Prediction of myelinated nerve fiber stimulation thresholds: limitations of linear models

IEEE Trans Biomed Eng. 2004 Feb;51(2):229-36. doi: 10.1109/TBME.2003.820382.


Computer models of neurons are used to simulate neural behavior, and are important tools for designing neural prostheses. Computation time remains an issue when simulating large numbers of neurons or applying models to real time applications. Warman et al. developed a method to predict excitation thresholds for axons using linear models and a predetermined critical voltage. We calculated threshold prediction error as a function of the location of an extracellular electrode using two different axon models to examine further threshold prediction using linear models. Threshold prediction error was low (<3% error) under the conditions examined by Warman et al., but under more general conditions, threshold prediction error was as high as 23.6%. Linear models were limited as effective tools for single fiber threshold prediction because accuracy was dependent on the nonlinear and linear models used, and any parameter that affected the extracellular potential distribution. Threshold prediction could be improved by appropriately choosing the membrane conductance of the linear model, but determination of an optimal conductance was computationally expensive. Finally, although single fiber threshold prediction error was partially masked when considering the input-output (I/O) properties of populations of axons, relatively large errors still occurred in population I/O curves generated with linear models.

Publication types

  • Comparative Study
  • Evaluation Study
  • Research Support, U.S. Gov't, P.H.S.
  • Validation Study

MeSH terms

  • Axons / physiology*
  • Cell Membrane / physiology*
  • Computer Simulation
  • Differential Threshold / physiology
  • Electric Stimulation*
  • Electromagnetic Fields
  • Linear Models*
  • Membrane Potentials / physiology
  • Models, Neurological*
  • Nerve Fibers, Myelinated / physiology*
  • Nonlinear Dynamics*
  • Recruitment, Neurophysiological / physiology*
  • Reproducibility of Results
  • Sensitivity and Specificity