Brain functional connectivity predicts outcome in comatose patients after cardiac arrest
Thomas Kustermann (Lausanne | CH)
Clinical outcome prognostication in comatose patients following cardiac arrest is currently based on multimodal assessments of clinical and electrophysiological markers. This requires the intervention of trained experts and lacks standardization. Quantitative Electroencephalography (EEG) analysis could provide complementary and unbiased information about patients’ chances of recovery. Here we investigate the properties of EEG-based functional brain networks in comatose patients and their predictive power regarding patient’s outcome. During the first day of coma, we used 63 channel EEG to prospectively record resting state activity in comatose patients after cardiac arrest. Of the 92 patients included in our study, 55 survived beyond unresponsive wakefulness. Functional networks were based on the ‘debiased weighted phase lag index’ computed over epochs of five seconds. We derived topological features, including clustering coefficient, path length, modularity and participation coefficient. For all topological measures, we investigated their variance over time and computed predictive values for patients’ outcome by splitting the sample in training and test datasets. Group-level analysis revealed significantly different network organization during the first day in survivors and non-survivors. Time variance of path length provided the best test set prediction of good outcome on the first day of coma (PPV:.85, CI:.55-.98, Specificity:.89, 95% CI:.65-.99). Excluding patients with epileptiform activity would have eliminated all false positive predictions. Overall, the time variance of path length in functional connectivity is highly informative of patients’ outcome as survivors exhibit a richer repertoire of path length than non-survivors during the first day.
MRI-EEG correlation for outcome prediction in post-anoxic myoclonus, a multicenter study
Isabelle Beuchat (lausanne | CH)
Post-anoxic myoclonus are historically considered a strong predictor of poor prognostic after cardiac arrest (CA). However, more recent studies, proved that a subset of these patients do not have such a dire outcome. We aim to examine the prognostic ability of EEG and MRI to identify patients with good outcome in a multi-centric retrospective study.
Adults with post-anoxic myoclonus who had an MRI within 15 days after CA were retrospectively identified in four prospective CA registry. Functional outcome was assessed through the ability to follow command at discharge and through cerebral performance category (CPC) at 3 months with CPC 1-2 defined good outcome. EEG and MRI were prospectively interpreted by local neuroradiologist and neurophysiologist.
78 patients (median age 56 years, 37% women) had an MRI with a median of 4 days after CA, 61 had anoxic brain injury (10 subcortical, 43 cortical, 35 in the basal ganglia and 19 diffuse) in T2/FLAIR (n=2), DWI/ADC (n=1) or in both (n=71) sequences, 14 had no MRI lesions and 3 other lesions (stroke, subarachnoid hemorrhage). EEG was continuous in 23 (29%), reactive in 18 (23%) and 73 (94%) showed epileptiform pattern. Myoclonus occurred early (within 48h from CA) in 53 (69%). 61 patients (78%) died, 11 (14%) follow command at discharge and 7 (9 %) presented good outcome at three months. Absence of anoxic brain injury was found in 10/11 of patients who follow command at discharge (predicting good outcome with 91% sensitivity -95%CI: 59%-99%, 90% specificity -95%CI: 80%-96% and 59% positive predictive value -95%CI: 41%-75%). Continuous EEG background was described in 11/11 of patients who follow command at discharge (predicting good outcome with 100% sensitivity -95%CI: 72%-100%, 83% specificity -95%CI: 71%-90% and 48% positive predictive value -95%CI: 35%-60%). Continuous EEG in combination with absence of anoxic brain injury predict ability to follow command at discharge with 100% sensitivity (95%CI: 63%-100%), 98% specificity (95%CI: 89%-100%) and 90% positive predictive value (95%CI: 53%-98%) and good outcome at 3 months with 83% sensitivity (95%CI: 36%-96%), 92% specificity (95%CI: 81%-98%) and 56% positive predictive value (95%CI: 31%-77%).
MRI in conjunction with EEG demonstrate a strong prognostic performance to identify patients with post-anoxic myoclonus who will regain consciousness and a good prognostic performance to identify patients with good outcome at three months.
A score to identify patients with good outcome after early epileptiform EEG following cardiac arrest
Giuseppina Barbella (Lausanne | CH)
Aims. Epileptiform patterns, occurring in about 1/3 of comatose patients after cardiac arrest (CA), are often but not invariably associated with poor outcome (1, 2). Our aim is to explore whether a combination of particular EEG patterns among those with epileptiform activity may identify patients with favourable outcome.
