Patient-specific visual neglect severity estimation for stroke patients with neglect using EEG

Area of Science:

• Neuroscience
• Medical Technology
• Signal Processing

Background:

• Spatial neglect (SN) is a common neurological deficit in stroke patients, often resulting from unilateral brain injuries.
• Current SN assessment tools like the Behavioral Inattention Test and Catherine Bergego Scale have limitations in detailing the full extent and severity of neglect across a patient's field of view (FOV).
• Existing EEG-based systems like AREEN show promise but require validation for patient-specific SN assessment to ensure generalizability.

Purpose of the Study:

• To develop and validate a novel EEG-based system for detailed spatial neglect severity assessment across a patient's FOV.
• To introduce an EEG-based combined spatio-temporal network (ESTNet) for analyzing both time and frequency domain EEG data relevant to SN.
• To implement a Bayesian fusion-based FOV correction system to enhance the accuracy of SN assessment by addressing noisy data.

Main Methods:

• Development of ESTNet, an EEG-based network processing time and frequency domain data to identify SN-related frequency bands.
• Integration of a Bayesian fusion system with AREEN's response time data for FOV correction and improved accuracy.
• Utilizing a proprietary dataset for extensive testing and validation of the proposed ESTNet and FOV correction system.

Main Results:

• ESTNet demonstrated superior performance compared to benchmark methods, achieving 79.62% accuracy, 76.71% sensitivity, and 86.36% specificity.
• Saliency maps were generated to enhance model interpretability and explore clinical correlations.
• The combined ESTNet and Bayesian fusion approach showed significant potential for generalized neglect assessment.

Conclusions:

• The novel ESTNet, coupled with Bayesian fusion-based FOV correction, presents a powerful tool for precise and generalized spatial neglect assessment.
• This approach overcomes limitations of existing methods by providing detailed FOV-specific SN severity information.
• The findings support the clinical utility of this EEG-based system for stroke patient rehabilitation and management.