Surrogate Models for Selective Neural Stimulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The advancement of bioelectronic therapies is limited by inadequate activation of targeted nerve fibers and co-activation of non-targeted fibers, due to the complex relationship between applied stimuli and nerve fiber responses, which varies across individuals and species, and is computationally costly to model effectively.
Innovation Solution
A method using surrogate models, including machine learning components, to estimate neuronal responses to electrical stimulation, optimizing electrode geometry and stimulation parameters for selective activation or block of specific nerve fibers, employing techniques like recurrent neural networks and global optimization algorithms to reduce computational costs and improve selectivity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If biophysical models are used to calculate neuronal response to stimulation, then prediction accuracy is improved, but computational cost increases significantly
Solution Approach 1:
The patent creates surrogate models that are simplified copies of complex biophysical models. These surrogate models capture the essential input-output relationships of neuronal response to stimulation parameters but use reduced computational complexity, enabling accurate predictions without the prohibitive computational cost of full biophysical simulations
Solution Approach 2:
The patent replaces traditional biophysical modeling approaches with machine learning-based surrogate models. This substitution transforms the computational mechanism from solving complex differential equations to using trained neural networks, dramatically reducing computational requirements while maintaining prediction accuracy
2Reliability
If model-based optimization is performed to achieve selective nerve fiber activation, then therapeutic benefit is improved, but computational time increases
Solution Approach 1:
The patent performs preliminary training of surrogate models using data from biophysical simulations. Once trained, these surrogate models can rapidly evaluate different stimulation parameter combinations during optimization, eliminating the need to run expensive biophysical simulations for every optimization iteration while maintaining the ability to achieve selective nerve fiber activation
3Adaptability or versatility
If comprehensive biophysical modeling is used to understand individual variations in neural response, then personalization capability is improved, but computational complexity increases
Solution Approach 1:
The patent develops surrogate models that can be trained on patient-specific data from biophysical simulations or experimental measurements. Each surrogate model captures the local characteristics of an individual's neural response to stimulation, enabling personalized prediction and optimization without requiring the full computational complexity of comprehensive biophysical modeling for each patient
Data Source
AI summary
The present disclosure describes novel systems and methods to estimate the response of neurons to electrical stimulation and to determine the optimal electrode geometry and parameters of stimulation to activate or block specific targeted groups of neurons without activation or block of non-targeted groups of neurons.