Portable RF brain scanner enables stroke type detection using deep learning.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neurology

Background:

• Stroke is a major global health concern requiring prompt diagnosis.
• Accurate differentiation between stroke subtypes is critical for effective treatment.

Purpose of the Study:

• To develop and validate a deep-learning model for non-invasive stroke subtype differentiation.
• To assess the model's performance using a novel radio-frequency brain scanner.

Main Methods:

• Utilized a 16-antenna radio-frequency array non-invasive brain scanner.
• Employed masked autoencoder-based self-supervised learning and contrastive strategies.
• Trained and tested a deep-learning model on clinical data.

Main Results:

• Achieved 92% sensitivity and 85% specificity for hemorrhagic vs. non-hemorrhagic stroke detection.
• Attained 95% sensitivity and 80% specificity for ischemic vs. non-ischemic stroke detection.
• Observed patterns related to relative dielectric permittivity across different conditions.

Conclusions:

• The deep-learning model shows significant potential for reliable stroke subtype differentiation.
• RF-based measurement-domain analysis offers insights into dielectric differences for stroke diagnosis.
• This technology could improve stroke management through accurate and rapid diagnosis.