Journal of the Neurological Sciences 336 (2014) 138–145
Contents lists available at ScienceDirect
Journal of the Neurological Sciences journal homepage: www.elsevier.com/locate/jns
Epileptic discharges specifically affect intrinsic connectivity networks during absence seizures Zhiqiang Zhang a,⁎,1, Wei Liao a,b,1, Zhengge Wang a, Qiang Xu a, Fang Yang c, Dante Mantini d,e, Qing Jiao a, Lei Tian f, Yijun Liu g, Guangming Lu a,⁎ a
Department of Medical Imaging, Jinling Hospital, Nanjing University School of Medicine, Nanjing 210002, China Center for Cognition and Brain Disorders and the Affiliated Hospital, Hangzhou Normal University, Hangzhou 310015, China Department of Neurology, Jinling Hospital, Nanjing University School of Medicine, Nanjing 210002, China d Department of Experimental Psychology, University of Oxford, Oxford OX1 3UD, United Kingdom e Department of Health Sciences and Technology, ETH Zurich, Zurich 8057, Switzerland f Department of Neurosurgery, Jinling Hospital, Nanjing University School of Medicine, Nanjing 210002, China g Department of Neuroscience & Psychiatry, University of Florida, Gainesville, FL, USA b c
a r t i c l e
i n f o
Article history: Received 13 July 2013 Received in revised form 30 September 2013 Accepted 16 October 2013 Available online 24 October 2013 Keywords: Absence seizures fMRI Generalized spike and wave discharge Intrinsic connectivity network EEG Independent component analysis
a b s t r a c t Intrinsic connectivity network (ICN) technique provides a feasible way for evaluating cognitive impairments in epilepsy. This EEG–fMRI study aims to comprehensively assess the alterations of ICNs affected by generalized spike-and-wave discharge (GSWD) during absence seizure (AS). Twelve fMRI sessions with GSWD, and individually paired non-GSWD sessions were acquired from 16 patients with AS. Ten ICNs corresponding to seizure origination and cognitive processes were extracted using independent component analysis. Intra- and inter-network connectivity alterations of the ICNs were observed through comparisons between GSWD and non-GSWD sessions. Sequential correlation analysis between GSWD and the ICN time courses addressed the immediate effects of GSWD on ICNs during AS. GSWD-related increase of intra-network connectivity was found only in the thalamus, and extensive decreases were found in the ICNs corresponding to higher-order cognitive processes including the default-mode network, dorsal attention network, central executive network and salience network. The perceptive networks and motor network were less affected by GSWD. Sequential correlation analysis further demonstrated different responses of the ICNs to GSWD. In addition to GSWDrelated functional excitation in the thalamus and functional suspension in the default-mode network, this study revealed extensive inhibitions in the other ICNs corresponding to higher-order cognitive processes, and spared perceptive and motor processes in AS. GSWD elevated synchronization of brain network activity and sequentially affected the ICNs. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Typical absence epilepsy is a common type of idiopathic generalized epilepsy in childhood; it is featured by sudden, brief impairment of consciousness, accompanied by a 2.5–4 Hz generalized spike and wave discharges (GSWD). It has been proposed that the cognitive processes underlying consciousness are selectively affected by GSWD during absence seizure (AS) [1]. Using general-linear-model (GLM) and timeseries analyses, simultaneous electroencephalography (EEG) and fMRI have depicted the spatial and temporal properties of brain activation
⁎ Corresponding authors at: Department of Medical Imaging, Jinling Hospital, 305# Eastern Zhongshan Rd., Nanjing 210002, China. Tel.: +86 25 80860185; fax: +86 25 84804659. E-mail addresses: [email protected] (Z. Zhang), [email protected] (G. Lu). 1 The authors contribute equally to this work. 0022-510X/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jns.2013.10.024
during AS. GSWD-related activation in the thalamus and deactivation in the medial frontal and posterior cingulate cortices are typically reported in EEG–fMRI studies [2–4]. These brain structures have been linked to seizure generation and deficits in the default-mode of brain function [2–4], respectively. In addition, distributed deactivation in the frontal and parietal regions has been suggested impaired attention and spared motion processes in AS [5,6]. Moreover, these brain regions presented different temporal patterns responding to GSWD during the evolution course of seizures. The accumulating imaging evidence may support the proposal that multiple cognitive processes are specifically involved in AS [7,8]. However, the precise alterations of brain processes associated with GSWD and the relationship among them have yet to be thoroughly assessed. Cognitive impairments in epilepsy have been recently related to the alteration of intrinsic connectivity networks (ICNs) [9–11]. By measuring the correlation of spontaneous hemodynamic fluctuations, resting-state fMRI has been used to link ICNs to specific cognitive processes, such as self-awareness, attention, control and perceptions [12–14].
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
Thus measuring connectivity within and across ICNs may permit examining the integrity of brain circuits related to consciousness [15–17]. This study investigated the intra- and inter-network alterations of extensive ICNs in AS, and addressed temporal evolutions of GSWD effect on the ICNs using sequential correlation analysis, which was expected to contribute to a better understanding of the neural correlates of consciousness impairments in absence epilepsy.
2. Methods 2.1. Patients Sixteen patients with childhood absence epilepsy were recruited in this study (detailed in Table 1). They met the following criteria: (i) clinical diagnosis of childhood absence epilepsy based on International League Against Epilepsy criteria [18]; (ii) EEG with typical 2.5–4 Hz bilateral ictal GSWD and normal background activity; (iii) no additional seizure types, such as myoclonic, tonic–clonic, or partial seizures; (iv) no known structural brain abnormality in routine MRI and other neurological disorders. All patients had informed consent signed by their legal guardians, and all human study procedures were approved by the institutional review boards at Jinling Hospital, Nanjing University School of Medicine.
2.2. Simultaneous EEG and fMRI data acquisitions All patients successfully underwent simultaneous EEG and fMRI data acquisitions on a 32 channels MRI-compatible EEG (Brain Product, Munich, Germany) and a 3 T MRI scanner (Siemens Trio, Erlangen, Germany). The patients were instructed to keep rest and not fall in sleep. Foam pads were used to help secure the EEG leads, minimize motion, and improve patient comfort. For EEG recordings, the electrode FCz was set as the reference and electrocardiography was recorded using an electrode placed on the back. Data were transmitted via an optic fiber cable from the amplifier placed inside the scanner room to a computer outside the scanner room. For MRI data scanning, the functional data were acquired using a T2*-weighted single-shot echo planar imaging sequence (TR/TE = 2000 ms/40 ms, FA = 90°, matrix = 64 × 64, FOV = 24 × 24 cm, thickness/gap = 4.0 mm/0.4 mm, 2 sessions with each consisting of 500 volumes each, collected after five dummy volumes). Threedimensional magnetization prepared rapid acquisition gradient-echo T1 anatomical images (TR/TE = 2300 ms/2.98 ms, FA = 9°, matrix = 256 × 256, FOV = 256 × 256 mm2, and slice thickness = 1 mm) were obtained as a structural reference.
Table 1 Demographic and clinical information of involved patients. Sex/Nos.
Age, y
Onset age, y
Medication
Nos. and durations of GSWD (in s) in AS
F1 F2 F3 F4 M1 M2 F5 F6 M3 F7 M4 F8 Mean ± Std
12 6 5 11 7 10 7 8 9 18 7 10 9.1 ± 3.5
9 4 4 9 5 9 7 6 6 10 8 3 6.7 ± 2.3
None SV, LTG, LEV None None SV, LTG, LEV SV SV SV None SV SV SV, LTG /
2 (75) 2 (17) 3 (16) 1 (14) 2 (15) 3 (33) 8 (73) 1 (12) 1 (9) 1 (16) 2 (39) 1 (16) 2.2 ± 1.9 (28.8 ± 22.8)
Abbreviations: F: Female, M: male; SV: Sodium Valproate; LTG: Lamotrigine; LEV: levetriacetam; AS: Absence seizure; GSWD: Generalized spike-wave discharge.