Methods. We retrospectively analysed a registry of comatose post-CA patients with epileptiform EEG within 3 days, admitted at two centres (CHUV, Sion Hospital; January 2013 - February 2019). EEGs at 12-36h and 36-72h from CA were scored with the ACNS nomenclature (3) (background, reactivity, status epilepticus, seizures). EEG features were compared according to outcome (CPC 1-3 vs 4-5) at three months. Significant EEG variables were combined in a score assessed with ROC curves; its correlation with serum neuron-specific enolase (NSE) was tested. Validation was obtained on an external independent cohort (BWH, Boston, USA).
Results. Among 488 patients, 107 had early epileptiform EEG. CPC 1-3, reached in 18 (17%), was associated with absence of epileptiform abnormalities and background continuity ≥ 50% at 12-36h (p < 0.00001 each), reactivity at 12-36h and 36-72h (p < 0.0001 each), normal background amplitude (p = 0.0004) and SIRPIDs at 36-72h (p = 0.0001). A 6-points score including these variables with a cut-off of ≥ 2 had sensitivity of 100% and specificity of 70% for CPC 1-3 (AUC = 0.98, 95% CI 0.94-1). A robust negative correlation was found between increasing EEG score and NSE peak values (r = -0.46, p = 0.0001). In the validation cohort, score ≥ 2 was 100% sensitive and 88% specific for Best CPC 1-3 (AUC = 0.96, 95% CI 0.91-1).
Correlation of somatosensory evoked potentials amplitude after cardiac arrest with other outcome prognosticators
Giuseppina Barbella (Lausanne/ Monza (Italy) | CH)
In comatose patients after cardiac arrest (CA), bilateral absence of cortical somato-sensory evoked potentials (SSEPs) responses is considered a specific poor outcome predictor. Recent studies 1,2 suggest that their amplitude may bear additional prognostic information. Our aim is to explore the SSEP amplitude relationship with other known prognosticators.
Cortical SSEPs amplitudes were measured (N20/P25 peak-to-peak, Fz reference)1,2 in consecutive patients of the Lausanne CA registry. We assessed their correlation with pupillary light reflex (PLR), motor response at 72h, EEG reactivity and absence of epileptiform features, and serum NSE peak. An amplitude cut-off was sought; its added value predicting poor prognosis (CPC 4-5 within 3 months) in addition to other tests was explored using ROC curves.
Among 158 patients, 114 (72%) men, aged 62.5y (± 14.6), 86 (54%) awakened within three months. Mean SSEPs amplitude was 3.08 (2.03) in awakeners vs 1.25 (1.71) uV in the others (p < 0.0001).
SSEP amplitudes were negatively correlated with serum NSE (r = -0.379; p < 0.0001), positively with both EEG at 12-36h and 36-72h components (respectively, r = 0.462, p < 0.0001; r = 0.448; p < 0.0001), and positively with present motor response and PLR (both p < 0 .0001).
A SSEP amplitude ≤ 0.41 uV was 100% specific (95% CI, 96-100%), and 47 % sensitive (95% CI, 35-59%) for CPC 4-5 (AUC = 0.802, 95%CI, 0.730- 0.874).
A score including EEG (reactivity, epileptiform features at 12-36h), clinical (presence of myoclonus, absence of PLR) and biochemical parameters (peak NSE) was 100% specific (95%CI, 96-100%) and 38% sensitive (95%CI, 26-51%) for CPC 4-5, with an AUC = 0.889 (95%CI 0.831-0.947). Adding SSEPs ≤ 0.41 uV, sensitivity was 44% (95%CI, 32-57%), and AUC=0.839 (95% CI 0.768-0.910). This difference was not statistically significant (p = 0.13).
Even if SSEPs amplitudes correlate with clinical outcome and other recognized prognostic variables, adding them to a multimodal prognostication including EEG, clinical and biochemical variables does not enhance poor outcome prediction, suggesting that SSEP are redundant in clinical practice if considering others predictors.
High-density ECoG improves the detection of high frequency oscillations that predict seizure outcome.
Ece Boran (Zurich | CH)
Objectives: Residual fast ripples (FR) in the intraoperative ECoG are highly specific predictors of postsurgical seizure recurrence. However, a FR is generated by a small patch of cortical tissue. Spatial sampling with standard electrodes may thus miss clinically relevant information.
