Abstract:Modeling biological dynamical systems is challenging due to the interdependence of different system components, some of which are not fully understood. To fill existing gaps in our ability to mechanistically model physiological systems, we propose to combine neural networks with physics-based models. Specifically, we demonstrate how we can approximate missing ordinary differential equations (ODEs) coupled with known ODEs using Bayesian filtering techniques to train the model parameters and simultaneously estimate dynamic state variables. As a study case we leverage a well-understood model for blood circulation in the human retina and replace one of its core ODEs with a neural network approximation, representing the case where we have incomplete knowledge of the physiological state dynamics. Results demonstrate that state dynamics corresponding to the missing ODEs can be approximated well using a neural network trained using a recursive Bayesian filtering approach in a fashion coupled with the known state dynamic differential equations. This demonstrates that dynamics and impact of missing state variables can be captured through joint state estimation and model parameter estimation within a recursive Bayesian state estimation (RBSE) framework. Results also indicate that this RBSE approach to training the NN parameters yields better outcomes (measurement/state estimation accuracy) than training the neural network with backpropagation through time in the same setting.
Abstract:A deep neural network (DNN) that can reliably model muscle responses from corresponding brain stimulation has the potential to increase knowledge of coordinated motor control for numerous basic science and applied use cases. Such cases include the understanding of abnormal movement patterns due to neurological injury from stroke, and stimulation based interventions for neurological recovery such as paired associative stimulation. In this work, potential DNN models are explored and the one with the minimum squared errors is recommended for the optimal performance of the M2M-Net, a network that maps transcranial magnetic stimulation of the motor cortex to corresponding muscle responses, using: a finite element simulation, an empirical neural response profile, a convolutional autoencoder, a separate deep network mapper, and recordings of multi-muscle activation. We discuss the rationale behind the different modeling approaches and architectures, and contrast their results. Additionally, to obtain a comparative insight of the trade-off between complexity and performance analysis, we explore different techniques, including the extension of two classical information criteria for M2M-Net. Finally, we find that the model analogous to mapping the motor cortex stimulation to a combination of direct and synergistic connection to the muscles performs the best, when the neural response profile is used at the input.