139
2.3. Data preprocessing The EEG data was offline-processed to remove gradient and ballistocardiogram artifacts using the Brain Vision Analyzer 2.0 software. The GSWDs were marked on artifact-removed EEG by an experienced neurologists (Yang) and an electroencephalographer (Tian). The fMRI data preprocessing was performed using a software package SPM8 (http://www.fil.ion.ucl.ac.uk/spm). After slice-timing adjustment and realignment for head-motion correction, data were realigned to the corresponding anatomical images, warped into the anatomical MNI152 space using a 12-parameter affine linear transformation, resliced at a resolution of 3 × 3 × 3 mm3, and spatially smoothed using an isotropic Gaussian kernel (8 mm full width at half maximum). In order to match data segments with and without GSWD in each individual, we divided the full amount of fMRI data into four subsessions comprising 250 volumes. Accordingly, we selected pairs of data segments from 12 patients who presented seizures during scanning. Each pair consisted of a GSWD sub-session and a matched non-GSWD sub-session. The following criteria were used for data selection: (i) Each GSWD sub-session contained seizure events. Seizure event was defined to occur if the GSWD was longer than 6 s [19,20] and the events had the same duration and morphology as clinically confirmed events on the routine EEG using the International League Against Epilepsy guidelines[4]. A total of 32 events of GSWD were included. The GSWD durations across events are 11.6 ± 8.9 s, and across subsession are 29.6 ± 22.6 s. (ii) For the non-GSWD sub-sessions, no discharge occurred 18 s before or after the selected data segment. (iii) The sub-sessions containing large motion (more than 1.5 mm or 1.5°) or the ones that could not be matched in the same subject were excluded (see Supplementary Fig. 1). There were no significant differences of head motion between the two data groups (Paired t-test, t = −0.6, p = 0.52 for translation and t = −0.8, p = 0.48 for rotation). 2.4. Data analysis 2.4.1. Sequential HRF generation and dynamic GLM framework In line with previous studies [21,22], we first generated a sequence of hemodynamic response functions (HRFs) to model the dynamic BOLD changes induced by GSWD. Sequential HRFs consisted of 49 successive gamma functions of FWHM of 5.2 s, peak = 5.4 s, centered at 0 s and spaced 1 s between one another. At the individual level, we convolved a boxcar function expressing the timing of seizure events with sequential HRFs shifted between 24 s before to 24 s after the GSWD (HRF-24 to HRF + 24), to generate a series of regressors that modeled the BOLD responses. Then, we performed separate t-tests within a general-linear-model (GLM) framework for each regressor with specific HRF, thereby producing 49 t-maps representing GSWDrelated BOLD activation modeled with different HRFs. The dynamic GLM framework allows us to observe the dynamic BOLD changes before and after seizure onset [22]. For group-analysis, we used one-sample t-tests (p b 0.05, AlphaSim correction) to determine regions showing significant GSWDs-related BOLD changes. Sequential-HRF based dynamic GLM framework was mainly used as a reference for the subsequent ICA analysis. 2.4.2. Group comparison analysis for intra- and inter-network connectivity of ICNs Subsequently, we performed group independent component analysis (ICA) to examine ICNs in the patients. We used the GIFT software (version 2.0d; http://icatb.sourceforge.net/) to both fMRI sub-sessions related to GSWD and non-GSWD. Forty-four spatially independent components were decomposed after component number estimation using the minimum description length criteria [23]. The decomposition produced a set of components, each with a spatial map of intensities and a representative time course. The spatial maps were converted to zscores, which reflect the degree to which synchronous activity occurs
140
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
across spatially independent but widely distributed brain regions, i.e., the degree of functional connectivity [24]. Based on our hypothesis derived from literatures in the field of absence seizures [5,11,21,22,25] and functional brain networks [12,26], we selected ten well-identified networks [12,26,27] as components-of-interest using spatial correlation analyses. These components include the thalamic network (THA), posterior and anterior default-mode (aDMN and pDMN), dorsal attention network (DAN), central executive network (CEN), salience (SN), sensory (SEN), auditory (AN), visual (VN) and motor networks (MN). The reference image used to identify the thalamic network was obtained from our GLM analysis, whereas the other reference images were most studied and well defined in the previous studies (see Supplementary Fig. 2) [9,28–30]. Considering the small sample-size of our data, we also performed a leave-one-out (LOO) analysis to test the reproducibility of each ICN extracted by ICA (see Supplementary Fig. 3). The effect of GSWD on ICNs was examined by comparing ICNs during GSWD and non-GSWD states. To this end, we used voxel-wise paired-t tests (pb0.05, AlphaSim correction) on the ICN maps associated with the two states. This permitted investigating alterations of synchronous activation in each ICN. We restricted this comparison to predefined masks generated by calculating one-sample t-tests for GSWD and non-GSWD states separately. Moreover, we correlated activation intensities (averaged z values in the regions showing altered connectivity) with the duration of GSWD using a Spearman correlation analysis (p b 0.05, Bonferroni correction). Subsequently, we also measured functional network connectivity (FNC) (http://icatb.sourceforge.net/) to investigate the interactions among ICNs. Following this approach, we investigated the temporal relationships among ICN time courses by computing a constrained maximal lagged correlation (maximal lag = 10 s) [31]. We calculated a one-sample t-test on pair-wise correlations across subjects for the GSWD and non-GSWD states separately, and we also compared the two states using paired t-tests (p b 0.05). 2.4.3. Sequential correlation analysis between GSWD and ICNs We further investigated the effect of GSWD on the ICNs in AS using sequential HRF-based correlation analysis. The GSWD modeling signals (obtained by convolving sequential HRFs with a GSWD-based boxcar function) were sequentially correlated with the time course of each ICN [32]. By extracting the maximum among these correlations, we defined a reference time lag at which a specific ICN was most likely affected by GSWD. The correlation coefficients were transformed into z-scores using the Fisher's r-to-z transformation, and were given as input to a one-sample t-test (p b 0.05, Bonferroni correction). We also used the peak of the GSWD modeling signals averaged across subjects to define the latency of ICN maximum activation/deactivation. In addition, to explicitly demonstrate the relationship between GSWD and the time courses of ICNs, we showed the averaged time courses across all the GSWD events over a 60 s window centered at the time of the GSWD onset. All the data analysis steps are illustrated in Supplementary Fig. 4. 3. Results The GLM-based analysis with sequential HRFs revealed dynamic properties of GSWD-related BOLD activation and deactivation (Supplementary Fig. 5). Each ICN extracted by ICA showed highly reproducibility (Supplementary Fig. 3). Voxel-wise paired t-tests demonstrated that the intra-network connectivity was altered in each ICN as affected by GSWD. The components corresponding to the thalamus and the AN showed increased connectivity in the GSWD state compared to the non-GSWD state. In contrast, the aDMN, pDMN, DAN and the CEN showed decreased connectivity in GSWD state. Furthermore, GSWD durations were found to be positively correlated with the intra-network connectivity (expressed by z-values) in the thalamus, and negatively
correlated with connectivity intensities in the pDMN, DAN and CEN (Fig. 1). FNC analysis revealed that the inter-network communications among ICNs were largely altered during GSWD state. First, the number of significant connections in the GSWD state was more than the nonGSWD state (21 vs. 14). Secondly, comparison analysis showed that the thalamus, aDMN and pDMN were most affected networks, particularly in terms of the number of inter-network connections and activation time lags (Fig. 2). For the sequential correlation analyses between the GSWD modeling signals and the time course of ICNs, the thalamus showed positive correlation, while the aDMN, pDMN, DAN, CEN and SN all showed negative correlations (p b 0.05, Bonferroni correction). None of the perceptive network showed significant correlation between the time series and the modeling signals. The sequence number of the GSWDmodeling signals of each ICN was used to define the latency of ICN activation/deactivation. We obtained the following latencies for the different ICNs: THA (HRF +9, i.e., starting time at 9 s after GSWD onset), aDMN (HRF +9), pDMN (HRF +11), DAN (HRF +15), CEN (HRF +11) and SN (HRF +17). These results were consistent with those of sequential correlation analyses (Fig. 3). 4. Discussion By employing an ICA-based functional connectivity approach, this study provided a comprehensive analysis of the effects of GSWD on a set of ICNs in AS. (1) ICA and GLM analyses both revealed that, while the thalamus showed positive BOLD response to GSWD, extensive cortical regions and ICNs had negative responses to GSWD. (2) ICNs corresponding to the higher-order cognitive processes, including the CEN, DAN and SN, were severely influenced by GSWD; whereas the perceptive networks and the MN showed much less effects of GSWD in AS. (3) Temporal analysis using averaged time courses and sequential correlation analyses provided converging results, and revealed a differential pattern of activation for the ICNs affected by GSWD. In line with previous works [19,22], we used a traditional GLM framework to show the dynamic properties of GSWD-related activation in the thalamus. Furthermore, we found extensive cortical structures not only covering regions of the DMN, but also including the regions of the DAN and CEN showed dynamic deactivation. These findings may suggest origination of epileptic discharges and dysfunction of the cortical regions [22,33]. Moreover, the dynamic properties of BOLD responses implicate that different regions were involved sequentially in the initiation, maintenance and termination of AS [19,22]. Importantly, we employed ICA to investigate ICN changes during AS. Different with the GLM framework for activation localization, ICA is a data-driven technique that decomposes the whole data set into several ICNs characterized by a specific spatial pattern and does not require any prior assumption related to the HRF [34,35]. Specifically, the ICNs reflect the architecture of coherent intrinsic activity in the brain. The DMN, DAN, CEN, SN and perceptive networks have been well linked to cognitive processes of self-awareness/episodic memory, top-down attention, cognitive control, personal salience and sensory integration, respectively [12,14,36,37]. By using spatial correlation with predefined templates, it was possible to identify a set of ICNs corresponding to specific cognitive functional processes [12,14,31,36], as well as regions associated with epileptic activation [34,35]. The thalamus not only showed positive response to GSWD, but also increased intra-network connectivity in the GSWD state relative to the non-GSWD state. Conversely, other ICNs corresponding to high-order cognitive processes, including the aDMN, pDMN, CEN, DAN and SN showed negative responses to GSWD, and decreased intra-network connectivity in the GSWD states relative to the non-GSWD state. Our results concerning positive BOLD responses in the thalamus and negative responses in a large portion of the cortex during GSWD are in accordance with previous findings [19,21,22]. Moreover, the GSWD
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
141
Fig. 1. Alterations of intra-network connectivity in the GSWD sub-session revealed by group comparisons. Voxel-based comparison of each ICN (z-map) using paired t-tests between GSWD and non-GSWD sub-sessions. In the GSWD state, the thalamus network and the AN showed increased intra-network connectivity, while the aDMN, pDMN, DAN and CEN showed decreased activation relative to the non-GSWD sub-session. There was no group difference in the SN, SEN, VN and MN. Moreover, for each ICN showing altered intra-network connectivity, the activation intensities (z-scores) in the regions showing group difference were correlated with GSWD durations (Spearman correlation, p b 0.05, Bonferroni correction). Positive correlation was found in the thalamus network, and negative correlations were found in the pDMN, DAN and CEN. No significant correlation was found in the aDMN and AN.
