Article Text

Original research
Detection of early neurological deterioration using a quantitative electroencephalography system in patients with large vessel occlusion stroke after endovascular treatment
  1. Yujia Yan1,2,3,4,
  2. Xingwei An1,3,4,
  3. Yuxiang Ma2,
  4. Zeliang Jiang5,
  5. Yang Di1,3,4,
  6. Tingting Li1,3,4,
  7. Honglin Wang1,3,4,
  8. Hecheng Ren2,
  9. Lin Ma2,
  10. Bin Luo2,
  11. Ying Huang2
    1. 1Tianjin Key Laboratory of Brain Science and Neuroengineering, Academy of Medical Engineering and Translational Medicine, Tianjin University, Tianjin, People's Republic of China
    2. 2Department of Neurosurgery, Tianjin Huanhu Hospital, Tianjin, People's Republic of China
    3. 3Haihe Laboratory of Brain-Computer Interaction and Human-Machine Integration, Tianjin, People's Republic of China
    4. 4State Key Laboratory of Advanced Medical Materials and Devices, Tianjin University, Tianjin, People's Republic of China
    5. 5Department of Psychology, Hebei Normal University, Shijiazhuang, Hebei, People's Republic of China
    1. Correspondence to Professor Ying Huang; yinghuang0000{at}163.com

    Abstract

    Background Early neurological deterioration (END) is a serious complication in patients with large vessel occlusion (LVO) stroke. However, modalities to monitor neurological function after endovascular treatment (EVT) are lacking. This study aimed to evaluate the diagnostic accuracy of a quantitative electroencephalography (qEEG) system for detecting END.

    Methods In this prospective, nested case–control study, we included 47 patients with anterior circulation LVO stroke and 34 healthy adults from different clinical centers in Tianjin, China, from May 2023 to January 2024. Patients with stroke underwent EEG at admission and after EVT. The diagnostic accuracy of qEEG features for END was evaluated by receiver operating characteristic curve analysis, and the feasibility was evaluated by the percentage of artifact-free data and device-related adverse events.

    Results 14 patients with stroke had END (29.8%, 95% CI 16.2% to 43.4%), with most developed within 12 hours of recanalization (n=11). qEEG features showed significant correlations with National Institutes of Health Stroke Scale score and infarct volume. After matching, 13 patients with END and 26 controls were included in the diagnostic analysis. Relative alpha power demonstrated the highest diagnostic accuracy for the affected and unaffected hemispheres. The optimal electrode positions were FC3/4 in the unaffected hemisphere, and F7/8 and C3/4 in the affected hemisphere. No device-related adverse events were reported.

    Conclusion The qEEG system exhibits a high diagnostic accuracy for END and may be a promising tool for monitoring neurological function. The identification of optimal electrode positions may enhance device convenience.

    Clinical trial registration ChiCTR 2300070829.

    • EEG
    • Thrombectomy
    • Stroke

    Data availability statement

    Data are available upon reasonable request.

    http://creativecommons.org/licenses/by-nc/4.0/

    This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

    Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    WHAT IS ALREADY KNOWN ON THIS TOPIC

    • Electroencephalography (EEG) has the potential to detect large vessel occlusion (LVO) in patients with stroke in a prehospital setting. However, its effectiveness in monitoring neurological function after endovascular treatment (EVT) remains unknown.

    WHAT THIS STUDY ADDS

    • In the present study, quantitative EEG (qEEG) demonstrated high diagnostic accuracy for early neurological deterioration (END).

    • qEEG features showed significant correlations with neurological function. Both hemispheres exhibited similar results regardless of the side of the ischemic lesion.

    • No device-related adverse events were observed.

    HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

    • This study suggests an available modality for monitoring neurological function in LVO stroke patients. Short recording time, and easy application of the electrode cap are clear advantages of this technique and provide evidence of the potential transition from bench to bedside.

    Introduction

    The incidence rate of early neurological deterioration (END) broadly ranges from 10% to 40% owing to the different criteria of definitions and therapeutic modalities.1 2 Nevertheless, END has consistently been associated with a poor prognosis.3–5 The practice primarily relies on physicians’ assessment of the National Institutes of Health Stroke Scale (NIHSS) score; therefore, there has been a widespread delay in diagnosis. Electroencephalography (EEG), with a higher temporal resolution,6 may be a promising monitoring tool for neurological function in patients with stroke. EEG has the potential to detect large vessel occlusion (LVO) stroke in the prehospital setting.7 However, studies of the application of EEG after recanalization have not been conducted. Thus, the primary objective of this study was to evaluate the feasibility and diagnostic accuracy of the quantitative EEG (qEEG) system for END detection.

