[Paper Review] Electric Field Calculation and PNS Prediction for Head and Body Gradient Coils

Peter Roemer, Trevor Wade+4 more|arXiv (Cornell University)|Dec 15, 2020

Cardiovascular Function and Risk Factors4 citations

TL;DR

This paper presents a computationally efficient method for calculating induced electric fields and predicting peripheral nerve stimulation (PNS) in head and body gradient coils using simplified vector potential modeling. The approach achieves ~5% prediction error for body gradients and ~20% for symmetric head gradients, with sub-second computation times enabling rapid optimization of PNS-minimizing gradient coil designs.

ABSTRACT

Purpose: To demonstrate and validate E-field calculation and PNS prediction methods that are accurate, computationally efficient and that could be used to inform regulatory standards. Methods: We describe a simplified method for calculating the spatial distribution of induced E-field over the volume of a body model given a gradient coil vector potential field. The method is easily programmed without finite element or finite difference software, allowing for straightforward and computationally-efficient E-field evaluation. Using these E-field calculations and a range of body models, population-weighted PNS thresholds are determined using established methods and compared against published experimental PNS data for two head gradient coils and one body gradient coil. Results: A head-gradient-appropriate chronaxie value of 669us was determined by meta-analysis. Prediction errors between our calculated PNS parameters and the corresponding experimentally measured values were ~5% for the body gradient and ~20% for the symmetric head gradient. Our calculated PNS parameters matched experimental measurements to within experimental uncertainty for 73% of deltaGmin estimates and 80% of SRmin estimates. Computation time is seconds for initial E-field maps and milliseconds for E-field updates for different gradient designs, allowing for highly efficient iterative optimization of gradient designs and enabling new dimensions in PNS-optimal gradient design. Conclusions: We have developed accurate and computationally efficient methods for prospectively determining PNS limits, with specific application to head gradient coils.

Motivation & Objective

• To develop an accurate, computationally efficient method for predicting peripheral nerve stimulation (PNS) in MRI gradient coils.
• To validate E-field calculation and PNS prediction against published experimental data for head and body gradient coils.
• To determine population-weighted PNS thresholds using meta-analysis of existing data.
• To enable prospective PNS limit estimation for regulatory standards and coil design optimization.

Proposed method

• A simplified method is used to calculate the spatial distribution of induced electric fields from a gradient coil's vector potential field without requiring finite element or finite difference software.
• The method relies on analytical approximations of the vector potential to compute E-fields over whole-body models efficiently.
• Chronaxie values are derived via meta-analysis of experimental PNS data to calibrate PNS thresholds for different coil types.
• PNS thresholds are calculated using established biophysical models and compared against experimental deltaGmin and SRmin values.
• The approach allows for rapid E-field map generation in seconds and E-field updates in milliseconds, enabling iterative coil design optimization.
• Multiple body models are used to determine population-weighted PNS thresholds and improve generalizability.

Experimental results

Research questions

• RQ1 Can a simplified, non-FEM-based method accurately predict electric field distributions in head and body gradient coils?
• RQ2 What is the optimal chronaxie value for head gradient coils based on meta-analysis of experimental PNS data?
• RQ3 How closely do the predicted PNS thresholds match experimentally measured values for body and head gradient coils?
• RQ4 Can the method achieve sufficient computational efficiency to support real-time iterative optimization of gradient coil designs?
• RQ5 To what extent do the predicted PNS parameters fall within experimental uncertainty bounds?

Key findings

• A head-gradient-appropriate chronaxie value of 669 μs was determined through meta-analysis of experimental PNS data.
• Prediction error was approximately 5% for the body gradient coil and 20% for the symmetric head gradient coil.
• 73% of predicted deltaGmin values matched experimental measurements within experimental uncertainty.
• 80% of predicted SRmin values matched experimental measurements within experimental uncertainty.
• Initial E-field map computation took seconds, with E-field updates for new designs achievable in milliseconds, enabling efficient iterative optimization.
• The method enables prospective PNS limit estimation suitable for regulatory standards and PNS-optimized gradient coil design.