Stroke damage detection using classification trees on electrical bioimpedance cerebral spectroscopy measurements.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Medical Imaging

Background:

• Stroke is a leading cause of death globally, necessitating improved diagnostic methods.
• Current stroke imaging technologies have limitations, highlighting the need for portable, non-invasive, and cost-effective alternatives.
• Electrical bioimpedance (EBI) measurements from the head may offer valuable clinical insights into cerebral tissue changes post-stroke.

Purpose of the Study:

• To investigate the potential of Electrical Bioimpedance Spectroscopy (EBIS) for differentiating healthy and stroke-damaged cerebral tissue.
• To extract and classify features from EBI measurements indicative of stroke-related tissue alterations.

Main Methods:

• Recorded 720 EBIS measurements from two head regions across nine subjects, including three with unilateral hemorrhagic stroke.
• Extracted features based on structural and frequency-dependent properties of cerebral tissue.
• Utilized a classification tree model with Leave-One-Out Cross-Validation (LOO-CV) for performance assessment.

Main Results:

• Achieved full classification of damaged and undamaged cerebral tissue in three hierarchical steps.
• Demonstrated the utility of classification features derived from Cole parameters, spectral information, and EBIS measurement geometry.
• LOO-CV confirmed the classification tree's performance.

Conclusions:

• EBI measurements contain valuable information for assessing brain tissue health after stroke.
• Classification features derived from EBIS show potential for differentiating healthy from stroke-damaged brain tissue.
• Further validation with larger patient cohorts is recommended to confirm these findings.