    Methods

    Study design and participants

    We conducted a prospective, nested case–control study consisting of three components (figure 1). We enrolled 62 patients with anterior circulation LVO stroke in Tianjin Huanhu Hospital and 53 healthy adults in Faculty of Medicine, Tianjin University from May 2023 to January 2024. All participants were right-handed.8 Participants with pre-existing neurological conditions (eg, previous ischemic stroke/cerebral hemorrhage, brain tumor, seizures) and a history of psychiatric or treatment with central nervous system (CNS) suppressive medications, such as sedatives and antipsychotics, were initially excluded. Among the patients with stroke, patients receiving endovascular treatment (EVT) under general anesthesia (n=3), with unsuccessful recanalization at the final angiogram (modified Treatment In Cerebral Infarction (mTICI) <2b,9 n=2), with poor EEG recording quality (n=1), and who received CNS suppressive medication (diazepam) for convulsive status epilepticus after EVT (n=1) were excluded. In section I, 47 eligible patients with stroke and 34 healthy controls matched for age, sex, and years of education were included to describe the qEEG features.

    Figure 1

    Flow diagram of study. CNS, central nervous system; EEG, electroencephalography; END, early neurological deterioration; EVT, endovascular treatment; LVO, large vessel occlusion; qEEG, quantitative EEG.

    In section II, for the nested case–control study design, patients with END were defined as cases. For each case, two controls without END were randomly selected and matched for age (±3 years), sex, and years of education (±3 years) from the same cohort. A total of 13 cases and 26 controls were included in END diagnostic analyses.

    In section III, based on the above results, we identified the qEEG features with the highest diagnostic accuracy as indicators of interest (IOIs) through receiver operating characteristic (ROC) curve analysis to identify the optimal electrode positions among all electrodes.

    Electroencephalography acquisition and preprocessing

    Patients with stroke underwent EEG at the bedside on admission (baseline) and after EVT (after recanalization). Patients who developed END underwent immediate EEG (the delay between EEG acquisition and onset of END was no more than 20 min), whereas patients who did not develop END underwent EEG at approximately the 12th and 24th hour after recanalization (if END had occurred within 12th hour, the 24th hour EEG acquisition was canceled). The process of EEG acquisition is shown in online supplemental figure S1,A.

    Supplemental material

    All participants (patients with stroke and healthy controls) were in the resting state and lying on a bed with closed eyes. The qEEG system (Grael EEG, Australia) included an Ag/AgCl (silver-silver chloride) semi-dry electrode cap, an EEG amplifier, a data receiver, and computer software. All electrodes were placed according to the international 10–20 system, covering the anterior circulation vascular territory (online supplemental figure S1,B). EEGs were performed by specialist researchers (blinded to the participants’ information) for 10–15 min (including preparatory steps). The detailed methodology for preprocessing is provided in the online supplemental material. Each sample was retained as artifact-free data for at least 3 min for quantitative analysis.

    Quantitative analysis

    Three quantitative features were considered: spectral power, coherence (Coh), and small world (SW). All the features were calculated for both hemispheres. The relative band power was calculated as the ratio of absolute band power to total power.10 The delta/alpha ratio (DAR), theta/alpha ratio (TAR) and (delta+theta)/(alpha+beta) ratio (DTABR) were also calculated. To evaluate brain symmetry, we determined the pairwise-derived Brain Symmetry Index (pdBSI) in the frequency range of 1–45 Hz.11 Coh and SW were used to describe the functional connectivity of the brain networks. The detailed methodology of the quantitative analysis is provided in the online supplemental material.