related increased and decreased intra-network connectivity might recapitulate the functional excitation of the thalamus and functional suspensions of the cortical networks in the intrinsic brain activities [33]. In addition to those of the DMN [19,21,22] and DAN [5,6], we for the first time reported the functional inhibitions in the higher-order processes of CEN and SN during AS. The ICNs corresponding to the perceptive networks and the MN showed less alteration responsive to GSWD. The AN presented increased functional connectivity in the GSWD state, but no correlation with GSWD was found. Furthermore, none of the perceptive networks showed significant correlation between the GSWD-modeling signals and the ICN time-courses. The GLM analysis revealed minor deactivation in the regions of the sensory, visual and motor cortices. These findings
supported the proposal that the motor process for simpler task is spared in AS [1,6], and further indicated that the lower-order perceptive processes are also less affected by GSWD. FNC analysis revealed that inter-networks relationships among ICNs are altered in AS [22,27]. More inter-networks communications in the GSWD sub-session suggest a hyper-synchronization of large-scale brain networks [38,39] induced by GSWD. These results may also imply that the intrinsic processes relative to cognitive activities are hijacked by epileptic discharges. Regarding to the critical roles of the thalamus and the DMN played in the maintenance of level of consciousness [8,15,40], the largest alterations of the inter-networks communications in the thalamus and DMN suggest the crucial roles of these two structures in AS. The specific alterations in the other ICNs
142
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
Fig. 2. Alterations of inter-network connectivity in the GSWD sub-session revealed by group comparisons. FNC analysis showed the inter-network communication among these ICNs. Visual inspection revealed that the amount of inter-network connections (one-sample t-tests for each sub-session) was increased in the GSWD sub-sessions relative to the non-GSWD sub-sessions (21 vs. 14). Paired t-test comparisons (p b 0.05) between two states revealed that the thalamus network and the DMN showed the most prominent alterations.
may implicate selective impairments of cognitive processes for the different extents of consciousness [4,19]. Both sequential correlation analysis and averaged time course analysis demonstrated a direct effect of GSWD on ICNs. With either approaches, ICNs showed different temporal responses with respect to GSWD. The thalamus network had a response peak at a latency of about 14s (9s+5.4s) after GSWD onset, which was almost simultaneous to that of aDMN but went in opposite direction. This result is in line with
those of previous studies [4,19], which used a temporal analysis based on regions of interest. Other ICNs such as pDMN, CEN, DAN and SN had later response peaks after GSWD onset, while we found no significant correlation in the perceptive networks and the MN. These results were largely consistent with those revealed by GLM analysis. Specifically, the sequence of ICN response peaks may provide valuable information on the hierarchy of networks that subserving impaired cognitive processes in AS. Moreover, we also noted the inconsistent findings between
Fig. 3. Averaged ICN time courses and sequential correlation with GSWD-modeling signals. For each ICN, we averaged the time courses across all GSWD events over a 60 s window centered at the time of GSWD onset. The curve line denotes the mean, and the shade denotes std of the z values. This analysis revealed that the thalamus network had positive BOLD response (red line); the higher-order cognitive networks including the aDMN, pDMN, CEN, DAN and SN showed negative BOLD responses (blue line) to GSWD at various peak times. The inset frames indicate sequential correlation results between sequential GSWD modeling signals and the ICN time courses. The stars denote significant correlations (p b 0.05, Bonferroni correction). The thalamus showed maximal positive correlation (t = 5.70, p = 1.67 × 10−8) when using HRF + 9. The aDMN (HRF + 9, t = −5.76, p = 1.17 × 10−8), pDMN (HRF + 11, t = −7.42, p = 2.90 × 10−13), CEN (HRF + 11, t = −6.45, p = 1.85 × 10−10), DAN (HRF + 15, t = −4.84, p = 1.52 × 10−6) and SN (HRF + 17, t = −5.84, p = 7.10 × 10−9) all showed negative correlations. The perceptive network and the MN showed no significant correlation between their time courses and the modeling signals. These results were largely consistent with the sequential correlation analysis. Specifically, the peak time of the average time courses was indeed similar to the optimal time lag estimated using the sequential correlation analysis plus the peak latency of the HRF (i.e., 5.4 s).
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
143
144
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
the ICA and GLM analysis, e.g., the time point of peak response of the thalamus. We considered that utilization of specific HRF in GLM and free of prior assumption related to the HRF in ICA might be the reason. It is important to mention that a number of potential limitations apply to our study. First, resting-state fMRI has proven to be a valid tool for detecting epileptic discharges-related activation [34,35] and for better understanding cognitive impairments in epilepsy [9,11,27], but the relationship between the resting-state ICNs and task-induced activations is still partially unclear [12,13]. Future studies with more imaging and simultaneous behavioral data acquisition are needed. Secondly, we used ICA, a data-driven method for resting-state fMRI analysis, to detect a set of ICNs related to specific cognitive processes [12,14]. It is worth noting that ICA results might be influenced by model order estimation, and may be difficult to interpret due to the complexity of the output structure [13,41]. Thirdly, this study has a relatively small sample size and future studies on a large number of patients are necessary to corroborate our findings. However, the high reproducibility of ICN maps supports the robustness of our results (Supplementary Fig. 3). Finally, the effect of anti-epileptic drugs on ICNs and different types of epileptic discharges were not addressed in the present study. 5. Conclusion This study comprehensively investigated the effect of GSWD on ICNs during AS. We confirmed previous findings concerning thalamic excitation and DMN suspension. Moreover, the results revealed extensive inhibitions in the ICNs corresponding to the high-order cognitive processes, such as the DAN, CEN and SN in AS; while the ICNs corresponding to the perceptive and motor networks seemed to be only less influenced by GSWD effects. Moreover, these ICNs presented enhancement in the synchronization of the large-scale brain network activity, suggesting aberrant relationship among the cognitive processes in consciousness system in AS. Finally, the sequential effects of GSWD on different ICNs may potentially reveal different roles of these networks in AS. Overall, our findings may contribute to a better understanding of the pathophysiological mechanisms underlying cognitive impairments during AS. Conflicts of interest The authors declare that they have no conflict of interest. Acknowledgment This research was supported by the Natural Science Foundation of China (Grant nos. 81271553, 81201155, 30971019, 81020108022) and Grants for Young Scholar of Jinling Hospital (Grant nos. 2011060, 201 1045). Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.jns.2013.10.024. References [1] Aldenkamp A, Arends J. The relative influence of epileptic EEG discharges, short nonconvulsive seizures, and type of epilepsy on cognitive function. Epilepsia 2004;45(1):54–63. [2] Salek-Haddadi A, Lemieux L, Merschhemke M, Friston KJ, Duncan JS, Fish DR. Functional magnetic resonance imaging of human absence seizures. Ann Neurol 2003;53(5):663–7. [3] Moeller F, Siebner HR, Wolff S, Muhle H, Granert O, Jansen O, et al. Simultaneous EEG–fMRI in drug-naive children with newly diagnosed absence epilepsy. Epilepsia 2008;49(9):1510–9.
