Stroke analysis and recognition in functional near-infrared spectroscopy signals using machine learning methods.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Medical Imaging

Background:

• Stroke is a leading cause of disability and mortality globally.
• Rapid diagnosis is crucial for effective stroke treatment and improved patient outcomes.
• Current diagnostic methods may have limitations in speed and accessibility.

Purpose of the Study:

• To develop and validate an automatic classification framework for stroke detection using functional near-infrared spectroscopy (fNIRS).
• To differentiate between hemorrhagic stroke, ischemic stroke, and normal subjects based on cerebral oxygenation patterns.
• To assess the potential of fNIRS in shortening the onset-to-diagnosis time for stroke.

Main Methods:

• Utilized functional near-infrared spectroscopy (fNIRS) to measure cerebral oxygen saturation and hemoglobin concentrations in the frontal lobes.
• Applied wavelet time-frequency analysis to extract features from fNIRS signals.
• Trained machine learning models using extracted features for classification of stroke types and healthy controls.

Main Results:

• The developed framework achieved an accuracy greater than 85% in classifying stroke types and normal subjects.
• Data augmentation further improved model accuracy to over 90%.
• Identified significant differences in cerebral oxygenation signals among patient groups and healthy individuals.

Conclusions:

• The proposed fNIRS-based framework demonstrates high accuracy in distinguishing between hemorrhagic stroke, ischemic stroke, and normal subjects.
• This technology shows significant potential for improving the speed and accuracy of stroke diagnosis.
• fNIRS offers a noninvasive, real-time monitoring tool that could revolutionize acute stroke management.