Bio-inspired spiking neural network for modeling and optimizing adaptive vertigo therapy

Area of Science:

• Computational Neuroscience
• Bio-inspired AI
• Neurovestibular Modeling

Background:

• Vertigo is a common neurovestibular disorder with limited personalized treatments.
• Current treatment approaches for vertigo often lack precision and adaptability.
• Understanding vestibular system dysfunction is crucial for developing effective therapies.

Purpose of the Study:

• To develop a bio-inspired spiking neural network (SNN) model simulating vestibular dysfunction and recovery.
• To model pathological states like hair cell hypofunction and synaptic disruption.
• To establish a computational foundation for AI-driven, adaptive vestibular therapy.

Main Methods:

• Utilized Leaky Integrate-and-Fire (LIF) neurons with spike-timing-dependent plasticity (STDP).
• Mimicked the vestibular pathway with biologically plausible layers (hair cells, afferents, cerebellar integrators).
• Incorporated a reinforcement-based feedback mechanism for therapy simulation.

Main Results:

• Simulated a 48-62% drop and 38% recovery in cerebellar spike activity during adaptation.
• Demonstrated real-time feasibility with an average simulation runtime of 4 seconds per epoch.
• The model is scalable and suitable for neuromorphic platforms.

Conclusions:

• The SNN model provides a computational framework for in silico testing of vestibular rehabilitation strategies.
• Enables real-time monitoring of vestibular dysfunction and personalized therapy development.
• Establishes a foundation for adaptive, explainable, and hardware-compatible AI-driven vestibular therapy.