Continuous EEG Predicts Delayed Cerebral Ischemia After SAH

Abstract

Code and data used in the prospective study:

Rosenthal ES, Biswal S, Zafar SF, O'Connor KL, Bechek S, Shenoy AV, Boyle EJ, Shafi MM, Gilmore EJ, Foreman BP, Gaspard N, Leslie-Mazwi TM, Rosand J, Hoch DB, Ayata C, Cash SS, Cole AJ, Patel AB, Westover MB. Continuous electroencephalography predicts delayed cerebral ischemia after subarachnoid hemorrhage: A prospective study of diagnostic accuracy. Ann Neurol. 2018 May;83(5):958-969. doi: 10.1002/ana.25232. Epub 2018 May 16. PMID: 29659050; PMCID: PMC6021198.

The data and code are available here: https://github.com/bdsp-core/SAH-Annals-2018

Background

This study examined cEEG's ability to predict delayed ischemic neurologic decline (DIND) and delayed cerebral ischemia (DCI) in subarachnoid hemorrhage patients. The dataset includes cEEG recordings and clinical metadata from 103 patients monitored at Massachusetts General Hospital between 2013-2015, along with analysis code used to evaluate specific EEG features as predictors of DCI.

Summary of Key Findings:
The study demonstrated that continuous EEG monitoring can accurately predict delayed cerebral ischemia in subarachnoid hemorrhage patients before clinical symptoms appear. Key findings include:

• High predictive accuracy: cEEG showed >95% sensitivity and >80% specificity for predicting subsequent DIND/DCI
• Early warning: The median time between EEG changes and clinical DCI was 1.9 days, with 82% of cases having >12 hours of warning
• Multiple predictive features: Both background EEG deterioration and new epileptiform abnormalities were predictive, with late-onset or worsening epileptiform patterns showing particularly strong association
• Risk-stratified performance: The predictive value remained robust across different baseline risk levels, though positive predictive value was higher in high-risk patients (94%) compared to low-risk patients (76%)

This data and code is provided to enable reproduction of the findings and further investigation of EEG-based early warning algorithms for DCI. The included code implements the multi-state survival analysis and diagnostic accuracy assessments described in the original publication.

Methods

See the original paper. 

Data Description

The repository contains the complete dataset and analysis code from Rosenthal et al. (2018) examining continuous EEG prediction of delayed cerebral ischemia. The primary source data file "sah_data_MASTER_FOR_ANALYSIS_v5.xlsx" contains the clinical and EEG monitoring data collected from 103 subarachnoid hemorrhage patients. Additional data files (.txt, .csv, and .mat formats) are intermediate files generated during the analysis process.

The code base consists of:

• Three R scripts performing multi-state survival analysis
• 17 MATLAB analysis scripts (prefixed with "a_step") that generate the paper's figures and results
• Multiple MATLAB helper functions (prefixed with "fcn") providing utilities for data processing and visualization
• Generated output files including figures (.png, .fig) and statistical results

Usage Notes

The analysis must be run in a specific sequence to reproduce the paper's results:

1. First execute the R scripts in this order:

• mState_Bootstrap_SeSpPpvNpv.R
• mState_Overall.R
• mState_predictionCurves.R

2. For the MATLAB analysis, users have two options:

• Run a_000_OverallStats.m, which will automatically execute all analysis steps in sequence
• Run individual analysis scripts manually in alphabetical order (a_step0b through a_step10)

The code was developed during the research process and while not optimized for production use, it provides a complete record of the analysis methods. Researchers may find it particularly useful for:

• Reproducing the study's findings
• Understanding the implementation of multi-state survival analysis
• Adapting the methods for similar clinical prediction studies

Ethics

In this dataset, all identifiable patient information has been removed.

Acknowledgements

Please cite this paper if you use the data or code:

Rosenthal ES, Biswal S, Zafar SF, O'Connor KL, Bechek S, Shenoy AV, Boyle EJ, Shafi MM, Gilmore EJ, Foreman BP, Gaspard N, Leslie-Mazwi TM, Rosand J, Hoch DB, Ayata C, Cash SS, Cole AJ, Patel AB, Westover MB. Continuous electroencephalography predicts delayed cerebral ischemia after subarachnoid hemorrhage: A prospective study of diagnostic accuracy. Ann Neurol. 2018 May;83(5):958-969. doi: 10.1002/ana.25232. Epub 2018 May 16. PMID: 29659050; PMCID: PMC6021198.

Conflicts of Interest

None