silif-dbs-neuromorphic-controller

name	silif-dbs-neuromorphic-controller
description	Neuromorphic Silicon Neuron Controller for Adaptive Deep Brain Stimulation in Parkinson's Disease - CMOS-implemented SiLIF-DBS controller achieving 5.85%/uW beta suppression efficiency with 75% power reduction vs open-loop
tags	["neuromorphic","deep-brain-stimulation","parkinsons","CMOS","adaptive-dbs","LIF","biomarker"]
activation_words	["SiLIF-DBS","neuromorphic DBS","adaptive deep brain stimulation","Parkinson's","beta suppression","CMOS neuron"]
arxiv_id	2607.05453

Neuromorphic Silicon Neuron Controller for Adaptive Deep Brain Stimulation

Paper Info

• Title: Neuromorphic Silicon Neuron Controller for Adaptive Deep Brain Stimulation in Parkinson's Disease
• arXiv: 2607.05453
• Authors: Md Abu Bakr Siddique, Jakub Orłowski, Yan Zhang, Hongyu An
• Date: 2026-07-08
• Categories: cs.AR, cs.NE
• DOI: 10.1145/3822454.3822465

Core Contribution

First circuit-level realization of a neuromorphic adaptive deep brain stimulation (aDBS) controller using CMOS technology. The SiLIF-DBS (Silicon Leaky Integrate-and-Fire DBS) controller achieves:

• 75% power reduction vs open-loop stimulation
• 5.85%/μW suppression efficiency for pathological beta activity
• Closed-loop operation driven by STN-LFP biomarkers

System Architecture

SiLIF-DBS Controller

• Implemented in CMOS (metal-oxide-semiconductor) technology
• Uses Leaky Integrate-and-Fire (LIF) neuron model in silicon
• Processes beta-band subthalamic nucleus local field potentials (STN-LFPs)
• Control biomarker: average rectified value (Beta ARV)

Closed-Loop Validation Framework

• Embedded within Parkinsonian cortico-basal ganglia computational model
• Driven by beta-band STN-LFPs as physiological input
• Beta ARV extracted as real-time control signal
• Stimulation adjusted based on biomarker threshold crossings

Key Results

Metric	SiLIF-DBS	Open-Loop DBS
Power consumption	25% of baseline	100% (baseline)
Beta suppression efficiency	5.85%/μW	~1.5%/μW
Pathological beta activity	Strongly suppressed	Moderately suppressed

Methodology

Biomarker Extraction

1. Record STN-LFPs from implanted electrode
2. Bandpass filter for beta band (13-30 Hz)
3. Compute average rectified value (ARV) as amplitude envelope
4. Threshold crossing triggers stimulation adjustment

Neuromorphic Implementation

• LIF neuron dynamics in analog CMOS circuits
• Subthreshold operation for ultra-low power
• Event-driven stimulation delivery (only when biomarker exceeds threshold)
• Hardware-software co-design for implantable form factor

Clinical Significance

1. Adaptive stimulation — tracks motor symptom fluctuations in real-time
2. Energy efficiency — critical for implantable pulse generators (IPGs) with limited battery
3. Reduced side effects — stimulation only when needed, minimizing tissue damage
4. Scalable — CMOS fabrication enables mass production

Limitations

• Validated only in computational model, not in vivo
• Single biomarker (Beta ARV) may not capture all symptom dimensions
• CMOS implementation details (process node, area) not fully specified
• Long-term stability of silicon neuron characteristics not addressed

Related Skills

• [[neuromorphic-computing]]
• [[deep-brain-stimulation]]
• [[adaptive-control]]
• [[cmos-neural-circuits]]
• [[parkinsons-disease]]

Implementation Notes

The key innovation is moving aDBS from software algorithms to actual silicon circuits. Previous aDBS systems run algorithms on general-purpose processors consuming too much power for chronic implantation. The SiLIF approach uses the physics of silicon neurons (subthreshold MOSFET operation) to perform biomarker detection and stimulation decision-making at the circuit level, achieving orders-of-magnitude power savings.