EquiliMind: An EEG-IMU Multimodal Fusion System for Early Warning of Meniere's Disease Episodes

Abstract

Meniere's disease is a chronic inner-ear disorder associated with recurrent vertigo and balance impairment; sudden vertigo can result in falls and threaten patients' daily safety. To address the limited capability of conventional symptom assessment for continuous risk monitoring, this paper presents EquiliMind, a multimodal early-warning system. The system integrates electroencephalography (EEG) and an inertial measurement unit (IMU) to synchronously obtain neural activity and body-posture information through a wearable EEG device and motion sensors, and employs EEGNet for balance and motion-state recognition. The experimental protocol included four states: normal standing, single-leg standing, straight walking, and protected simulated falling. Each state was recorded 10 to 15 times for 5 s per trial, yielding 40 to 60 valid task trials and 200 to 300 s of state recordings. EEG data were processed by band-pass filtering, artifact removal, and average referencing, while IMU data were used to extract resultant acceleration, angular-velocity magnitude, and posture-angle change features. Experimental results show that the developed model can distinguish different balance and motion states and detect simulated fall events within 500ms. Channel analysis indicates relatively high relevance scores for T7 and T8, suggesting that temporal channels play an important role in vertigo and balance-risk recognition. The system establishes a complete technical route from multimodal state perception to real-time risk warning, providing a new method for wearable safety monitoring of patients with Meniere's disease.