[4] Carney PW, Masterton RA, Harvey AS, Scheffer IE, Berkovic SF, Jackson GD. The core network in absence epilepsy. Differences in cortical and thalamic BOLD response. Neurology 2010;75(10):904–11. [5] Killory BD, Bai X, Negishi M, Vega C, Spann MN, Vestal M, et al. Impaired attention and network connectivity in childhood absence epilepsy. Neuroimage 2011;56(4):2209–17. [6] Berman R, Negishi M, Vestal M, Spann M, Chung MH, Bai X, et al. Simultaneous EEG, fMRI, and behavior in typical childhood absence seizures. Epilepsia 2010;51(10): 2011–22. [7] Blumenfeld H, Taylor J. Why do seizures cause loss of consciousness? Neuroscientist 2003;9(5):301–10. [8] Blumenfeld H. Impaired consciousness in epilepsy. Lancet Neurol 2012;11(9):814–26. [9] Zhang Z, Lu G, Zhong Y, Tan Q, Yang Z, Liao W, et al. Impaired attention network in temporal lobe epilepsy: a resting FMRI study. Neurosci Lett 2009;458(3):97–101. [10] Vlooswijk MC, Jansen JF, de Krom MC, Majoie HM, Hofman PA, Backes WH, et al. Functional MRI in chronic epilepsy: associations with cognitive impairment. Lancet Neurol 2010;9(10):1018–27. [11] Yang T, Luo C, Li Q, Guo Z, Liu L, Gong Q, et al. Altered resting-state connectivity during interictal generalized spike-wave discharges in drug-naive childhood absence epilepsy. Hum Brain Mapp 2012. http://dx.doi.org/10.1002/hbm.22025. [12] Zhang D, Raichle ME. Disease and the brain's dark energy. Nat Rev Neurol 2010;6(1): 15–28. [13] Fox MD, Raichle ME. Spontaneous fluctuations in brain activity observed with functional magnetic resonance imaging. Nat Rev Neurosci 2007;8(9):700–11. [14] Mantini D, Perrucci MG, Del Gratta C, Romani GL, Corbetta M. Electrophysiological signatures of resting state networks in the human brain. Proc Natl Acad Sci U S A 2007;104(32):13170–5. [15] Boly M, Phillips C, Tshibanda L, Vanhaudenhuyse A, Schabus M, Dang-Vu TT, et al. Intrinsic brain activity in altered states of consciousness: how conscious is the default mode of brain function? Ann N Y Acad Sci 2008;1129:119–29. [16] Vanhaudenhuyse A, Noirhomme Q, Tshibanda LJ, Bruno MA, Boveroux P, Schnakers C, et al. Default network connectivity reflects the level of consciousness in noncommunicative brain-damaged patients. Brain 2010;133(Pt 1):161–71. [17] Greicius MD, Kiviniemi V, Tervonen O, Vainionpaa V, Alahuhta S, Reiss AL, et al. Persistent default-mode network connectivity during light sedation. Hum Brain Mapp 2008;29(7):839–47. [18] Engel Jr J. A proposed diagnostic scheme for people with epileptic seizures and with epilepsy: report of the ILAE Task Force on Classification and Terminology. Epilepsia 2001;42(6):796–803. [19] Bai X, Vestal M, Berman R, Negishi M, Spann M, Vega C, et al. Dynamic time course of typical childhood absence seizures: EEG, behavior, and functional magnetic resonance imaging. J Neurosci 2010;30(17):5884–93. [20] Li Q, Luo C, Yang T, Yao Z, He L, Liu L, et al. EEG–fMRI study on the interictal and ictal generalized spike-wave discharges in patients with childhood absence epilepsy. Epilepsy Res 2009;87(2–3):160–8. [21] Moeller F, LeVan P, Muhle H, Stephani U, Dubeau F, Siniatchkin M, et al. Absence seizures: individual patterns revealed by EEG–fMRI. Epilepsia 2010;51(10):2000–10. [22] Benuzzi F, Mirandola L, Pugnaghi M, Farinelli V, Tassinari CA, Capovilla G, et al. Increased cortical BOLD signal anticipates generalized spike and wave discharges in adolescents and adults with idiopathic generalized epilepsies. Epilepsia 2012;53 (4):622–30. [23] Li YO, Adali T, Calhoun VD. Estimating the number of independent components for functional magnetic resonance imaging data. Hum Brain Mapp 2007;28(11): 1251–66. [24] Calhoun VD, Kiehl KA, Liddle PF, Pearlson GD. Aberrant localization of synchronous hemodynamic activity in auditory cortex reliably characterizes schizophrenia. Biol Psychiatry 2004;55(8):842–9. [25] Luo C, Li Q, Lai Y, Xia Y, Qin Y, Liao W, et al. Altered functional connectivity in default mode network in absence epilepsy: a resting-state fMRI study. Hum Brain Mapp 2011;32(3):438–49. [26] Moeller F, Maneshi M, Pittau F, Gholipour T, Bellec P, Dubeau F, et al. Functional connectivity in patients with idiopathic generalized epilepsy. Epilepsia 2011;52(3):515–22. [27] Luo C, Li Q, Xia Y, Lei X, Xue K, Yao Z, et al. Resting state basal ganglia network in idiopathic generalized epilepsy. Hum Brain Mapp 2011. http://dx.doi.org/10.1002/ hbm.21286. [28] Wang Z, Lu G, Zhang Z, Zhong Y, Jiao Q, Tan Q, et al. Altered resting state networks in epileptic patients with generalized tonic–clonic seizures. Brain Res 2011;1374:134–41. [29] Liao W, Chen H, Feng Y, Mantini D, Gentili C, Pan Z, et al. Selective aberrant functional connectivity of resting state networks in social anxiety disorder. Neuroimage 2010;52(4):1549–58. [30] Allen EA, Damaraju E, Plis SM, Erhardt EB, Eichele T, Calhoun VD. Tracking whole-brain connectivity dynamics in the resting state. Cereb Cortex 2012. http://dx.doi.org/10.1093/cercor/bhs352. [31] Vlassenko AG, Rundle MM, Raichle ME, Mintun MA. Regulation of blood flow in activated human brain by cytosolic NADH/NAD+ ratio. Proc Natl Acad Sci U S A 2006;103(6):1964–9. [32] Mantini D, Corbetta M, Perrucci MG, Romani GL, Del Gratta C. Large-scale brain networks account for sustained and transient activity during target detection. Neuroimage 2009;44(1):265–74. [33] Gotman J, Grova C, Bagshaw A, Kobayashi E, Aghakhani Y, Dubeau F. Generalized epileptic discharges show thalamocortical activation and suspension of the default state of the brain. Proc Natl Acad Sci U S A 2005;102(42):15236–40. [34] Rodionov R, De Martino F, Laufs H, Carmichael DW, Formisano E, Walker M, et al. Independent component analysis of interictal fMRI in focal epilepsy: comparison
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
[35]
[36]
[37]
[38]
[39] [40] [41]
with general linear model-based EEG-correlated fMRI. Neuroimage 2007;38(3):488–500. LeVan P, Tyvaert L, Moeller F, Gotman J. Independent component analysis reveals dynamic ictal BOLD responses in EEG–fMRI data from focal epilepsy patients. Neuroimage 2010;49(1):366–78. Damoiseaux JS, Rombouts SA, Barkhof F, Scheltens P, Stam CJ, Smith SM, et al. Consistent resting-state networks across healthy subjects. Proc Natl Acad Sci U S A 2006;103(37):13848–53. Seeley WW, Menon V, Schatzberg AF, Keller J, Glover GH, Kenna H, et al. Dissociable intrinsic connectivity networks for salience processing and executive control. J Neurosci 2007;27(9):2349–56. Meeren H, van Luijtelaar G, Lopes da Silva F, Coenen A. Evolving concepts on the pathophysiology of absence seizures: the cortical focus theory. Arch Neurol 2005;62(3):371–6. Snead III OC. Basic mechanisms of generalized absence seizures. Ann Neurol 1995;37(2):146–57. Blumenfeld H. Epilepsy and the consciousness system: transient vegetative state? Neurol Clin 2011;29(4):801–23. Fox MD, Greicius M. Clinical applications of resting state functional connectivity. Front Syst Neurosci 2010;4:19.
