Abstract
The derivation by Alan Hodgkin and Andrew Huxley of their famous neuronal conductance model relied on experimental data gathered using the squid giant axon. However, the experimental determination of conductances of neurons is difficult, in particular under the presence of spatial and temporal heterogeneities, and it is also reasonable to expect variations between species or even between different types of neurons of the same species.
We tackle the inverse problem of determining, given voltage data, conductances with non-uniform distribution in the simpler setting of a passive cable equation, both in a single or branched neurons. To do so, we consider the minimal error iteration, a computational technique used to solve inverse problems. We provide several numerical results showing that the method is able to provide reasonable approximations for the conductances, given enough information on the voltages, even for noisy data.
Similar content being viewed by others
References
Avdonin, S., & Bell, J. (2013). Determining a distributed parameter in a neural cable model via a boundary control method. Journal of Mathematical Biology, 67(1), 123–141.
Avdonin, S., & Bell, J. (2015). Determining a distributed conductance parameter for a neuronal cable model defined on a tree graph. Journal of Inverse Problems and Imaging, 9, 645–659.
Bédard, C., Béhuret, S., Deleuze, C., Bal, T., & Destexhe, A. (2012). Oversampling method to extract excitatory and inhibitory conductances from single-trial membrane potential recordings. Journal of Neuroscience Methods, 210(1), 3–14.
Bell, J. (1990). Introduction to theoretical neurobiology. volume 1: Linear cable theory and dendritic structure. volume 2: Nonlinear and stochastic theories (Henry C. Tuckwell). SIAM Review, 32(1), 158–160.
Bell, J., & Craciun, G. (2005). A distributed parameter identification problem in neuronal cable theory models. Mathematical Biosciences, 194(1), 1–19.
Bezanilla, F. (2008). Ion channels: from conductance to structure. Neuron, 60(3), 456–468.
Brown, T.H., Fricke, R.A., & Perkel, D.H. (1981). Passive electrical constants in three classes of hippocampal neurons. Journal of Neurophysiology, 46(4), 812–827.
Casale, A.E., Foust, A.J., Bal, T., & McCormick, D.A. (2015). Cortical interneuron subtypes vary in their axonal action potential properties. Journal of Neuroscience, 35(47), 15555–15567.
Cooley, J., & Dodge Jr, F. (1966). Digital computer solutions for excitation and propagation of the nerve impulse. Biophysical Journal, 6(5), 583–599.
Cox, S.J. (1998). A new method for extracting cable parameters from input impedance data. Mathematical Biosciences, 153(1), 1–12.
Cox, S.J. (2006). An adjoint method for channel localization. Mathematical Medicine and Biology, 23(2), 139–152.
Cox, S.J., & Griffith, B.E. (2001). Recovering quasi-active properties of dendritic neurons from dual potential recordings. Journal of Computational Neuroscience, 11(2), 95–110.
Cox, S.J., & Ji, L. (2000). Identification of the cable parameters in the somatic shunt model. Biological Cybernetics, 83(2), 151–159.
Cox, S.J., & Ji, L. (2001). Discerning ionic currents and their kinetics from input impedance data. Bulletin of Mathematical Biology, 63(5), 909–932.
Cox, S.J., & Raol, J.H. (2004). Recovering the passive properties of tapered dendrites from single and dual potential recordings. Mathematical Biosciences, 190(1), 9–37.
D’Aguanno, A., Bardakjian, B.L., & Carlen, P.L. (1986). Passive neuronal membrane parameters: comparison of optimization and peeling methods. IEEE Transactions on Biomedical Engineering, 33(12), 1188–1196.
de Myttenaere, A., Golden, B., Grand, B.L., & Rossi, F. (2016). Mean absolute percentage error for regression models. Neurocomputing, 192, 38 – 48. Advances in artificial neural networks, machine learning and computational intelligence.
Durand, D., Carlen, P., Gurevich, N., Ho, A., & Kunov, H. (1983). Electrotonic parameters of rat dentate granule cells measured using short current pulses and hrp staining. Journal of Neurophysiology, 50(5), 1080–1097.
Engl, H.W., Hanke, M., & Neubauer, A. (1996). Regularization of inverse problems, Vol. 375. Springer Science & Business Media.
Ermentrout, G.B., & Terman, D.H. (2010). Mathematical foundations of neuroscience, Vol. 35. Springer Science & Business Media.
George, S., & Sabari, M. (2017). Convergence rate results for steepest descent type method for nonlinear ill-posed equations. Applied Mathematics and Computation, 294, 169–179.
Grinvald, A., & Hildesheim, R. (2004). Vsdi: a new era in functional imaging of cortical dynamics. Nature Reviews Neuroscience, 5, 874–885.
