Related Experiment Video
Updated: Jun 28, 2025
06:37
Electroencephalography Network Indices as Biomarkers of Upper Limb Impairment in Chronic Stroke
Published on: July 14, 2023
870
Differentiating ischemic stroke patients from healthy subjects using a large-scale, retrospective EEG database and
William Peterson1, Nithya Ramakrishnan2, Krag Browder3
1University of Virginia, Charlottesville, VA, United States.
Summary
This study developed a machine learning model using a 3-minute electroencephalogram (EEG) to accurately distinguish ischemic stroke patients from healthy individuals, achieving high precision.
Area of Science:
- Neuroscience
- Medical Technology
- Artificial Intelligence
Background:
- Distinguishing ischemic stroke from healthy states is crucial for timely intervention.
- Electroencephalogram (EEG) offers a non-invasive method for brain activity assessment.
Purpose of the Study:
- To develop a machine learning model for differentiating ischemic stroke patients from healthy subjects.
- To utilize short, resting electroencephalogram (EEG) recordings for stroke detection.
Main Methods:
- A retrospective dataset of 2401 EEG recordings (374 stroke patients, 1385 healthy subjects) was compiled.
- Features from spectral and temporal domains of 3-minute EEG recordings were computed.
- A machine learning model was trained using this dataset, incorporating multiple recordings per subject.
Main Results:
- The model achieved an Area Under the Curve (AUC) of 0.95.
- Sensitivity was 93% and specificity was 86%.
- Including multiple recordings per subject improved sensitivity by 7%.
Conclusions:
- Machine learning applied to EEG data shows significant potential for identifying stroke patients.
- The developed model provides a timely (3-minute recording) and accurate diagnostic tool.
- This approach offers a precise and efficient method for stroke detection.
