tomography: a new method for localizing electrical activity in the brain.
Int J Psychophysiol, 18(1), pp.49–65.[8] Riedl, V. et al., 2014.
Local activity determines functional connectivity in the resting human brain: a simultaneous FDG-PET/fMRI study.
J Neurosci, 34(18), pp.6260–6.[9] Shah, N.J. et al., 2013.
Advances in multimodal neuroimaging: Hybrid MR-PET and MR-PET-EEG at 3 T and 9.4 T.
J Magn Reson, 229, pp.101–15.[10] Shirer WR, Ryali S, Rykhlevskaia E, Menon V, Greicius MD, 2012.
Decoding subject-driven cognitive states with whole-brain connectivity patterns.
Cereb. cortex, 22(1), pp.158–165.[11] Uhlhaas, P.J. & Singer, W., 2015.
Oscillations and neuronal dynamics in schizophrenia: The search for basic symptoms and translational opportunities.
Biol. Psychiatry, 77(12), pp.1001–1009.[12] Wehrl, H.F. et al., 2013.
Simultaneous PET-MRI reveals brain function in activated and resting state on metabolic , hemodynamic and multiple temporal scales.
Nature Medicine, 19(9), pp.1184–1189.