This study aims to validate a machine learning model that stratifies the risk of stroke in patients who present to the emergency department with dizziness or vertigo.

This is a cross-sectional, hospital-based study, with no randomization procedure. Over a 21-month period, approximately 600 subjects will be enrolled. The study will assess the risk of stroke in each patient using a machine-learning model. To detect ischemic or hemorrhagic stroke, each patient will undergo a non-contrast brain magnetic resonance imaging study. The predictive performance of the machine-learning model will be evaluated in terms of accuracy, precision, recall, F1 score, and area under the receiver operating characteristics curve.

Inclusion Criteria:

• Presenting to the emergency department with dizziness or vertigo
• Aged ≥ 20
• Dizziness or vertigo lasting for at least 1 hour and < 7 days
• Persistent dizziness or vertigo at the time of the examination at the emergency department

Exclusion Criteria:

• Having 5 or more prior episodes similar in quality, intensity, and duration to the current symptoms, with at least one episode more than a year prior and one within the past year
• Having frank stroke-like symptoms or signs, such as hemiparesis, facial asymmetry, dysarthria, dysphasia, ataxia, gaze palsy
• Consciousness disturbance, Glasgow coma scale < 13
• Severe dementia
• Recent stroke within 3 months
• Dizziness or vertigo thought to be the result of head injury, hypotension, medication/drug intoxication, hypoglycemia, electrolyte imbalance, anemia, hepatic encephalopathy, hydrocephalus, etc
• Concomitant head or neck trauma
• Severe degenerative cervical spondylosis
• Having contraindications to magnetic resonance imaging examination

Subjects who visit the emergency department.