differentiable-biophysical-simulation-neurostimulation

name	differentiable-biophysical-simulation-neurostimulation
description	Differentiable biophysical simulation framework for inferring Hodgkin-Huxley parameters from extracellular MEA data. Enables rapid biophysical inference and precise neurostimulation prediction without invasive intracellular recordings.
tags	["neuroscience","biophysical-modeling","hodgkin-huxley","neurostimulation","differentiable-simulation","MEA"]
source	arXiv:2607.04063v1
date	2026-07-05T00:00:00.000Z

Learning Biophysical Models of Large-Scale Multineuronal Data to Enable Precise Neurostimulation

Paper Information

• Title: Learning Biophysical Models of Large-Scale Multineuronal Data to Enable Precise Neurostimulation
• Authors: Amrith Lotlikar, Ian Christopher Tanoh, Praful Vasireddy, Andrew Lanpouthakoun, Ramandeep Vilkhu
• arXiv: 2607.04063v1
• Date: 2026-07-05
• Categories: q-bio.NC

Core Problem

Multi-compartment Hodgkin-Huxley (HH) models provide principled neural dynamics prediction but require invasive intracellular recordings for parameter fitting. This limits scalability to large neural populations and prevents capturing cell-specific properties in circuits.

Key Innovation

Framework to infer HH biophysical parameters from extracellular MEA (Multi-Electrode Array) measurements using:

1. Differentiable biophysical simulation
2. Simulation-based inference
3. Designed features of extracellular signals

Methodology

Differentiable Biophysical Simulation

Extracellular MEA Data
        ↓
Feature Extraction (waveform shape, spike timing, etc.)
        ↓
Differentiable HH Model Simulation
        ↓
Gradient-Based Parameter Inference
        ↓
Predicted Biophysical Parameters

Key Components

1. Feature Engineering: Extract informative features from extracellular recordings:

   • Extracellular waveform shapes
   • Spike timing patterns
   • Population activity features

2. Differentiable Simulation: Make HH model differentiable to enable gradient-based optimization:

   • Backpropagation through biophysical equations
   • Efficient parameter updates
   • Scalable to large neuron populations

3. Simulation-Based Inference: Use simulated data to train inference models:

   • Generate training data from known parameters
   • Learn mapping: features → biophysical parameters
   • Generalize to real experimental data

Applications

Predicting Neurostimulation Responses

Central translational neuroengineering goal: predict neural spiking responses to electrical stimulation

Use cases:

• Optimize stimulation parameters for therapeutic effect
• Minimize side effects by predicting off-target activation
• Personalize deep brain stimulation (DBS) protocols
• Design closed-loop stimulation systems

Large-Scale Circuit Modeling

• Fit HH parameters for hundreds of neurons simultaneously
• Capture cell-type specific properties
• Build biologically realistic circuit models
• Enable in-silico testing of interventions

Technical Advantages

Traditional Approach	This Framework
Intracellular recordings (invasive)	Extracellular MEA (scalable)
Single-cell fitting	Population-scale inference
Manual parameter tuning	Automated gradient-based optimization
Limited to simple models	Full multi-compartment HH models

Implementation Notes

• Simulation framework: Differentiable HH model
• Inference method: Simulation-based inference with gradient optimization
• Data source: High-density MEA recordings
• Output: Biophysical parameters (conductances, time constants, morphology)

Validation

• Predict spiking responses to electrical stimulation
• Match experimental data from MEA recordings
• Generalize across neurons and conditions

Related Work

• Hodgkin-Huxley models (classic biophysics)
• Neural mass models (population-level)
• Differentiable programming in neuroscience
• Simulation-based inference (SBI)
• Brain stimulation optimization

Activation Triggers

• Hodgkin-Huxley, HH model, biophysical modeling
• neurostimulation, DBS, brain stimulation
• differentiable simulation, simulation-based inference
• MEA, multi-electrode array, extracellular recordings
• parameter inference, neural parameter estimation
• computational neuroscience, biophysical parameters