Inter-Subject Prediction of Pediatric Emergence Delirium Using Feature Selection and Classification from Spontaneous EEG Signals

Abstract

Pediatric emergence delirium (PED) is a troublesome clinical phenomenon, which may cause serious safety problems and sequelae. PED prediction detects emergence delirium earlier before pediatric patients regain consciousness, providing early warning to enable timely treatment, thereby reducing the occurrence of adverse events. We proposed a feature selection and classification (FSC) method fusing marine predator algorithm (MPA) and k-nearest neighbors (KNN), named MPA-KNN-FSC, to achieve inter-subject PED prediction with spontaneous electroencephalogram (EEG) signals. To verify the feasibility and generalizability of this method, spontaneous 64-channel EEG signals were collected from 20 paediatric patients (1 to 12 years old) after anaesthetic surgery, and processed to extract 1792 features containing time domain, frequency domain, time-frequency domain and nonlinear features. MPA was then exploited to search the optimal feature subset and KNN parameters used for predicting PED. To resolve the sample imbalance problem, the MPA's fitness was constructed with the average area under the receiver operating characteristic curve (AUROC) obtained by the KNN classifier with five-fold cross-validation. Finally, the MPA-KNN-FSC method with AUROC-based fitness was compared with the FSC methods based on 10 well-known metaheuristic algorithms as well as the MPA-KNN-FSC method based on accuracy. The proposed AUROC-based MPA-KNN-FSC method achieved the best inter-subject PED prediction accuracy (77.90 ± 2.59%) and AUROC (0.871 ± 0.017) when compared with the FSC methods with ten metaheuristic algorithms and the accuracy-based MPA-KNN-FSC method. These results demonstrate the feasibility, effectiveness and generalizability of the proposed AUROC-based MPA-KNN-FSC method for automatic and unsupervised physiological monitoring of pediatric patients.

Note:
Funding Information: This work was supported in part by the National Key Research & Development Project (No. 2020YFC2003903) and the Key-Area Research and Development Program of Guangdong Province (No. 2019B010152001).

Conflict of Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Ethical Approval: The study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Zhongshan Ophthalmic Center (ID No. 2020KYPJ092). Prior to the collection of EEG data, all participants' parents or designated guardians had obtained informed consent.

Keywords: Pediatric emergence delirium, electroencephalogram (EEG), marine predator algorithm, k-nearest neighbors, feature selection and classification