    Data collection and criteria

    Demographic and clinical data, including age, sex, education, and medical history, were systematically collected. Time parameters, including the time of stroke onset (or time of last known well), initial MRI scanning, EEG acquisitions (baseline and after EVT), puncture, recanalization, and END onset (within 12 hours or 12–24 hours after recanalization), were systematically recorded. The infarct volume was calculated from the initial MRI scan (diffusion-weighted imaging lesion). The NIHSS scores were assessed by a neurologist at the time of EEG. Definition of anterior circulation LVO stroke was based on previous studies.7 12 Etiological subtypes of ischemic stroke were classified according to the Trial of Org 10 172 in Acute Stroke Treatment criteria.13 EVT included mechanical thrombectomy and bridging thrombolysis. All procedures were performed according to standard protocols.14 The recanalization status was assessed using the mTICI scale, and successful reperfusion was defined as mTICI 2b or 3 on the final angiogram.9

    Outcomes

    The primary outcome was the diagnostic accuracy of the qEEG features for END, evaluated using the area under the ROC curve (AUC). END was defined as an increase of ≥4 points in total NIHSS score or ≥1 points in 1a (level of consciousness) or death between the baseline and within 24 hours after recanalization,1 3 for clear (eg, parenchymal hemorrhage, progression of initial infarction or new infarctions, malignant edema, procedural complications) or unexplained reasons.15 The secondary outcome was the feasibility of the qEEG system, expressed as the percentage of artifact-free data and device-related adverse events.

    Statistical analysis

    The Shapiro-Wilk test was used to assess normality. Continuous variables are reported as median (IQR). Categorical variables are reported as frequencies (percentages). We used the Mann-Whitney U tests for continuous variables to examine differences between patients with stroke and healthy controls and between the ‘END’ and ‘non-END’ samples. The Pearson χ2 or Fisher exact test were used for categorical variables. Spearman’s linear correlations were tested between qEEG features, NIHSS scores, and infarct volumes. In section II, considering that baseline qEEG features primarily hold predictive value for END, we discarded them from the diagnostic analysis. After matching for age, sex, and education, the EEG samples (after recanalization) were labeled according to whether ‘END’ had developed. In total, 13 EEG samples were labeled with ‘END’ and 35 with ‘non-END’ as the matched cohort for diagnostic analysis. The diagnostic accuracy was evaluated by AUC, and the cut-off point was determined by the Youden index. Sensitivity, specificity, positive predictive value, and negative predictive value with 95% confidence intervals (95% CI) were also reported. The qEEG features with the highest AUC values were selected as the IOIs. The optimal electrode position was identified based on the highest AUC value of the IOI among all electrodes. All statistical analyses were performed using R (version 4.2.1) and SPSS (version 25.0) software. The Power (1-β err prob) of IOI was performed by post-hoc analysis with two-sided testing (α=0.05) using the G*Power software (version 3.1). Two-sided with a P value <0.05 was considered statistically significant.

    Results

    The characteristics and process measures of the patients with stroke and healthy controls are shown in table 1. There were no significant differences in demographic features between the two cohorts (P>0.05). In healthy controls, qEEG features showed no significant differences between the hemispheres. Thus, the average of the bilateral hemispheres could be used as a reference (figure 2, red line) for comparison with patients with LVO. The results in section I showed significant differences in the relative band powers (delta, theta, alpha, and beta bands), Coh (delta and theta bands), and SW (delta, theta, alpha, and beta bands) between the healthy controls and patients with stroke in both the unaffected and affected hemispheres (figure 2). Brain symmetry results showed that pdBSI values were significantly higher in patients with stroke for theta (P=0.004), beta (P=0.002), and gamma (P=0.025) bands than in healthy controls. The detailed results are shown in online supplemental table S1.

    Figure 2

    Quantitative electroencephalography features between the healthy subjects and large vessel occlusion stroke patients. Red line represents the average of bilateral hemispheres in healthy subjects. *Statistically significant (P<0.05).

    Table 1

    Characteristics and process measures of stroke patients and healthy controls.

    For patients with stroke, the median time from onset (or time of last known well) to the baseline EEG acquisition was 7.6 (IQR 5.9–9.5) hours. Fourteen patients had END (29.8%, 95% CI 16.2% to 43.4%), with most developed within 12 hours of recanalization (n=11, 78.6%). Detailed experimental time parameters and patient characteristics, including medical history, stroke information, treatment profiles, and causes of END, are presented in table 1. The correlations among qEEG features, NIHSS scores, and infarct volumes are shown in online supplemental table S2. The values of the frequency band powers in the alpha, beta, and gamma bands were significantly negatively correlated with the NIHSS score and infarct volume, whereas DAR, TAR, and DTABR were positively correlated (online supplemental figure S2,S3). Regarding secondary outcomes, the median percentage of artifact-free data was 62% (IQR 47–71%), and no device-related adverse events were reported.