Glossary AN: auditory network AS: absence seizures BOLD: blood oxygen level-dependent CAE: childhood absence epilepsy CEN: central executive network DAN: dorsal attention network DMN: default-mode network FNC: functional network connectivity GLM: general-linear-model GSWD: generalized spike and wave discharge HRF: hemodynamic response function ICA: independent component analysis ICN: intrinsic connectivity networks MN: motor network SEN: sensory network SN: salience network VN: visual network
145
Contents lists available at ScienceDirect
Journal of the Neurological Sciences journal homepage: www.elsevier.com/locate/jns
Epileptic discharges specifically affect intrinsic connectivity networks during absence seizures Zhiqiang Zhang a,⁎,1, Wei Liao a,b,1, Zhengge Wang a, Qiang Xu a, Fang Yang c, Dante Mantini d,e, Qing Jiao a, Lei Tian f, Yijun Liu g, Guangming Lu a,⁎ a
Department of Medical Imaging, Jinling Hospital, Nanjing University School of Medicine, Nanjing 210002, China Center for Cognition and Brain Disorders and the Affiliated Hospital, Hangzhou Normal University, Hangzhou 310015, China Department of Neurology, Jinling Hospital, Nanjing University School of Medicine, Nanjing 210002, China d Department of Experimental Psychology, University of Oxford, Oxford OX1 3UD, United Kingdom e Department of Health Sciences and Technology, ETH Zurich, Zurich 8057, Switzerland f Department of Neurosurgery, Jinling Hospital, Nanjing University School of Medicine, Nanjing 210002, China g Department of Neuroscience & Psychiatry, University of Florida, Gainesville, FL, USA b c
a r t i c l e
i n f o
Article history: Received 13 July 2013 Received in revised form 30 September 2013 Accepted 16 October 2013 Available online 24 October 2013 Keywords: Absence seizures fMRI Generalized spike and wave discharge Intrinsic connectivity network EEG Independent component analysis
a b s t r a c t Intrinsic connectivity network (ICN) technique provides a feasible way for evaluating cognitive impairments in epilepsy. This EEG–fMRI study aims to comprehensively assess the alterations of ICNs affected by generalized spike-and-wave discharge (GSWD) during absence seizure (AS). Twelve fMRI sessions with GSWD, and individually paired non-GSWD sessions were acquired from 16 patients with AS. Ten ICNs corresponding to seizure origination and cognitive processes were extracted using independent component analysis. Intra- and inter-network connectivity alterations of the ICNs were observed through comparisons between GSWD and non-GSWD sessions. Sequential correlation analysis between GSWD and the ICN time courses addressed the immediate effects of GSWD on ICNs during AS. GSWD-related increase of intra-network connectivity was found only in the thalamus, and extensive decreases were found in the ICNs corresponding to higher-order cognitive processes including the default-mode network, dorsal attention network, central executive network and salience network. The perceptive networks and motor network were less affected by GSWD. Sequential correlation analysis further demonstrated different responses of the ICNs to GSWD. In addition to GSWDrelated functional excitation in the thalamus and functional suspension in the default-mode network, this study revealed extensive inhibitions in the other ICNs corresponding to higher-order cognitive processes, and spared perceptive and motor processes in AS. GSWD elevated synchronization of brain network activity and sequentially affected the ICNs. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Typical absence epilepsy is a common type of idiopathic generalized epilepsy in childhood; it is featured by sudden, brief impairment of consciousness, accompanied by a 2.5–4 Hz generalized spike and wave discharges (GSWD). It has been proposed that the cognitive processes underlying consciousness are selectively affected by GSWD during absence seizure (AS) [1]. Using general-linear-model (GLM) and timeseries analyses, simultaneous electroencephalography (EEG) and fMRI have depicted the spatial and temporal properties of brain activation
⁎ Corresponding authors at: Department of Medical Imaging, Jinling Hospital, 305# Eastern Zhongshan Rd., Nanjing 210002, China. Tel.: +86 25 80860185; fax: +86 25 84804659. E-mail addresses: [email protected] (Z. Zhang), [email protected] (G. Lu). 1 The authors contribute equally to this work. 0022-510X/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jns.2013.10.024
during AS. GSWD-related activation in the thalamus and deactivation in the medial frontal and posterior cingulate cortices are typically reported in EEG–fMRI studies [2–4]. These brain structures have been linked to seizure generation and deficits in the default-mode of brain function [2–4], respectively. In addition, distributed deactivation in the frontal and parietal regions has been suggested impaired attention and spared motion processes in AS [5,6]. Moreover, these brain regions presented different temporal patterns responding to GSWD during the evolution course of seizures. The accumulating imaging evidence may support the proposal that multiple cognitive processes are specifically involved in AS [7,8]. However, the precise alterations of brain processes associated with GSWD and the relationship among them have yet to be thoroughly assessed. Cognitive impairments in epilepsy have been recently related to the alteration of intrinsic connectivity networks (ICNs) [9–11]. By measuring the correlation of spontaneous hemodynamic fluctuations, resting-state fMRI has been used to link ICNs to specific cognitive processes, such as self-awareness, attention, control and perceptions [12–14].
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
Thus measuring connectivity within and across ICNs may permit examining the integrity of brain circuits related to consciousness [15–17]. This study investigated the intra- and inter-network alterations of extensive ICNs in AS, and addressed temporal evolutions of GSWD effect on the ICNs using sequential correlation analysis, which was expected to contribute to a better understanding of the neural correlates of consciousness impairments in absence epilepsy.
2. Methods 2.1. Patients Sixteen patients with childhood absence epilepsy were recruited in this study (detailed in Table 1). They met the following criteria: (i) clinical diagnosis of childhood absence epilepsy based on International League Against Epilepsy criteria [18]; (ii) EEG with typical 2.5–4 Hz bilateral ictal GSWD and normal background activity; (iii) no additional seizure types, such as myoclonic, tonic–clonic, or partial seizures; (iv) no known structural brain abnormality in routine MRI and other neurological disorders. All patients had informed consent signed by their legal guardians, and all human study procedures were approved by the institutional review boards at Jinling Hospital, Nanjing University School of Medicine.
2.2. Simultaneous EEG and fMRI data acquisitions All patients successfully underwent simultaneous EEG and fMRI data acquisitions on a 32 channels MRI-compatible EEG (Brain Product, Munich, Germany) and a 3 T MRI scanner (Siemens Trio, Erlangen, Germany). The patients were instructed to keep rest and not fall in sleep. Foam pads were used to help secure the EEG leads, minimize motion, and improve patient comfort. For EEG recordings, the electrode FCz was set as the reference and electrocardiography was recorded using an electrode placed on the back. Data were transmitted via an optic fiber cable from the amplifier placed inside the scanner room to a computer outside the scanner room. For MRI data scanning, the functional data were acquired using a T2*-weighted single-shot echo planar imaging sequence (TR/TE = 2000 ms/40 ms, FA = 90°, matrix = 64 × 64, FOV = 24 × 24 cm, thickness/gap = 4.0 mm/0.4 mm, 2 sessions with each consisting of 500 volumes each, collected after five dummy volumes). Threedimensional magnetization prepared rapid acquisition gradient-echo T1 anatomical images (TR/TE = 2300 ms/2.98 ms, FA = 9°, matrix = 256 × 256, FOV = 256 × 256 mm2, and slice thickness = 1 mm) were obtained as a structural reference.
Table 1 Demographic and clinical information of involved patients. Sex/Nos.
Age, y
Onset age, y
Medication
Nos. and durations of GSWD (in s) in AS
F1 F2 F3 F4 M1 M2 F5 F6 M3 F7 M4 F8 Mean ± Std
12 6 5 11 7 10 7 8 9 18 7 10 9.1 ± 3.5
9 4 4 9 5 9 7 6 6 10 8 3 6.7 ± 2.3
None SV, LTG, LEV None None SV, LTG, LEV SV SV SV None SV SV SV, LTG /
2 (75) 2 (17) 3 (16) 1 (14) 2 (15) 3 (33) 8 (73) 1 (12) 1 (9) 1 (16) 2 (39) 1 (16) 2.2 ± 1.9 (28.8 ± 22.8)
Abbreviations: F: Female, M: male; SV: Sodium Valproate; LTG: Lamotrigine; LEV: levetriacetam; AS: Absence seizure; GSWD: Generalized spike-wave discharge.