Methods: We analyzed FR rates in the intraoperative ECoG of 22 patients that underwent resective epilepsy surgery. We used standard electrodes with 10 mm inter-contact spacing (standard ECoG) in 14 surgeries and high-density grid electrodes with 5 mm spacing (hd-ECoG) in 8 surgeries. We detected FR using a previously validated automatic detector.
Results: Postoperative seizure freedom was achieved in 13/22 (59%) cases. Across all 42 ECoG recordings, FR rates were higher for hd-ECoG than for standard ECoG. In the 13 seizure free patients (ILAE 1), no residual FR were detected (specificity = 100%). In the 10 patients with seizure recurrence (ILAE > 1), residual FR were detected in 2/2 hd-ECoG and 1/8 standard ECoG (Accuracy ACCstandard ECoG = 50%, CI [23% 77%], ACChd-ECoG = 100%, CI [63% 100%]).
Conclusion: Denser spatial sampling by hd-ECoG improved FR detection and thus seizure outcome prediction compared to standard ECoG.
Significance: Hd-ECoG may advance seizure freedom after epilepsy surgery.
High-Frequency oscillations in scalp EEG mirror seizure frequency in pediatric focal epilepsy
Ece Boran (Zurich | CH)
Objective: High-frequency oscillations (HFO) are promising EEG biomarkers of epileptogenicity. While the evidence supporting their significance derives mainly from invasive recordings, recent studies have extended these observations to HFO recorded in the widely accessible scalp EEG. Here, we investigated whether scalp HFO in drug-resistant focal epilepsy correspond to epilepsy severity, if they are related to invasive HFO and how they are affected by surgical therapy.
Methods: In 11 children with drug-resistant focal epilepsy that underwent epilepsy surgery, we recorded pre- and postsurgical scalp EEG with a custom-made low-noise amplifier (LNA), in addition to a commercial device (CD). In four of these children, we also recorded intraoperative electrocorticography (ECoG). To detect clinically relevant HFO in both scalp EEG and ECoG, we applied a previously validated automated detector in the time-frequency domain. We compared the scalp HFO rates with the seizure frequency and the HFO location between scalp EEG and ECoG.
Results: Scalp HFO rates showed a significant positive correlation with seizure frequency (R2 = 0.84, p < 0.001). Overall, scalp HFO rates were higher in patients with active epilepsy (17 recordings, p = 0.006, PPV = 93%, NPV = 100%, accuracy = 94% CI [71% 100%]) and decreased following successful epilepsy surgery. Higher scalp HFO rates were detected with the LNA compared to the CD (p < 0.001). The location of the highest HFO rates in scalp EEG matched the location of the highest HFO rates in ECoG.
Significance: HFO in scalp EEG mirror seizure frequency, and thus disease severity, in children with drug-resistant focal epilepsy. The LNA considerably improves detectability, and the automated detector ensures a prospective, bias-free definition of clinically relevant HFO in scalp EEG. This study is the first step towards using non-invasively recorded scalp HFO for therapy monitoring in patients affected by epilepsy.
Brain-computer interfaces based on cortical source activity during attempted movements reconstructed from high-density EEG in patients with amyotrophic lateral sclerosis
Christoph Pokorny (Geneva | CH)
For people with motor neuron diseases, such as amyotrophic lateral sclerosis (ALS), a brain-computer interface (BCI) may be the only way to provide a means of communication. Bypassing any muscular output function, a BCI directly translates brain patterns into control signals for communication and environmental control. To this end, we studied the brain patterns in response to different movement attempts in healthy subjects and in five people with ALS with different severity of paralysis, ranging from partial paralysis to complete locked-in syndrome.
The subjects performed a delayed instructed movement task with a range of movements. Movement instructions were presented on a screen or as spoken words from a loudspeaker, and were followed after two seconds by a go cue. As the subjects performed the task, we recorded high-density electroencephalogram (EEG), electrooculogram, and electromyogram using a 128-channel recording system (ANT Neuro). We co-registered the EEG electrode positions with the structural magnetic resonance imaging scans to accurately reconstruct the sources of cortical activity. We then calibrated a regularized linear discriminant analysis decoder on movement-related cortical potentials (MRCPs) and event-related desynchronization and synchronization (ERDS) responses in the beta frequency bands.
We characterized the subjects’ motor-evoked responses, even in the absence of any actual movements. We found preparatory responses and MRCPs in the low-frequency range, as well as ERDS responses present in different frequency bands for different subjects. Surprisingly, these responses could be identified in single trials. Our decoder successfully detected feet, wrist and finger movements in an asynchronous test scenario.