    For section II, the ‘END’ samples exhibited significantly higher values in relative delta power, DAR, and TAR in bilateral hemispheres, whereas lower values in relative alpha power (bilateral hemispheres) and SW in alpha band (unaffected hemisphere) compared with the ‘non-END’ samples (online supplemental table S3, matched cohort). The results of the diagnostic measurements for the primary outcome are reported in online supplemental table S4. The AUCs of relative alpha power were 0.86 and 0.85 in the affected and unaffected hemispheres, respectively, which were selected as the optimal indicators for END diagnosis (figure 3A). Powers (1-β err prob) of relative alpha power were 0.84 and 0.95 in the affected and unaffected hemispheres, respectively.

    Figure 3

    Identification of the optimal qEEG features and electrode locations. (A) ROC curves for bilateral hemispheric qEEG features. (B) Difference in relative alpha power between the END and non-END samples. The boxes represent the upper and lower quartiles, the short black line inside the box represents the median, the whiskers represent the 5th and 95th percentiles, and the circles represent the extreme values. (C) Scatterplots of significant correlation between relative alpha power and NIHSS score, and infarct volume, respectively. (D) Brain topographic mapping based on the AUC values of relative alpha power for each electrode. The gray boxes represent the optimal electrode positions. AUC, area under the ROC curve; END, early neurological deterioration; NIHSS, National Institutes of Health Stroke Scale; qEEG, quantitative electroencephalography; ROC, receiver operating characteristic.

    In section III, the optimal electrode positions with the highest AUC value for relative alpha power were FC3/4 (AUC 0.88) in the unaffected hemisphere and F7/8 (AUC 0.85) and C3/4 (AUC 0.85) in the affected hemisphere (figure 3D). Detailed results are presented in online supplemental table S5.

    Discussion

    In the present study we found that, using a qEEG system in patients with LVO stroke, a high diagnostic accuracy for END was achieved, and no device-related adverse events were reported. Furthermore, certain qEEG features correlated with the NIHSS score and infarct volume. Both hemispheres showed similar results regardless of the side of the ischemic lesion.

    EEG can serve as a diagnostic tool and biomarker to predict outcomes in patients with ischemic stroke. However, few studies have analyzed the features of LVO stroke. The qEEG features can be briefly summarized in a dichotomous category: increased low spectrum (delta, theta) and decreased high spectrum (alpha, beta). These values correlate with the severity of neurological impairment. Cramer et al16 revealed a positive correlation between DTABR and NIHSS score. Jonathan et al7 17 reported that an increased TAR could identify patients with LVO stroke in a prehospital setting. Several studies have reported different ischemic stroke subtypes and variations in the time of acquisition; however, the results were consistent with those of the dichotomous category.18–20

    Despite differences in the causes of END, qEEG features consistently showed the aforementioned dichotomous results, particularly in terms of spectral features. Functional regions constantly interact with each other and maintain a dynamic balance between segregation and integration characteristics.21 Brain function abnormalities are associated with dysfunctional connectivity and network structures. Several studies have reported the correlation between functional connectivity and outcomes, and SW and Coh can be used as prognostic biomarkers.22 23 In this study, SW showed lower values in the alpha band in the END samples than in the non-END samples. A decrease in the SW represents an increase in the integration of the brain network. However, the AUC values of the SW parameters did not reach the diagnostic accuracy priority; thus, we discarded the parameters of functional connectivity.

    Furthermore, both hemispheres exhibited similar results, regardless of the ischemic lesion side. Given the absence of mechanistic studies, we hypothesized, based on evidence from the literature, that alterations in the unaffected hemispheres are mediated by the transcallosal pathway.24 Bilateral abnormal alterations are advantageous for clinical applications and can enhance generalizability beyond bilateral monitoring. Notably, alterations in the unaffected hemisphere are not the same as those in the affected hemisphere. We calculated the pdBSI to evaluate brain symmetry. The results showed significantly higher pdBSI values for the theta, beta, and gamma bands. Several theoretical models, such as the vicariation, interhemispheric competition, and bimodal balance-recovery models, explain the pathophysiological mechanisms between the affected and unaffected hemispheres.25 Interhemispheric imbalance is an important target for neuromodulation in non-invasive brain stimulation techniques aimed at promoting neurorehabilitation.

