AI and Eye Tracking Separate Vertigo from Stroke

Differentiating between a harmless inner ear condition and a life-threatening brainstem stroke in dizzy patients is a critical medical challenge. This article explores how audiologists and vestibular specialists combine infrared video-oculography (VOG) and artificial intelligence (AI) to make rapid, accurate diagnoses. By precisely tracking involuntary eye movements known as nystagmus and analyzing the data through machine learning algorithms, clinicians can reliably distinguish benign paroxysmal positional vertigo (BPPV) from central nervous system strokes, ultimately saving lives and optimizing emergency care.

The Diagnostic Challenge: BPPV vs. Brainstem Stroke

When a patient presents with acute dizziness or vertigo, the underlying cause generally falls into one of two categories: peripheral (inner ear) or central (brain).

The most common peripheral cause is BPPV, which occurs when tiny calcium carbonate crystals become dislodged in the inner ear canals. While disorienting, BPPV is harmless and easily treated with physical maneuvers. Conversely, a brainstem or cerebellar stroke is a central cause that can mimic the exact symptoms of BPPV. Because traditional brain imaging like CT scans often miss early-stage posterior circulation strokes, specialists rely on examining the eyes—the direct window to the vestibular system—to make the distinction.

Infrared Video-Oculography (VOG) in Vestibular Testing

To evaluate a dizzy patient, specialists look for nystagmus, which is an involuntary, rhythmic jerking of the eyes. However, patients with inner ear disorders can often suppress these eye movements simply by focusing on a stationary object in a lit room.

Infrared video-oculography solves this problem. The patient wears light-tight goggles equipped with high-speed infrared cameras. In total darkness, the patient cannot fixate on anything, allowing the true nature of their nystagmus to emerge. As the clinician performs positional tests or the HINTS exam (Head Impulse, Nystagmus, Test of Skew), the VOG system tracks the pupils and records the exact velocity, direction, and torsion of the eye movements.

• BPPV Patterns: VOG reveals nystagmus that is brief, fatigues with repeated testing, and follows a specific vertical-torsional trajectory aligned with the affected inner ear canal.
• Stroke Patterns: VOG detects "red flag" central signs, such as pure vertical nystagmus (straight up or down), nystagmus that changes direction when the patient looks different ways, or eye movements that do not fatigue over time.

How Artificial Intelligence Enhances Diagnosis

While VOG provides rich data, interpreting subtle eye movements requires specialized training that isn't always available in urgent care settings. Furthermore, the human eye cannot detect micro-movements occurring in milliseconds. This is where artificial intelligence transforms the diagnostic process.

AI algorithms, trained on thousands of hours of VOG recordings from confirmed BPPV and stroke patients, analyze the eye-tracking data in real time.

1. Micro-Movement Detection

Deep learning models measure the slow-phase velocity of nystagmus and detect micro-saccades (tiny, rapid eye adjustments) during head movement tests. The AI can instantly calculate the exact gain of the vestibulo-ocular reflex, a metric that is typically normal in stroke patients but abnormal in certain peripheral conditions.

2. 3D Torsional Tracking

Tracking the rolling motion (torsion) of the eye is notoriously difficult. Computer vision algorithms map the unique patterns of the patient's iris, allowing the software to accurately measure 3D eye rotation. This confirms whether the eye movement matches the exact anatomical plane of an inner ear canal, confirming BPPV.

3. Automated Risk Scoring

By synthesizing multiple data points—such as nystagmus direction, velocity, latency, and fixation suppression—the AI generates an immediate probability score. If the algorithm detects central patterns, it flags the patient for immediate neurological intervention and an MRI. If it confirms BPPV, the audiologist or specialist can immediately perform the correct canalith repositioning maneuver to cure the patient on the spot.

Clinical Impact

The integration of infrared VOG and AI removes the subjectivity from dizziness evaluations. It empowers audiologists and vestibular specialists to act as a vital triage line, preventing unnecessary neuroimaging for BPPV patients while ensuring that those experiencing a brainstem stroke receive time-sensitive, life-saving care.