好色先生

好色先生

Explore the latest content from across our publications
Join The AAN

Log In

Forgot Password?
Create New Account

Loading... please wait

Abstract Details

Development of a Machine Learning Model for Supporting Brain Death Using EEG Suppression Ratio
Epilepsy/Clinical Neurophysiology (EEG)
P11 - Poster Session 11 (5:30 PM-6:30 PM)
1-012

To develop a machine learning model for supporting brain death using EEG suppression ratio (SR).

EEG has been used ancillary testing for brain breath diagnosis. Electrocerebral inactivity (ECI) is mandatory for declaring brain death in Korea, but there are challenging situations where we don’t have a clear-cut ECI. The delayed diagnosis may detain the process of organ transplantation.
The SR was analyzed with installed software (Persyst® v13) in EEG machine. Each EEG leads’ data were divided into 10 seconds epochs. Suppression was defined as <3μv amplitude lasting ≥0.5 seconds, which is the default threshold values of the software. 180 values (1800 seconds/10 seconds) of SR during 30 minutes routine EEG were obtained and averaged. Therefore, the SR was calculated in percentage. 180 patients were included in this study consisted of 81 patients with brain death and 99 patients with unresponsive wakefulness syndrome. The SRs on EEG were analyzed with machine learning (ML) algorithm, which included 70% of training set and 30% of test set for brain death prediction. Supervised ML models including quadratic discriminant analysis (QDA), naïve bayes, logistic regression, linear discriminant analysis, and light gradient-boosting machine were applied.
Among supervised ML models, QDA model demonstrated the highest performance and precisely defined cutoff value of the SR for brain death that was >74.76% as the maximum SR index, achieving AUC value of 0.9806 ± 0.02. Additionally, QDA model also outperformed the other models in terms of accuracy, precision, recall, and F1-score, which were 0.9120 ± 0.05, 0.9304 ± 0.04, 0.9120 ± 0.05 and 0.9101 ± 0.05, respectively.
Development of a machine learning model for supporting brain death using SR may be achievable via QDA model. This will allow us not to delay the diagnosis of brain death, therefore, may expedite the process of organ transplantation.
Authors/Disclosures
GyeongMo Sohn, MD (Inje University – Haeundae Paik Hospital)
PRESENTER 		Dr. Sohn has nothing to disclose.
Sung Eun Kim No disclosure on file