Computer vision for automated seizure detection and classification: A systematic review.
Brandon M Brown1, Aidan M H Boyne1, Adel M Hassan1
1Department of Neurology, Baylor College of Medicine, Houston, Texas, USA.
Epilepsia
|March 1, 2024
Summary
Computer vision (CV) offers efficient video seizure detection and classification. While promising for epilepsy care, standardized validation and privacy concerns hinder widespread adoption.
Area of Science:
- Computer Vision
- Epilepsy Research
- Medical Technology
Background:
- Video seizure detection and classification using computer vision (CV) is an emerging, cost-effective technology.
- Epilepsy monitoring units face resource limitations, highlighting the need for automated analysis tools.
Approach:
- A systematic literature review was conducted on PubMed, Embase, and Web of Science (2000-2023).
- 45 studies met inclusion criteria, focusing on CV model architectures for video seizure analysis.
- Excluded were reviews, nonhuman studies, and those with insufficient data quality.
Key Points:
- CV models demonstrate significant growth and impressive accuracy in seizure detection and classification.
- CV offers rapid, scalable detection, potentially reducing SUDEP (sudden unexpected death in epilepsy) and easing monitoring burdens.
- Model performance varies across clinical seizure phenotypes.
Conclusions:
- CV holds substantial potential to improve epilepsy management and monitoring.
- Key challenges include the lack of standardized validation and patient privacy concerns.
- Further research should focus on model validation across diverse datasets and environments.