Hodgkin, A.l., & Huxley, A.F.A. (1952). A quantitative description of membrane current and its application to conduction and excitation in nerve. The Journal of Physiology, 117(4), 500–544.
Holmes, W.R., & Rall, W. (1992). Estimating the electrotonic structure of neurons with compartmental models. Journal of Neurophysiology, 68(4), 1438–1452.
Jack, J., & Redman, S. (1971). An electrical description of the motoneurone, and its application to the analysis of synaptic potentials. The Journal of Physiology, 215(2), 321.
Kaltenbacher, B., Neubauer, A., & Scherzer, O. (2008). Iterative regularization methods for nonlinear ill-posed problems, Vol. 6, Walter de Gruyter, Roslyn.
Kashef, B., & Bellman, R. (1974). Solution of the partial differential equation of the hodgkin-huxley model using differential quadrature. Mathematical Biosciences, 19(1-2), 1–8.
Kawato, M. (1984). Cable properties of a neuron model with non-uniform membrane resistivity. Journal of Theoretical Biology, 111(1), 149–169.
Kirsch, A. (2011). An introduction to the mathematical theory of inverse problems, Vol. 120, Springer Science & Business Media.
Kobayashi, R., Shinomoto, S., & Lansky, P. (2011). Estimation of time-dependent input from neuronal membrane potential. Neural Computation, 23(12), 3070–3093.
Kreyszig, E. (1978). Introductory functional analysis with applications Vol. 1. New York: Wiley.
Mandujano, J.A.V. (2020). Examples to estimate conductances: Codes avaliable at https://github.com/mandujanovalle/inverse-problem-in-the-cable-equation.
Mascagni, M. (1990). The backward euler method for numerical solution of the hodgkin–huxley equations of nerve conduction. SIAM Journal on Numerical Analysis, 27(4), 941–962.
Mascagni, M.V., Sherman, A.S., & et al. (1989). Numerical methods for neuronal modeling. Methods in Neuronal Modeling, 2.
Nayak, A. (2019). A new parameter-free regularization method for inverse problems. arXiv:1903.03130.
Neubauer, A. (2018). A new gradient method for ill-posed problems. Numerical Functional Analysis and Optimization, 39(6), 737–762.
Pierce, A. (1979). Unique identification of eigenvalues and coefficients in a parabolic problem. SIAM Journal on Control and Optimization, 17(4), 494–499.
Rall, W. (1959). Branching dendritic trees and motoneuron membrane resistivity. Experimental neurology, 1(5), 491–527.
Rall, W. (1960). Membrane potential transients and membrane time constant of motoneurons. Experimental neurology, 2(5), 503–532.
Rall, W. (1962). Theory of physiological properties of dendrites. Annals of the New York Academy of Sciences, 96(4), 1071–1092.
Rall, W. (1977). Core conductor theory and cable properties of neurons. Comprehensive Physiology, 1, 39–97.
Rall, W., Burke, R., Holmes, W., Jack, J., Redman, S., & Segev, I. (1992). Matching dendritic neuron models to experimental data. Physiological Reviews, 72(4), S159–S186.
Schierwagen, A. (1990). Identification problems in distributed parameter neuron models. Automatica, 26(4), 739–755.
Schutter, E.D. (2009). Computational modeling methods for neuroscientists. Cambridge: The MIT Press.
Stuart, G., & Spruston, N. (1998). Determinants of voltage attenuation in neocortical pyramidal neuron dendrites. Journal of Neuroscience, 18(10), 3501–3510.
Tadi, M., Klibanov, M.V., & Cai, W. (2002). An inversion method for parabolic equations based on quasireversibility. Computers & Mathematics with Applications, 43(8), 927–941.
Tuckwell, H.C. (1988). Introduction to theoretical neurobiology: volume 2, nonlinear and stochastic theories Vol. 8. Cambridge: Cambridge University Press.
Vich, C., Berg, R.W., Guillamon, A., & Ditlevsen, S. (2017). Estimation of synaptic conductances in presence of nonlinear effects caused by subthreshold ionic currents. Frontiers in Computational Neuroscience, 11, 69.
Vich, C., & Guillamon, A. (2015). Dissecting estimation of conductances in subthreshold regimes. Journal of Computational Neuroscience, 39(3), 271–287.
White, J.A., Manis, P.B., & Young, E.D. (1992). The parameter identification problem for the somatic shunt model. Biological Cybernetics, 66(4), 307–318.
Willms, A.R., Baro, D.J., Harris-Warrick, R. M., & Guckenheimer, J. (1999). An improved parameter estimation method for hodgkin-huxley models. Journal of Computational Neuroscience, 6(2), 145–168.
