Rapid estimation of cortical neuron activation thresholds by transcranial magnetic stimulation using convolutional neural networks

Abstract

BackgroundTranscranial magnetic stimulation (TMS) can modulate neural activity by evoking action potentials in subpopulations of cortical neurons. The TMS-induced electric field (E-field) can be simulated in subject-specific head models derived from MR images, but the spatial distribution of the E-field alone does not predict the physiological response. Coupling E-field models to populations of biophysically realistic neuron models yields insights into the activation mechanisms of TMS, but the significant computational cost associated with these models limits their use and eventual translation to clinically relevant applications. ObjectiveThe objective was to develop computationally efficient estimators of the activation thresholds of multi-compartmental cortical neuron models in response to TMS-induced E-field distributions. MethodsMulti-scale models combining anatomically accurate finite element method (FEM) simulations of the TMS E-field with layer-specific representations of cortical neurons were used to generate a large dataset of activation thresholds. 3D convolutional neural networks (CNNs) were trained on these data to predict the activation threshold of specific model neurons given the local E-field distribution. Using training and test data from different head models, the CNN estimator was compared to an approach using the uniform E-field approximation to estimate thresholds in the non-uniform TMS-induced E-field. ResultsThe 3D CNNs were more accurate than the uniform E-field approach, with mean absolute percent error (MAPE) on the test dataset below 2.5% compared to 5.9 - 9.8% with the uniform E-field approach. Further, there was a strong correlation between the CNN predicted and actual thresholds for all cell types (R2 > 0.96) compared to the uniform E-field approach (R2 = 0.62 - 0.91). The CNNs estimate thresholds with a 2 - 4 orders of magnitude reduction in the computational cost of the multi-compartmental neuron models. Conclusion3D CNNs can estimate rapidly and accurately the TMS activation thresholds of biophysically realistic neuron models using sparse samples of the local E-field, enabling simulating responses of large neuron populations or parameter space exploration on a personal computer.