University of Twente Student Theses


Early prognostication in postanoxic coma patients based on EEG with reduced electrode sets

Merkerk, M.N. van (2021) Early prognostication in postanoxic coma patients based on EEG with reduced electrode sets.

[img] PDF
Abstract:INTRODUCTION: Continuous electroencephalography (cEEG) is increasingly used for prognostication after cardiac arrest (CA). The first 24 hours of EEG after resuscitation are reported to have the highest prognostic value for neurological outcome. Easy-to-apply adhesive electrodes, like the BrainStatus, or a reduced number of electrodes can both contribute to early start of cEEG monitoring. Furthermore, quantitative EEG analysis can be a helpful tool to support prediction of neurological outcome. OBJECTIVE: During this study, we investigated the possibilities for using adhesive electrodes and reduced montages for cEEG monitoring while preserving the predictive value. METHODS: Continuous EEG was obtained from postanoxic comatose patients after CA admitted to the ICU for all parts of the study. Monitoring with the BrainStatus was done prospectively, simultaneously with conventional EEG monitoring, and with intermittent check-ups. A visual scoring application was used to score EEG epochs from the BrainStatus. Secondly, a total of 221 EEG epochs was visually scored in a 4-channel frontotemporal montage. All scored classifications were compared to the classifications in the 9-channel bipolar montage. Lastly, a logistic regression (LR) model was used for the prediction of neurological outcome. Features extracted from reduced electrode montages were used as input, and model performance was compared to the full montage. RESULTS: Classification agreement of EEG background patterns between the BrainStatus and the bipolar montage was moderate (κ = 0.48), based on six patients. Agreement between the frontotemporal montage and the bipolar montage was substantial (κ = 0.76). Sensitivity and specificity for predicting neurological outcome were not significantly different for the frontotemporal montage compared to the bipolar montage. Baseline performance of the LR model for the bipolar montage equals an AUC of 0.897 (0.885-0.908), Se100 of 77% (74-79) for predicting poor outcome and Se95 of 71% (68-74) for good outcome at 12 hours after CA. For 24 hours after CA, the AUC equals 0.879 (0.873-0.886), with Se100 of 63% (62-65) for predicting poor outcome and Se95 of 50% (47-52) for good outcome. Performance of the model on the frontotemporal montage significantly increases for Se100 at 12h after CA, and for AUC and Se100 at 24h after CA. CONCLUSION: Intermittent check-ups did not improve the signal quality of the BrainStatus recordings. Reducing the number of electrodes from 9 to 4 does not affect EEG classification or prognostic accuracy in patients with postanoxic coma. For qEEG analysis, the 4-channel montage showed significant better performance for outcome prediction than the 9-channel montage.
Item Type:Essay (Master)
Faculty:TNW: Science and Technology
Subject:44 medicine
Programme:Technical Medicine MSc (60033)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page