139
2.3. Data preprocessing The EEG data was offline-processed to remove gradient and ballistocardiogram artifacts using the Brain Vision Analyzer 2.0 software. The GSWDs were marked on artifact-removed EEG by an experienced neurologists (Yang) and an electroencephalographer (Tian). The fMRI data preprocessing was performed using a software package SPM8 (http://www.fil.ion.ucl.ac.uk/spm). After slice-timing adjustment and realignment for head-motion correction, data were realigned to the corresponding anatomical images, warped into the anatomical MNI152 space using a 12-parameter affine linear transformation, resliced at a resolution of 3 × 3 × 3 mm3, and spatially smoothed using an isotropic Gaussian kernel (8 mm full width at half maximum). In order to match data segments with and without GSWD in each individual, we divided the full amount of fMRI data into four subsessions comprising 250 volumes. Accordingly, we selected pairs of data segments from 12 patients who presented seizures during scanning. Each pair consisted of a GSWD sub-session and a matched non-GSWD sub-session. The following criteria were used for data selection: (i) Each GSWD sub-session contained seizure events. Seizure event was defined to occur if the GSWD was longer than 6 s [19,20] and the events had the same duration and morphology as clinically confirmed events on the routine EEG using the International League Against Epilepsy guidelines[4]. A total of 32 events of GSWD were included. The GSWD durations across events are 11.6 ± 8.9 s, and across subsession are 29.6 ± 22.6 s. (ii) For the non-GSWD sub-sessions, no discharge occurred 18 s before or after the selected data segment. (iii) The sub-sessions containing large motion (more than 1.5 mm or 1.5°) or the ones that could not be matched in the same subject were excluded (see Supplementary Fig. 1). There were no significant differences of head motion between the two data groups (Paired t-test, t = −0.6, p = 0.52 for translation and t = −0.8, p = 0.48 for rotation). 2.4. Data analysis 2.4.1. Sequential HRF generation and dynamic GLM framework In line with previous studies [21,22], we first generated a sequence of hemodynamic response functions (HRFs) to model the dynamic BOLD changes induced by GSWD. Sequential HRFs consisted of 49 successive gamma functions of FWHM of 5.2 s, peak = 5.4 s, centered at 0 s and spaced 1 s between one another. At the individual level, we convolved a boxcar function expressing the timing of seizure events with sequential HRFs shifted between 24 s before to 24 s after the GSWD (HRF-24 to HRF + 24), to generate a series of regressors that modeled the BOLD responses. Then, we performed separate t-tests within a general-linear-model (GLM) framework for each regressor with specific HRF, thereby producing 49 t-maps representing GSWDrelated BOLD activation modeled with different HRFs. The dynamic GLM framework allows us to observe the dynamic BOLD changes before and after seizure onset [22]. For group-analysis, we used one-sample t-tests (p b 0.05, AlphaSim correction) to determine regions showing significant GSWDs-related BOLD changes. Sequential-HRF based dynamic GLM framework was mainly used as a reference for the subsequent ICA analysis. 2.4.2. Group comparison analysis for intra- and inter-network connectivity of ICNs Subsequently, we performed group independent component analysis (ICA) to examine ICNs in the patients. We used the GIFT software (version 2.0d; http://icatb.sourceforge.net/) to both fMRI sub-sessions related to GSWD and non-GSWD. Forty-four spatially independent components were decomposed after component number estimation using the minimum description length criteria [23]. The decomposition produced a set of components, each with a spatial map of intensities and a representative time course. The spatial maps were converted to zscores, which reflect the degree to which synchronous activity occurs
140
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
across spatially independent but widely distributed brain regions, i.e., the degree of functional connectivity [24]. Based on our hypothesis derived from literatures in the field of absence seizures [5,11,21,22,25] and functional brain networks [12,26], we selected ten well-identified networks [12,26,27] as components-of-interest using spatial correlation analyses. These components include the thalamic network (THA), posterior and anterior default-mode (aDMN and pDMN), dorsal attention network (DAN), central executive network (CEN), salience (SN), sensory (SEN), auditory (AN), visual (VN) and motor networks (MN). The reference image used to identify the thalamic network was obtained from our GLM analysis, whereas the other reference images were most studied and well defined in the previous studies (see Supplementary Fig. 2) [9,28–30]. Considering the small sample-size of our data, we also performed a leave-one-out (LOO) analysis to test the reproducibility of each ICN extracted by ICA (see Supplementary Fig. 3). The effect of GSWD on ICNs was examined by comparing ICNs during GSWD and non-GSWD states. To this end, we used voxel-wise paired-t tests (pb0.05, AlphaSim correction) on the ICN maps associated with the two states. This permitted investigating alterations of synchronous activation in each ICN. We restricted this comparison to predefined masks generated by calculating one-sample t-tests for GSWD and non-GSWD states separately. Moreover, we correlated activation intensities (averaged z values in the regions showing altered connectivity) with the duration of GSWD using a Spearman correlation analysis (p b 0.05, Bonferroni correction). Subsequently, we also measured functional network connectivity (FNC) (http://icatb.sourceforge.net/) to investigate the interactions among ICNs. Following this approach, we investigated the temporal relationships among ICN time courses by computing a constrained maximal lagged correlation (maximal lag = 10 s) [31]. We calculated a one-sample t-test on pair-wise correlations across subjects for the GSWD and non-GSWD states separately, and we also compared the two states using paired t-tests (p b 0.05). 2.4.3. Sequential correlation analysis between GSWD and ICNs We further investigated the effect of GSWD on the ICNs in AS using sequential HRF-based correlation analysis. The GSWD modeling signals (obtained by convolving sequential HRFs with a GSWD-based boxcar function) were sequentially correlated with the time course of each ICN [32]. By extracting the maximum among these correlations, we defined a reference time lag at which a specific ICN was most likely affected by GSWD. The correlation coefficients were transformed into z-scores using the Fisher's r-to-z transformation, and were given as input to a one-sample t-test (p b 0.05, Bonferroni correction). We also used the peak of the GSWD modeling signals averaged across subjects to define the latency of ICN maximum activation/deactivation. In addition, to explicitly demonstrate the relationship between GSWD and the time courses of ICNs, we showed the averaged time courses across all the GSWD events over a 60 s window centered at the time of the GSWD onset. All the data analysis steps are illustrated in Supplementary Fig. 4. 3. Results The GLM-based analysis with sequential HRFs revealed dynamic properties of GSWD-related BOLD activation and deactivation (Supplementary Fig. 5). Each ICN extracted by ICA showed highly reproducibility (Supplementary Fig. 3). Voxel-wise paired t-tests demonstrated that the intra-network connectivity was altered in each ICN as affected by GSWD. The components corresponding to the thalamus and the AN showed increased connectivity in the GSWD state compared to the non-GSWD state. In contrast, the aDMN, pDMN, DAN and the CEN showed decreased connectivity in GSWD state. Furthermore, GSWD durations were found to be positively correlated with the intra-network connectivity (expressed by z-values) in the thalamus, and negatively
correlated with connectivity intensities in the pDMN, DAN and CEN (Fig. 1). FNC analysis revealed that the inter-network communications among ICNs were largely altered during GSWD state. First, the number of significant connections in the GSWD state was more than the nonGSWD state (21 vs. 14). Secondly, comparison analysis showed that the thalamus, aDMN and pDMN were most affected networks, particularly in terms of the number of inter-network connections and activation time lags (Fig. 2). For the sequential correlation analyses between the GSWD modeling signals and the time course of ICNs, the thalamus showed positive correlation, while the aDMN, pDMN, DAN, CEN and SN all showed negative correlations (p b 0.05, Bonferroni correction). None of the perceptive network showed significant correlation between the time series and the modeling signals. The sequence number of the GSWDmodeling signals of each ICN was used to define the latency of ICN activation/deactivation. We obtained the following latencies for the different ICNs: THA (HRF +9, i.e., starting time at 9 s after GSWD onset), aDMN (HRF +9), pDMN (HRF +11), DAN (HRF +15), CEN (HRF +11) and SN (HRF +17). These results were consistent with those of sequential correlation analyses (Fig. 3). 4. Discussion By employing an ICA-based functional connectivity approach, this study provided a comprehensive analysis of the effects of GSWD on a set of ICNs in AS. (1) ICA and GLM analyses both revealed that, while the thalamus showed positive BOLD response to GSWD, extensive cortical regions and ICNs had negative responses to GSWD. (2) ICNs corresponding to the higher-order cognitive processes, including the CEN, DAN and SN, were severely influenced by GSWD; whereas the perceptive networks and the MN showed much less effects of GSWD in AS. (3) Temporal analysis using averaged time courses and sequential correlation analyses provided converging results, and revealed a differential pattern of activation for the ICNs affected by GSWD. In line with previous works [19,22], we used a traditional GLM framework to show the dynamic properties of GSWD-related activation in the thalamus. Furthermore, we found extensive cortical structures not only covering regions of the DMN, but also including the regions of the DAN and CEN showed dynamic deactivation. These findings may suggest origination of epileptic discharges and dysfunction of the cortical regions [22,33]. Moreover, the dynamic properties of BOLD responses implicate that different regions were involved sequentially in the initiation, maintenance and termination of AS [19,22]. Importantly, we employed ICA to investigate ICN changes during AS. Different with the GLM framework for activation localization, ICA is a data-driven technique that decomposes the whole data set into several ICNs characterized by a specific spatial pattern and does not require any prior assumption related to the HRF [34,35]. Specifically, the ICNs reflect the architecture of coherent intrinsic activity in the brain. The DMN, DAN, CEN, SN and perceptive networks have been well linked to cognitive processes of self-awareness/episodic memory, top-down attention, cognitive control, personal salience and sensory integration, respectively [12,14,36,37]. By using spatial correlation with predefined templates, it was possible to identify a set of ICNs corresponding to specific cognitive functional processes [12,14,31,36], as well as regions associated with epileptic activation [34,35]. The thalamus not only showed positive response to GSWD, but also increased intra-network connectivity in the GSWD state relative to the non-GSWD state. Conversely, other ICNs corresponding to high-order cognitive processes, including the aDMN, pDMN, CEN, DAN and SN showed negative responses to GSWD, and decreased intra-network connectivity in the GSWD states relative to the non-GSWD state. Our results concerning positive BOLD responses in the thalamus and negative responses in a large portion of the cortex during GSWD are in accordance with previous findings [19,21,22]. Moreover, the GSWD
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
141
Fig. 1. Alterations of intra-network connectivity in the GSWD sub-session revealed by group comparisons. Voxel-based comparison of each ICN (z-map) using paired t-tests between GSWD and non-GSWD sub-sessions. In the GSWD state, the thalamus network and the AN showed increased intra-network connectivity, while the aDMN, pDMN, DAN and CEN showed decreased activation relative to the non-GSWD sub-session. There was no group difference in the SN, SEN, VN and MN. Moreover, for each ICN showing altered intra-network connectivity, the activation intensities (z-scores) in the regions showing group difference were correlated with GSWD durations (Spearman correlation, p b 0.05, Bonferroni correction). Positive correlation was found in the thalamus network, and negative correlations were found in the pDMN, DAN and CEN. No significant correlation was found in the aDMN and AN.
