Effective classification for neonatal brain injury using EEG feature selection based on elastic net regression and improved crow search algorithm
Neonatal brain injury carries the risk of neurological sequelae such as epileptic seizures, cerebral palsy, intellectual disability, and even death. Classification methods based on electroencephalography (EEG) signals and machine learning algorithms are crucial for assessing neonatal brain injury. H...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
PeerJ Inc.
2025-07-01
|
| Series: | PeerJ Computer Science |
| Subjects: | |
| Online Access: | https://peerj.com/articles/cs-3000.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Neonatal brain injury carries the risk of neurological sequelae such as epileptic seizures, cerebral palsy, intellectual disability, and even death. Classification methods based on electroencephalography (EEG) signals and machine learning algorithms are crucial for assessing neonatal brain injury. However, classification methods that utilise all features from the original EEG signals may result in lengthy training and classification times, thereby reducing the performance of the classification system. This article presents a novel classification system with a proposed feature selection method for assessing neonatal brain injury, in which the feature selection method is combined using elastic net (EN) regression and an improved crow search algorithm (ICSA), named EN-ICSA. In the EN-ICSA method, EN regression is used to conduct the pre-screening of features. The ICSA is utilised to select the essential figures further by introducing the dynamic perception probability for deciding whether to search locally or globally, a novel neighbor-following strategy for the local search and a global search strategy according to the crow’s search experience, resulting in accelerating the search efficiency while effectively avoiding falling into local optima. Experimental results demonstrate that the proposed system, based on support vector machine (SVM) with the EN-ICSA feature selection method, performs exceptionally well compared to other traditional machine learning and feature selection methods, achieving an accuracy of 91.94%, precision of 92.32%, recall of 89.85%, and F1-score of 90.82%. |
|---|---|
| ISSN: | 2376-5992 |