University of Twente Student Theses


Bedside quantitative cEEG monitoring on the Intensive Care for comatose patients after cardiac arrest.

Lievestro, H.J. (2019) Bedside quantitative cEEG monitoring on the Intensive Care for comatose patients after cardiac arrest.

[img] PDF
Abstract:Introduction: Electroencephalography (EEG) patterns within 24 hours after cardiac arrest (CA) have shown to reliably predict neurological outcome. These recordings may be simplified by using a less extensive electrode set. Subsequently, algorithm-aided EEG analysis may be used to identify regions of interest in high volumes of data and support untrained readers. In this dual study, we compared a sub-hairline EEG headband (BrainStatus) to a 9-channel full-head (FH) electrode set. Also, a 4-channel frontotemporal (FT) montage was derived from the FH set. Next, these FH/FT montages were used to re-train the existing cerebral recovery index (CRI) algorithm. Methods: EEGs were simultaneously recorded with the FH set and the BrainStatus in consecutive adult patients admitted after CA. EEG patterns were visually scored in 5-minute epochs at 24 hours after CA, and scoring agreement was evaluated using confusion matrices and Cohen’s kappa. The CRI algorithm was re-trained on 79 post-arrest EEGs from March 2014 through August 2018. Neurological outcome was dichotomised as good (Cerebral Performance Category (CPC) 1-2) or poor (CPC 3-5). For both the FH and FT montage, a random forest classifier was trained for each hour after CA and new thresholds were established. Sensitivity and specificity of each montage were evaluated. Results: Between July 2018 and January 2019, EEGs from 22 patients were recorded. At 24 hours after CA, patterns of 21 patients were available. The agreement for background patterns between the BrainStatus and FH set was fair (kappa = 0.32). Between the FT and FH montage, the agreement was substantial (kappa = 0.77). In a test set of 79 patients, the re-trained CRI with the FH montage predicted poor outcome at 24 hours after CA with sensitivity of 0.86 and specificity of 0.20. With the FT montage, sensitivity was 0.92 and specificity was 0.35. Conclusion: Visual classification of EEG patterns in patients with postanoxic coma with a FH electrode set cannot be replaced with the BrainStatus. However, four frontotemporal channels provide enough agreement, and can reliably be used for visual scoring of EEG background patterns. The specificities of the CRI algorithms, re-trained on the FH and FT montage, to predict poor neurological outcome at 24 hours after CA are unacceptably lower than the original CRI. Further optimisation of this limited electrode EEG monitoring set-up is recommended.
Item Type:Essay (Master)
Faculty:TNW: Science and Technology
Subject:44 medicine, 50 technical science in general
Programme:Technical Medicine MSc (60033)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page