SnNb₂O₆-Based Capacitive Memristive Synapses with Intrinsic LIF Dynamics for Neuromorphic Epilepsy Detection

Area of Science:

• Materials Science
• Neuroscience
• Electrical Engineering

Background:

• Neurological disorders require early diagnostic tools.
• Synaptic memristors offer potential for bio-inspired computing.
• Leaky-integrate-and-fire (LIF) dynamics are crucial for neuronal function emulation.

Purpose of the Study:

• To develop a SnNb2O6-based capacitive memristor emulating a biosynaptic neuron.
• To utilize the device for epileptic seizure monitoring.
• To integrate the memristor's LIF-type excitatory postsynaptic current (EPSC) with electroencephalogram (EEG) analysis.

Main Methods:

• Fabrication of a SnNb2O6-based capacitive memristor.
• Characterization of device stability and synaptic behavior under various stimulation protocols.
• Mechanistic analysis of capacitive memristor behavior.
• Implementation of LIF-type EPSC as convolution kernels in a 1D-CNN for EEG signal classification.

Main Results:

• The memristor demonstrated stable coupled bipolar switching and capacitive response over 500 cycles.
• A highly reproducible LIF-type three-stage EPSC was triggered by single-pulse stimulus.
• Device response transitioned between transient and persistent states by tuning readout-delay.
• A 95.8% classification accuracy for three epilepsy-related states was achieved on the Bonn dataset.

Conclusions:

• The capacitive memristor successfully emulates biosynaptic neuron dynamics.
• The device shows significant potential for on-chip epileptic seizure monitoring.
• Integration with 1D-CNN enables accurate epilepsy state classification from EEG signals.