Event Abstract

On-line integration of multiple neural network and musculoskeletal models

  • 1 Okinawa Institute of Science and Technology, Neural Computation Unit, Japan
  • 2 Honda Research Institute, Japan
  • 3 University of Heidelberg, IWR, Germany
  • 4 The University of Tokyo, Department of Mechano-Informatics, Japan

We are building a large-scale spiking neuron model of the thalamo-cortico-basal ganglia circuit with the aim of understanding its normal functions and the mechanisms underlying Parkinsonian symptoms, such as tremor. For this purpose, we need to combine multiple models incorporating the basal ganglia, the thalamus, the motor cortex, the spinal cord, and the musculoskeletal system. The minimum size models total 120k model neurons, though larger-scale versions need at least 1.8M neurons; this requires a large amount of computational power as well as the software infrastructure to connect them.

Models are often developed using different tools at separate locations. We need a system that supports distributed development and concurrent execution with efficient communication on parallel computers, so we use the MUSIC library, developed by INCF. It interconnects spiking neuron-level simulators on workstations and massive parallel systems, and comes with a C/C++ API and bindings to common simulators.

The Parkinsonian tremor is thought to originate in the basal ganglia through GPe and STN interaction, where inhibitory rebound causes rhythmic bursting. The oscillation is transmitted through thalamus and the cortical pyramidal tract neurons to spinal motor neuron pools that drives an opposing pairs of model muscles with Ia spindle feedback.

Connecting separate models has several benefits:

I) Each model is independent from the others, so we can use different simulators or separate versions with different built-in models or features as needed.

II) Model development is cleanly separated, and the interconnection structure provides a well-defined model API. Project members can work independently. Intermodel connection issues become clear at an early stage, and the integration work can happen in parallel with model development.

The cortical models are not yet tuned to produce realistic output and inputs to cortical layer 2/3 and the striatal direct pathways happen through preset Poisson sources. The motor neuron network lacks spindle II inputs and spinal networks, and there is no feedback to the CNS from the motor level. These are all issues we need to address in order to reproduce Parkinsonian tremor down to the motor level.

We are nevertheless already able to integrate them and let the integration work guide further model development. We show the current model architecture and performance figures for the interconnected system compared to the separate models.

Figure 1

Keywords: Basal Ganglia, Basal Ganglia Diseases, Thalamus, Motor Cortex, Muscle, Skeletal, Motor Neurons, large-scale modeling, spiking neural networks, Parkinson Disease

Conference: Neuroinformatics 2014, Leiden, Netherlands, 25 Aug - 27 Aug, 2014.

Presentation Type: Poster, not to be considered for oral presentation

Topic: Large-scale modeling

Citation: Moren J, Igarashi J, Shouno O, Sreenivasa M, Ayusawa K, Nakamura Y and Doya K (2014). On-line integration of multiple neural network and musculoskeletal models. Front. Neuroinform. Conference Abstract: Neuroinformatics 2014. doi: 10.3389/conf.fninf.2014.18.00015

Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.

The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.

Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.

For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.

Received: 04 Apr 2014; Published Online: 04 Jun 2014.

* Correspondence: Dr. Jan Moren, Okinawa Institute of Science and Technology, Neural Computation Unit, Onna-son, Okinawa, 904-0495, Japan, jan.moren@gmail.com