Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Development of machine learning models of peripheral nerve stimulation integrated into an online 3D platform for neuroprosthetic applications
Computational modeling of the PNS is a complex tool whose domain spans several physical scales. The long time and cost of their execution can be immensely reduced by the use of AI estimators, which will enable the development of an online 3D platform for real-time model manipulation and computation.
The efficacy of electrical stimulation of the peripheral nervous system in its several applications, such as in motor and somatosensory neuroprostheses, is strongly dependent on the choice of electrode and policy of stimulation. In-silico hybrid computational models have been developed to answer the need for optimization of the electrode's design, surgical placement and stimulation pattern in a flexible manner, reducing the need for animal and human testing. The increasing complexity of such models, driven by the ambition to closely represent the highly convoluted real anatomy and physiology of the peripheral nervous system, presents a critical challenge in terms of computation time, representation, and interpretation of the results.
Our research group has long-standing experience with in-silico models of the peripheral nervous system, which include solid modeling of nerves and their components, Finite Element Method for electric field simulation, and a neurophysiological module to estimate signal propagation along nerve fibers. We wish to share our knowledge with the academic and medical community by developing a tool which will allow for flexible manipulation of models and their parameters (such as choice of electrodes and stimulation patterns) with immediate estimation of the results.
_Electro-Neural model workflow:_ Available datasets for nerve geometry and structure are read into COMSOL. In this simulation environment, a three-dimensional FEM model with an electrical description of the nerve is created, with given parameters such as the injected current and flux at the active sites and impedance and conductivity throughout the nerve fascicle. Output of this simulation consists of the changes in electric field in the nervous tissue obtained from the intraneural stimulation. To complete the hybrid model, the result is processed by NEURON, another simulation environment. Neural fibers distribution inside the nerve fascicle is estimated and signal propagation induced by the evoked electrical field changes to provide a functional measure of the simulated neural interface.
The efficacy of electrical stimulation of the peripheral nervous system in its several applications, such as in motor and somatosensory neuroprostheses, is strongly dependent on the choice of electrode and policy of stimulation. In-silico hybrid computational models have been developed to answer the need for optimization of the electrode's design, surgical placement and stimulation pattern in a flexible manner, reducing the need for animal and human testing. The increasing complexity of such models, driven by the ambition to closely represent the highly convoluted real anatomy and physiology of the peripheral nervous system, presents a critical challenge in terms of computation time, representation, and interpretation of the results.
Our research group has long-standing experience with in-silico models of the peripheral nervous system, which include solid modeling of nerves and their components, Finite Element Method for electric field simulation, and a neurophysiological module to estimate signal propagation along nerve fibers. We wish to share our knowledge with the academic and medical community by developing a tool which will allow for flexible manipulation of models and their parameters (such as choice of electrodes and stimulation patterns) with immediate estimation of the results.
_Electro-Neural model workflow:_ Available datasets for nerve geometry and structure are read into COMSOL. In this simulation environment, a three-dimensional FEM model with an electrical description of the nerve is created, with given parameters such as the injected current and flux at the active sites and impedance and conductivity throughout the nerve fascicle. Output of this simulation consists of the changes in electric field in the nervous tissue obtained from the intraneural stimulation. To complete the hybrid model, the result is processed by NEURON, another simulation environment. Neural fibers distribution inside the nerve fascicle is estimated and signal propagation induced by the evoked electrical field changes to provide a functional measure of the simulated neural interface.
Machine learning estimators can dramatically reduce the complexity of these models, which require long and costly computations, from days to fractions of seconds. We consider an essential objective, within this frame, the development of an interactive web-based application on which our models and results are represented and manipulable in the 3D space, as a tool for researchers to estimate outcomes of electrical nerve stimulation and optimize electrode and stimulation parameters for the desired therapeutic effect. Extensive effort is being spent to deploy it in high quality standards. The complexity of the problem, that spans a very large physical scale, poses strong challenges that will require the development of original solutions for which individual initiative will be rewarded. Deep understanding of the biophysical context and of its representation in the model is expected to arise during an initial formative and then exploratory phase, that will guide an adaptation of the existing framework to the identified needs and requirements, to ultimately build and publish a complete web platform, leveraging 3D visualization to provide a clear and informative representation of the model, allow the choice of stimulation parameters, and aid the interpretation of the simulation results.
**Recommendable skills:** MATLAB; experience with 3D rendering (e.g. WebGL); SolidWorks; COMSOL.
**Time effort:** Master's thesis or Semester project.
Machine learning estimators can dramatically reduce the complexity of these models, which require long and costly computations, from days to fractions of seconds. We consider an essential objective, within this frame, the development of an interactive web-based application on which our models and results are represented and manipulable in the 3D space, as a tool for researchers to estimate outcomes of electrical nerve stimulation and optimize electrode and stimulation parameters for the desired therapeutic effect. Extensive effort is being spent to deploy it in high quality standards. The complexity of the problem, that spans a very large physical scale, poses strong challenges that will require the development of original solutions for which individual initiative will be rewarded. Deep understanding of the biophysical context and of its representation in the model is expected to arise during an initial formative and then exploratory phase, that will guide an adaptation of the existing framework to the identified needs and requirements, to ultimately build and publish a complete web platform, leveraging 3D visualization to provide a clear and informative representation of the model, allow the choice of stimulation parameters, and aid the interpretation of the simulation results.
**Recommendable skills:** MATLAB; experience with 3D rendering (e.g. WebGL); SolidWorks; COMSOL.
**Time effort:** Master's thesis or Semester project.
Dr. Stanisa Raspopovic, Assistant Professor, Head of Neuroengineering Lab, ETH Zürich, Switzerland; stanisa.raspopovic@hest.ethz.ch
Federico Ciotti, Doctoral student; Neuroengineering Lab, ETH Zürich, Switzerland; federico.ciotti@hest.ethz.ch
Dr. Stanisa Raspopovic, Assistant Professor, Head of Neuroengineering Lab, ETH Zürich, Switzerland; stanisa.raspopovic@hest.ethz.ch
Federico Ciotti, Doctoral student; Neuroengineering Lab, ETH Zürich, Switzerland; federico.ciotti@hest.ethz.ch