Yaşar, T.B., Wright, N.C., & Wessel, R. (2016). Inferring presynaptic population spiking from single-trial membrane potential recordings. Journal of Neuroscience Methods, 259, 13–21.
Zerlaut, Y., & Destexhe, A. (2017). Heterogeneous firing responses predict diverse couplings to presynaptic activity in mice layer v pyramidal neurons. PLoS Computational Biology, 13(4), e1005452.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interests
There is no conflict of interest.
Additional information
Action Editor: Alain Destexhe
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The first author would like to thank PCI-CNPq (301330/2020-4) for its financial suport. Also, the second author acknowledges the support of CNPq (grant 307392/2018-0) and FAPERJ (grant E-26/210.162/2019), and the third author acknowledges support from the research agency CNPq (grant 311087/2017-5), and from the AvH Foundation.
Electronic supplementary material
Below is the link to the electronic supplementary material.
Appendix A: Abstract Formulation
Appendix A: Abstract Formulation
In practice, V |Γ is the data and given such information and under the assumption that Eq. (3) holds, the inverse problem under consideration is to recover or approximate the conductances. The lack of stability, characteristic of ill-posed problems can be tamed by regularization methods (Engl et al. 1996; Kaltenbacher et al. 2008; Kirsch 2011), in particular by the minimal error method.
Consider for simplicity T > 0. Let Ω = {(t, x) : 0 ≤ t ≤ T, 0 ≤ x ≤ L}, and
It is well-known that H(F) and R(F) become Hilbert spaces under the inner products
and the associated norms \(\|f\|_{H(F)}=\langle f,f\rangle _{H(F)}^{1/2}\), \(\|f\|_{R(F)}=\langle f,f\rangle _{R(F)}^{1/2}\). Note that the inner product on R(F) depends on Γ, see Eq. (11), as follows:
where α1, α2 are as in Eq. (11).
The set \(D(F)=\left (L^{\infty }({\Omega })\right )^{N_{{\text {ion}}}}\subset H(F)\) is the Banach space of “essentially” bounded functions (see Kreyszig (1978) for precise definitions). Consider the operator \(F:D(F)\subset H(F)\rightarrow R(F)\) defined by F(G) = V |Γ. Our goal is to find an approximation for G using the minimal error iteration defined by Eq. (6).
In the next Theorem we show how to obtain Eq. (9) from Eq. (6).
Theorem 1
Consider the iteration in Eq. (6). Then Eq. (9) holds
Proof
Given Gk, δ ∈ D(F) and \(\boldsymbol {\theta }=(\theta _{1},\dots ,\theta _{N_{{\text {ion}}}})\in H(F)\), the Gâteux derivative of F at Gk, δ in the direction 𝜃 is given by
where Wk solves
and Vk, δ solves Eq. (3) with Gi replaced by \(G_{i}^{k,\delta }\). To obtain Eq. (28) from Eq. (27), it is enough to consider the difference between problem in Eq. (3) with coefficients Gk, δ + λ𝜃 and Gk, δ, divide by λ and take the limit λ → 0.
Let \(V^{k,\delta }|_{\Gamma }=F(\boldsymbol {G}^{k,\delta })\). From the minimal error iteration in Eq. (6), we gather that
By definition of adjoint operator,
from Eq. (27).
Although Eq. (29) yields an interesting relation, it carries an impeding dependence on 𝜃 through Wk. It is possible to avoid that by performing some “trick” manipulations.
Multiplying the first equation from (10) by − Wk, and integrating in the intervals [0, T] and [0, L] we gather that
Integrating by parts twice the first term from Eq. (30) with respect to the space variable, and using the boundary conditions for Wk we have
where we denote \({U^{k}_{x}}(t,x)W^{k}(t,x)|_{0}^{L}=U^{k}(t,L)W^{k}(t,L)\) − Uk(t, 0)Wk(t, 0). Similarly, integrating by parts the second term from Eq. (30) with respect to time and using the initial condition of Wk and the final condition of Uk, we gather that
Substituting Eqs. (31) and (32) in Eq. (30), it follows that
Substituting the first equation from Eq. (28) in the previous equation, we obtain
From the boundary conditions from Eq. (10), the following expression holds:
From the previous equation and the definition of the inner product in Eq. (26), we have
From Eqs. (29) and (33) we have
Since 𝜃 ∈ H(F) is arbitrary and \(L^{\infty }({\Omega })\) is dense in L2(Ω), we gather that the following iteration holds:
for all i ∈Ion.
□
Rights and permissions
About this article
Cite this article
Mandujano Valle, J.A., Madureira, A.L. & Leitão, A. A computational approach for the inverse problem of neuronal conductances determination. J Comput Neurosci 48, 281–297 (2020). https://doi.org/10.1007/s10827-020-00752-7
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10827-020-00752-7