related increased and decreased intra-network connectivity might recapitulate the functional excitation of the thalamus and functional suspensions of the cortical networks in the intrinsic brain activities [33]. In addition to those of the DMN [19,21,22] and DAN [5,6], we for the first time reported the functional inhibitions in the higher-order processes of CEN and SN during AS. The ICNs corresponding to the perceptive networks and the MN showed less alteration responsive to GSWD. The AN presented increased functional connectivity in the GSWD state, but no correlation with GSWD was found. Furthermore, none of the perceptive networks showed significant correlation between the GSWD-modeling signals and the ICN time-courses. The GLM analysis revealed minor deactivation in the regions of the sensory, visual and motor cortices. These findings
supported the proposal that the motor process for simpler task is spared in AS [1,6], and further indicated that the lower-order perceptive processes are also less affected by GSWD. FNC analysis revealed that inter-networks relationships among ICNs are altered in AS [22,27]. More inter-networks communications in the GSWD sub-session suggest a hyper-synchronization of large-scale brain networks [38,39] induced by GSWD. These results may also imply that the intrinsic processes relative to cognitive activities are hijacked by epileptic discharges. Regarding to the critical roles of the thalamus and the DMN played in the maintenance of level of consciousness [8,15,40], the largest alterations of the inter-networks communications in the thalamus and DMN suggest the crucial roles of these two structures in AS. The specific alterations in the other ICNs
142
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
Fig. 2. Alterations of inter-network connectivity in the GSWD sub-session revealed by group comparisons. FNC analysis showed the inter-network communication among these ICNs. Visual inspection revealed that the amount of inter-network connections (one-sample t-tests for each sub-session) was increased in the GSWD sub-sessions relative to the non-GSWD sub-sessions (21 vs. 14). Paired t-test comparisons (p b 0.05) between two states revealed that the thalamus network and the DMN showed the most prominent alterations.
may implicate selective impairments of cognitive processes for the different extents of consciousness [4,19]. Both sequential correlation analysis and averaged time course analysis demonstrated a direct effect of GSWD on ICNs. With either approaches, ICNs showed different temporal responses with respect to GSWD. The thalamus network had a response peak at a latency of about 14s (9s+5.4s) after GSWD onset, which was almost simultaneous to that of aDMN but went in opposite direction. This result is in line with
those of previous studies [4,19], which used a temporal analysis based on regions of interest. Other ICNs such as pDMN, CEN, DAN and SN had later response peaks after GSWD onset, while we found no significant correlation in the perceptive networks and the MN. These results were largely consistent with those revealed by GLM analysis. Specifically, the sequence of ICN response peaks may provide valuable information on the hierarchy of networks that subserving impaired cognitive processes in AS. Moreover, we also noted the inconsistent findings between
Fig. 3. Averaged ICN time courses and sequential correlation with GSWD-modeling signals. For each ICN, we averaged the time courses across all GSWD events over a 60 s window centered at the time of GSWD onset. The curve line denotes the mean, and the shade denotes std of the z values. This analysis revealed that the thalamus network had positive BOLD response (red line); the higher-order cognitive networks including the aDMN, pDMN, CEN, DAN and SN showed negative BOLD responses (blue line) to GSWD at various peak times. The inset frames indicate sequential correlation results between sequential GSWD modeling signals and the ICN time courses. The stars denote significant correlations (p b 0.05, Bonferroni correction). The thalamus showed maximal positive correlation (t = 5.70, p = 1.67 × 10−8) when using HRF + 9. The aDMN (HRF + 9, t = −5.76, p = 1.17 × 10−8), pDMN (HRF + 11, t = −7.42, p = 2.90 × 10−13), CEN (HRF + 11, t = −6.45, p = 1.85 × 10−10), DAN (HRF + 15, t = −4.84, p = 1.52 × 10−6) and SN (HRF + 17, t = −5.84, p = 7.10 × 10−9) all showed negative correlations. The perceptive network and the MN showed no significant correlation between their time courses and the modeling signals. These results were largely consistent with the sequential correlation analysis. Specifically, the peak time of the average time courses was indeed similar to the optimal time lag estimated using the sequential correlation analysis plus the peak latency of the HRF (i.e., 5.4 s).
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
143
144
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
the ICA and GLM analysis, e.g., the time point of peak response of the thalamus. We considered that utilization of specific HRF in GLM and free of prior assumption related to the HRF in ICA might be the reason. It is important to mention that a number of potential limitations apply to our study. First, resting-state fMRI has proven to be a valid tool for detecting epileptic discharges-related activation [34,35] and for better understanding cognitive impairments in epilepsy [9,11,27], but the relationship between the resting-state ICNs and task-induced activations is still partially unclear [12,13]. Future studies with more imaging and simultaneous behavioral data acquisition are needed. Secondly, we used ICA, a data-driven method for resting-state fMRI analysis, to detect a set of ICNs related to specific cognitive processes [12,14]. It is worth noting that ICA results might be influenced by model order estimation, and may be difficult to interpret due to the complexity of the output structure [13,41]. Thirdly, this study has a relatively small sample size and future studies on a large number of patients are necessary to corroborate our findings. However, the high reproducibility of ICN maps supports the robustness of our results (Supplementary Fig. 3). Finally, the effect of anti-epileptic drugs on ICNs and different types of epileptic discharges were not addressed in the present study. 5. Conclusion This study comprehensively investigated the effect of GSWD on ICNs during AS. We confirmed previous findings concerning thalamic excitation and DMN suspension. Moreover, the results revealed extensive inhibitions in the ICNs corresponding to the high-order cognitive processes, such as the DAN, CEN and SN in AS; while the ICNs corresponding to the perceptive and motor networks seemed to be only less influenced by GSWD effects. Moreover, these ICNs presented enhancement in the synchronization of the large-scale brain network activity, suggesting aberrant relationship among the cognitive processes in consciousness system in AS. Finally, the sequential effects of GSWD on different ICNs may potentially reveal different roles of these networks in AS. Overall, our findings may contribute to a better understanding of the pathophysiological mechanisms underlying cognitive impairments during AS. Conflicts of interest The authors declare that they have no conflict of interest. Acknowledgment This research was supported by the Natural Science Foundation of China (Grant nos. 81271553, 81201155, 30971019, 81020108022) and Grants for Young Scholar of Jinling Hospital (Grant nos. 2011060, 201 1045). Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.jns.2013.10.024. References [1] Aldenkamp A, Arends J. The relative influence of epileptic EEG discharges, short nonconvulsive seizures, and type of epilepsy on cognitive function. Epilepsia 2004;45(1):54–63. [2] Salek-Haddadi A, Lemieux L, Merschhemke M, Friston KJ, Duncan JS, Fish DR. Functional magnetic resonance imaging of human absence seizures. Ann Neurol 2003;53(5):663–7. [3] Moeller F, Siebner HR, Wolff S, Muhle H, Granert O, Jansen O, et al. Simultaneous EEG–fMRI in drug-naive children with newly diagnosed absence epilepsy. Epilepsia 2008;49(9):1510–9.