These results demonstrate the potential to develop an effective communication BCI based on high-density EEG cortical source reconstruction for people with locked-in syndrome.
Decoding gait events from high density EEG in healthy volunteers as preliminary step towards brain-controlled neuromodulation in people with paraplegia.
Andrea Galvez (Lausanne | CH)
Accurate decoding of gait events from electroencephalographic (EEG) recordings would provide the optimal information to trigger spinal cord stimulation in order to alleviate gait deficits and promote functional recovery in people with paraplegia. As a preliminary step towards this goal, we investigated cortical dynamics reconstructed from high-density EEG recordings (128 electrodes) during walking on a treadmill and overground in ten healthy volunteers. In order to track and study gait, we recorded bilateral electromyographic (EMG) activity from seven lower extremity muscles, and used a motion capture system to acquire three-dimensional kinematics during walking. We identified key gait events (foot strike and foot off) by analysis of EMG signals and kinematic parameters. We reconstructed cortical dynamics using EEG source imaging based on individual anatomy derived from structural magnetic resonance imaging scans. In accordance with previous literature, we found low gamma (28-40 Hz) amplitude modulations related to gait phases that occurred specifically in the leg motor cortical areas. Surprisingly, we also found that these gait-related amplitude modulations were detectable at a single trial level, which is a prerequisite for accurate decoding of gait events. We then calibrated a decoder that detected gait events from the reconstructed activity of cortical sources. Here, we present the first results of this non-invasive gait event decoding strategy.
Repetitive Ocular Vestibular Evoked Myogenic Potential (RoVEMP) Stimulation For Diagnosis of Myasthenia Gravis: Optimization Of Stimulation Parameters
Konrad P. Weber (Zürich | CH)
Early and accurate diagnosis is of great importance for the course and outcome of myasthenia gravis (MG). Recently, repetitive ocular vestibular evoked myogenic potential (RoVEMP) stimulation has been developed as a novel diagnostic tool for MG. Quantification of extraocular muscle response decrement after repetitive stimulation facilitates the often challenging diagnosis of MG. Comparison of different stimulation paradigms is needed to determine the most sensitive and specific parameters for detecting the characteristic RoVEMP decrement.
Repetitive bone-conducted oVEMPs were elicited in 18 MG patients and 20 healthy subjects. To determine the most sensitive and specific RoVEMP paradigm for decrement detection, we compared four different repetition rates (20Hz, 30Hz, 40Hz, 50Hz). In addition to the inferior oblique muscles, we recorded oVEMPs from the lateral rectus muscles.
Repetitive stimulation at all tested repetition rates with recordings from inferior oblique muscles allowed for effective differentiation between MG patients and healthy subjects. Among all repetition rates, 30Hz showed a trend towards superiority, with a sensitivity of 71% and a specificity of 94% (area under the curve (AUC) 0.88) when using the smaller decrement of the two eyes and -10% as cutoff. Considering the larger decrement for analysis (-9% as cutoff), sensitivity increased to 82%, but specificity decreased to 78% (AUC 0.81).
Our study suggests 30Hz repetitive oVEMP stimulation from the inferior oblique muscles as the most effective stimulation paradigm. Repetitive oVEMP stimulation with optimal parameters facilitates early and accurate diagnosis of ocular MG.
Characteristics of acute MRI examinations in focal non-convulsive status epilepticus – a retrospective analysis.
Lukas L. Imbach (Zurich | CH)
The mechanisms involved in the propagation and persistence of epileptic activity during status epilepticus (SE) are largely unclear. Proposed mechanisms include cortico-subcortical interactions resulting in a hyperactive wide-spread epileptogenic network. In this study, we set out to determine characteristic magnetic resonance imaging (MRI) findings during SE to identify affected brain structures during SE.
Materials and Methods
We retrospectively analyzed 641 consecutive patients with diagnoses of EEG-documented nonconvulsive SE according to Salzburg consensus criteria between March 15th 2001 and October 31st 2018. Inclusion criteria for MRI imaging were (I) MRI examination on the same day of the EEG or (II) between two consecutive EEG examinations in SE.