    Advantages of the technique

    The diagnosis of END relies on physicians or nurses to assess neurological status, leading to an almost unavoidable delay.26 The direct assessment of neurological function, short recording time, and easy application are clear advantages of this technique. It provides a modality for monitoring neurological function in patients after EVT and may potentially help reduce diagnostic delays. In future studies, exploring the correlation between physiological parameters, such as blood pressure, intracranial pressure, and fluid parameters, and qEEG features can provide a comprehensive monitoring approach for patients and serve as a possible selection criterion for future applications of therapeutic and neuroprotective agents. Moreover, identifying the optimal electrode position makes single-/multiple-electrode monitoring modes possible. Not restricting bilateral monitoring and not limiting the frequency of monitoring are the advantages of this device. Manganotti et al20 revealed that qEEG features correlated with the CT perfusion hypoperfused volume and the ischemic core. Expanding beyond postoperative monitoring, the correlation between EEG and imaging features provides optimism for future applications in preoperative assessment.

    Limitations

    Compared with continuous EEG, the saline electrodes used in this study could not facilitate long-term EEG recordings. Changing to dry/wet electrodes may solve this issue. The inability to visualize results in real time is a notable limitation of this technique because offline processing hinders its application. This issue can be addressed by algorithms that intercept 5–10 s segments of EEG data for continuous quantitative analysis, enabling an up-to-date display of qEEG features. Although this will result in a computational delay of a few seconds, optimization of the algorithm can minimize this effect. Another noteworthy limitation is that this study focused on quantitative analyses; the results of the qualitative analyses were not presented. In particular, epileptic discharges were monitored in nine (19.1%) patients during EEG recording after EVT (non-convulsive status), including sporadic epileptiform discharges, spike-and-wave, and rhythmic sharp activity. Most LVO stroke cases are associated with impaired consciousness. Abnormal EEG features are difficult to categorize, and often assumed to be due to ischemic stroke, leading to missed diagnoses of epilepsy, particularly non-convulsive seizures.27 Additionally, in patients with stroke, altered blood flow appears to lead to abnormal neural activity, which can further affect blood flow through the neurovascular unit (neurovascular coupling).28 It is difficult to clarify their causal relationship; therefore, this study cautiously concludes with relevant inferences. Finally, limited by the sample size of this study, we need to conduct a larger multicenter prospective study to validate our results.

    Tips to be noted

    Patients with LVO stroke frequently present with consciousness disorders. When using an electrode cap, the correct placement and appropriate tightness of the jaw strap should be ensured to avoid suffocation. During EEG acquisition, researchers should remain vigilant of the patient’s status, and the experiment should be terminated if deterioration is in progress. Although appropriate limb restraint can be provided, the use of sedatives should be avoided to prevent masking of the patient’s condition and corruption of the EEG data.

    Conclusions

    In this prospective study, the qEEG system exhibits high diagnostic accuracy for END, and its features correlate with NIHSS score and infarct volume. The identification of optimal electrode positions may reduce the number of electrodes and enhance device convenience. Not restricting to bilateral monitoring with high diagnostic accuracy of END, makes it a promising tool for monitoring neurological function. The application of the qEEG system in patients with LVO stroke should be validated in future studies.

    Data availability statement

    Data are available upon reasonable request.

    Ethics statements

    Patient consent for publication

    Ethics approval

    Tianjin Huanhu Hospital Ethics Committee approved the data collection procedures that involved the study participants (JH2022-104). Written informed consent for participation was obtained before EEG acquisition in accordance with the national legislation and the institutional requirements.

    Acknowledgments

    The authors thank all study participants for the invaluable contribution. Special thanks to Professor An and MED-BCI group for technical support with the quantitative electroencephalography system.

    References

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Footnotes

    • YY, XA and YM are joint first authors.

    • Contributors YY, XA and YM: conception and design of the study. YY and YM: data collection and manuscript drafting. YY, ZJ, YD, TL, and HW: analysis and interpretation of data. HR, BL, LM: data collection. YH and XA: study supervision, revision, editing and approval of final draft. Guarantor: YY.

    • Funding This project was supported by the Tianjin Science and Technology Program (20230108, 20JCZDJC00620) and Tianjin Municipal Health Commission (2023026).

    • Competing interests None declared.

    • Provenance and peer review Not commissioned; internally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.