[4] Carney PW, Masterton RA, Harvey AS, Scheffer IE, Berkovic SF, Jackson GD. The core network in absence epilepsy. Differences in cortical and thalamic BOLD response. Neurology 2010;75(10):904–11. [5] Killory BD, Bai X, Negishi M, Vega C, Spann MN, Vestal M, et al. Impaired attention and network connectivity in childhood absence epilepsy. Neuroimage 2011;56(4):2209–17. [6] Berman R, Negishi M, Vestal M, Spann M, Chung MH, Bai X, et al. Simultaneous EEG, fMRI, and behavior in typical childhood absence seizures. Epilepsia 2010;51(10): 2011–22. [7] Blumenfeld H, Taylor J. Why do seizures cause loss of consciousness? Neuroscientist 2003;9(5):301–10. [8] Blumenfeld H. Impaired consciousness in epilepsy. Lancet Neurol 2012;11(9):814–26. [9] Zhang Z, Lu G, Zhong Y, Tan Q, Yang Z, Liao W, et al. Impaired attention network in temporal lobe epilepsy: a resting FMRI study. Neurosci Lett 2009;458(3):97–101. [10] Vlooswijk MC, Jansen JF, de Krom MC, Majoie HM, Hofman PA, Backes WH, et al. Functional MRI in chronic epilepsy: associations with cognitive impairment. Lancet Neurol 2010;9(10):1018–27. [11] Yang T, Luo C, Li Q, Guo Z, Liu L, Gong Q, et al. Altered resting-state connectivity during interictal generalized spike-wave discharges in drug-naive childhood absence epilepsy. Hum Brain Mapp 2012. http://dx.doi.org/10.1002/hbm.22025. [12] Zhang D, Raichle ME. Disease and the brain's dark energy. Nat Rev Neurol 2010;6(1): 15–28. [13] Fox MD, Raichle ME. Spontaneous fluctuations in brain activity observed with functional magnetic resonance imaging. Nat Rev Neurosci 2007;8(9):700–11. [14] Mantini D, Perrucci MG, Del Gratta C, Romani GL, Corbetta M. Electrophysiological signatures of resting state networks in the human brain. Proc Natl Acad Sci U S A 2007;104(32):13170–5. [15] Boly M, Phillips C, Tshibanda L, Vanhaudenhuyse A, Schabus M, Dang-Vu TT, et al. Intrinsic brain activity in altered states of consciousness: how conscious is the default mode of brain function? Ann N Y Acad Sci 2008;1129:119–29. [16] Vanhaudenhuyse A, Noirhomme Q, Tshibanda LJ, Bruno MA, Boveroux P, Schnakers C, et al. Default network connectivity reflects the level of consciousness in noncommunicative brain-damaged patients. Brain 2010;133(Pt 1):161–71. [17] Greicius MD, Kiviniemi V, Tervonen O, Vainionpaa V, Alahuhta S, Reiss AL, et al. Persistent default-mode network connectivity during light sedation. Hum Brain Mapp 2008;29(7):839–47. [18] Engel Jr J. A proposed diagnostic scheme for people with epileptic seizures and with epilepsy: report of the ILAE Task Force on Classification and Terminology. Epilepsia 2001;42(6):796–803. [19] Bai X, Vestal M, Berman R, Negishi M, Spann M, Vega C, et al. Dynamic time course of typical childhood absence seizures: EEG, behavior, and functional magnetic resonance imaging. J Neurosci 2010;30(17):5884–93. [20] Li Q, Luo C, Yang T, Yao Z, He L, Liu L, et al. EEG–fMRI study on the interictal and ictal generalized spike-wave discharges in patients with childhood absence epilepsy. Epilepsy Res 2009;87(2–3):160–8. [21] Moeller F, LeVan P, Muhle H, Stephani U, Dubeau F, Siniatchkin M, et al. Absence seizures: individual patterns revealed by EEG–fMRI. Epilepsia 2010;51(10):2000–10. [22] Benuzzi F, Mirandola L, Pugnaghi M, Farinelli V, Tassinari CA, Capovilla G, et al. Increased cortical BOLD signal anticipates generalized spike and wave discharges in adolescents and adults with idiopathic generalized epilepsies. Epilepsia 2012;53 (4):622–30. [23] Li YO, Adali T, Calhoun VD. Estimating the number of independent components for functional magnetic resonance imaging data. Hum Brain Mapp 2007;28(11): 1251–66. [24] Calhoun VD, Kiehl KA, Liddle PF, Pearlson GD. Aberrant localization of synchronous hemodynamic activity in auditory cortex reliably characterizes schizophrenia. Biol Psychiatry 2004;55(8):842–9. [25] Luo C, Li Q, Lai Y, Xia Y, Qin Y, Liao W, et al. Altered functional connectivity in default mode network in absence epilepsy: a resting-state fMRI study. Hum Brain Mapp 2011;32(3):438–49. [26] Moeller F, Maneshi M, Pittau F, Gholipour T, Bellec P, Dubeau F, et al. Functional connectivity in patients with idiopathic generalized epilepsy. Epilepsia 2011;52(3):515–22. [27] Luo C, Li Q, Xia Y, Lei X, Xue K, Yao Z, et al. Resting state basal ganglia network in idiopathic generalized epilepsy. Hum Brain Mapp 2011. http://dx.doi.org/10.1002/ hbm.21286. [28] Wang Z, Lu G, Zhang Z, Zhong Y, Jiao Q, Tan Q, et al. Altered resting state networks in epileptic patients with generalized tonic–clonic seizures. Brain Res 2011;1374:134–41. [29] Liao W, Chen H, Feng Y, Mantini D, Gentili C, Pan Z, et al. Selective aberrant functional connectivity of resting state networks in social anxiety disorder. Neuroimage 2010;52(4):1549–58. [30] Allen EA, Damaraju E, Plis SM, Erhardt EB, Eichele T, Calhoun VD. Tracking whole-brain connectivity dynamics in the resting state. Cereb Cortex 2012. http://dx.doi.org/10.1093/cercor/bhs352. [31] Vlassenko AG, Rundle MM, Raichle ME, Mintun MA. Regulation of blood flow in activated human brain by cytosolic NADH/NAD+ ratio. Proc Natl Acad Sci U S A 2006;103(6):1964–9. [32] Mantini D, Corbetta M, Perrucci MG, Romani GL, Del Gratta C. Large-scale brain networks account for sustained and transient activity during target detection. Neuroimage 2009;44(1):265–74. [33] Gotman J, Grova C, Bagshaw A, Kobayashi E, Aghakhani Y, Dubeau F. Generalized epileptic discharges show thalamocortical activation and suspension of the default state of the brain. Proc Natl Acad Sci U S A 2005;102(42):15236–40. [34] Rodionov R, De Martino F, Laufs H, Carmichael DW, Formisano E, Walker M, et al. Independent component analysis of interictal fMRI in focal epilepsy: comparison
Z. Zhang et al. / Journal of the Neurological Sciences 336 (2014) 138–145
[35]
[36]
[37]
[38]
[39] [40] [41]
with general linear model-based EEG-correlated fMRI. Neuroimage 2007;38(3):488–500. LeVan P, Tyvaert L, Moeller F, Gotman J. Independent component analysis reveals dynamic ictal BOLD responses in EEG–fMRI data from focal epilepsy patients. Neuroimage 2010;49(1):366–78. Damoiseaux JS, Rombouts SA, Barkhof F, Scheltens P, Stam CJ, Smith SM, et al. Consistent resting-state networks across healthy subjects. Proc Natl Acad Sci U S A 2006;103(37):13848–53. Seeley WW, Menon V, Schatzberg AF, Keller J, Glover GH, Kenna H, et al. Dissociable intrinsic connectivity networks for salience processing and executive control. J Neurosci 2007;27(9):2349–56. Meeren H, van Luijtelaar G, Lopes da Silva F, Coenen A. Evolving concepts on the pathophysiology of absence seizures: the cortical focus theory. Arch Neurol 2005;62(3):371–6. Snead III OC. Basic mechanisms of generalized absence seizures. Ann Neurol 1995;37(2):146–57. Blumenfeld H. Epilepsy and the consciousness system: transient vegetative state? Neurol Clin 2011;29(4):801–23. Fox MD, Greicius M. Clinical applications of resting state functional connectivity. Front Syst Neurosci 2010;4:19.
Glossary AN: auditory network AS: absence seizures BOLD: blood oxygen level-dependent CAE: childhood absence epilepsy CEN: central executive network DAN: dorsal attention network DMN: default-mode network FNC: functional network connectivity GLM: general-linear-model GSWD: generalized spike and wave discharge HRF: hemodynamic response function ICA: independent component analysis ICN: intrinsic connectivity networks MN: motor network SEN: sensory network SN: salience network VN: visual network
145