We found 77 cases with acute MRI examinations in SE fulfilling the inclusion criteria. In 50 cases MRI and EEG were performed on the same day. DWI-restrictions were found in 92.2% of all cases. Neocortical, hippocampal (with or without involvement of the amygdala) and thalamic alterations occurred in 64.9%, 81.8% and 55.8% of the patients, respectively. The lateralization of the electroencephalographic focus was identified correctly in the MRI in 80%, 71.4% and 79.1% of the cases. We found overlapping signals between the EEG discharges and the cortical DWI-restrictions in 96% of all cases. In 88.4% of the patients with thalamic DWI restrictions, the dorso-medial thalamus was affected.
Acute, reversible DWI restrictions in MRI are common during focal SE und reflect the localization of the epileptic focus in the EEG. Moreover, the MRI findings suggest the activation of subcortical structures, in particular the dorso-medial thalamus, suggesting a pathophysiological activation of a wide-spread epileptogenic network in SE.
Multi-modal intraoperative electrophysiological mapping of the anterior nucleus of the thalamus in refractory epilepsy
Lukas L. Imbach (Zurich | CH)
Purpose: Deep brain stimulation (DBS) in the anterior nucleus of the thalamus (ANT) has been shown to be effective in reducing seizure frequency in focal epilepsy. However, treatment response varies considerably on an individual level. Previous studies showed that the anti-epileptic effect of anterior thalamic DBS may be dependent on the stimulation site within the ANT and functional connectivity from imaging data showed differences in the default network in ANT DBS responders as compared to non-responders.
Method: In this pilot study we aim to implement a multimodal mapping of the anterior nucleus of the thalamus based on intra-operative montoring. We recorded intraoperatively 16-channel surface EEG, local field potentials (LFP), and micro-electrode signals along the implantation trajectory. Data post-processing included spike-frequency analysis, spectral analysis of local field potentials and cortico-subcortical coherence. Here we present intra-operative data from 6 DBS trajectories in 3 patients.
Results: Upon entry in the anterior nucleus of the thalamus, we observed increased theta band (4-8 Hz) LFP activity within the first 4.5mm after entry in the anterior thalamus (entry was predicted based on direct MRI-targeting). Spectral coherence analysis revealed increased theta band coupling between the LFP and ipsilateral temporal EEG electrodes with maximal coherence between 3 and 7mm after entry. MER recordings did not reveal a consistent pattern to delineate the target region.
Conclusion: LFP theta activity and LFP-EEG coherence between the ANT and temporal surface EEG showed a consistent pattern to delineate the anterior nucleus of the thalamus. This pilot data corroborates the feasibility to use intra-operative electrophysiological monitoring for mapping the ANT. These observations could be used in future studies as potential biomarkers to determine the stimulation site within the ANT and to enhance treatment response after DBS implantation.
The Swiss Narcolepsy Scale (SNS) and its Short Form (sSNS) for the discrimination of narcolepsy in patients with hypersomnolence: A cohort study based on the Bern Sleep-Wake Database
Anelia Dietmann (Bern | CH)
Previous studies reported high sensitivity and specificity of the Swiss Narcolepsy Scale (SNS) for the diagnosis of narcolepsy type 1. We used data from the Bern Sleep-Wake Database to investigate the discriminating capacity of both the SNS and the Epworth Sleepiness Scale (ESS) to identify narcolepsy type 1 and type 2 in patients with central disorders of hypersomnolence (CDH) or sleepy patients with obstructive sleep apnea (OSA). In addition, we aimed to develop a simplified version of the SNS.
We used data from the Bern Sleep-Wake Database. The validation of the diagnoses was based on the third edition of the International Classification of Sleep Disorders (ICSD-3).
We created the two-item short-form SNS (sSNS), based on the discriminative capability of the models including all possible combinations of the five questions of the SNS.
Using the previously published co-efficiencies, we confirmed the high capacity of the SNS in identifying narcolepsy type 1. The updated SNS (based on new co-efficiencies and cut-off) and the sSNS showed high capacity and were both superior to ESS in identifying narcolepsy type 1. The sSNS correlated significantly with the SNS (r = -0.897, p < 0.001). No scale showed sufficient discrimination for narcolepsy type 2.
This is the largest cohort study that confirms the discriminating power of SNS for narcolepsy type 1 in patients with hypersomnolence and the first study to assess its discriminative power for narcolepsy type 2. The easy-to-use and easy-to-calculate short-form scale has a high discriminating power for narcolepsy type 1 and may be used as screening tool, especially among general practitioners, to identify patients and accelerate their referral to a center of expertise.