Shuangming Yang

Digital implementations of thalamocortical neuron models and its application in thalamocortical control using FPGA for Parkinson's disease

Due to the relay ability of sensory information and communication between cortical regions, the thalamocortical (TC) relay neuron plays an essential role in the therapy of Parkinsons disease. This paper first explores a series of efficient methods for the hardware implementation of TC relay neuron models, aiming to reproduce relevant biological behaviors and present appropriate feedback control in neural dynamics in thalamic systems. In addition, a modified two-dimensional TC neuron model is presented for convenient realization to decrease the complexity of the original model and promote the feasibility of the digital design, which shows significance for the large-scale network simulation of TC-based networks and the establishment of digital thalamus. A system-on-a-chip model-based control system is implemented on an FPGA using the modified TC neuron model, which is aimed at the real-time feedback control of tremor dominant Parkinsonian state. In this paper, the hardware syntheses and theoretical researches are given to illustrate the outstanding performance of the presented hardware implementation. The presented platform can be applied in both the brain-machine interface and the robotic control projects, and the proposed modular hardware framework can be extended to the real-time closed-loop treatments of other dyskinesia diseases. We first present efficient digital implementation of thalamocortical neuron model.Modified thalamocortical neuron model is proposed for a convenient implementation.This model can decrease the resource cost while retaining the biological dynamics.A digital thalamocortical model-based controller is designed using the modified model.The control platform can be applied in real-time control of Parkinsons disease.

Efficient implementation of a real-time estimation system for thalamocortical hidden Parkinsonian properties

Real-time estimation of dynamical characteristics of thalamocortical cells, such as dynamics of ion channels and membrane potentials, is useful and essential in the study of the thalamus in Parkinsonian state. However, measuring the dynamical properties of ion channels is extremely challenging experimentally and even impossible in clinical applications. This paper presents and evaluates a real-time estimation system for thalamocortical hidden properties. For the sake of efficiency, we use a field programmable gate array for strictly hardware-based computation and algorithm optimization. In the proposed system, the FPGA-based unscented Kalman filter is implemented into a conductance-based TC neuron model. Since the complexity of TC neuron model restrains its hardware implementation in parallel structure, a cost efficient model is proposed to reduce the resource cost while retaining the relevant ionic dynamics. Experimental results demonstrate the real-time capability to estimate thalamocortical hidden properties with high precision under both normal and Parkinsonian states. While it is applied to estimate the hidden properties of the thalamus and explore the mechanism of the Parkinsonian state, the proposed method can be useful in the dynamic clamp technique of the electrophysiological experiments, the neural control engineering and brain-machine interface studies.

A real-time FPGA implementation of a biologically inspired central pattern generator network

We engineer a basic CPG with conductance-based KomendantovKononenko neuron model.We propose a multiplier-less FPGA implementation method with low hardware cost.The neural dynamics are highlighted in the design in a biorealistic manner.We employ piecewise linearization method to obtain the reduced neuron model. Central pattern generators (CPGs) functioning as biological neuronal circuits are responsible for generating rhythmic patterns to control locomotion. In this paper, a biologically inspired CPG composed of two reciprocally inhibitory neurons was implemented on a reconfigurable FPGA with real-time computational speed and considerably low hardware cost. High-accuracy neural circuit implementation can be computationally expensive, especially for a high-dimensional conductance-based neuron model. Thus, we aimed to present an efficient multiplier-less hardware implementation method for the investigation of real-time hardware CPG (hCPG) networks. In order to simplify the hardware implementation, a modified neuron model without nonlinear parts was given to decrease the complexity of the original model. A simple CPG network involving two chemical coupled neurons was realized which represented the pyloric dilator (PD) and lateral pyloric (LP) neurons in the crustacean pyloric CPG. The implementation results of the hCPG network showed that rhythmic behaviors were successfully reproduced and the resource consumption was dramatically reduced by using our multiplier-less implementation method. The presented FPGA-based implementation of hCPG network with remarkable performance set a prototype for the realization of other large-scale CPG networks and could be applied in bio-inspired robotics and motion rehabilitation for locomotion control.

Digital neuromorphic implementation of the biologically inspired pallidal oscillator

This paper proposes a modified biologically conductance-based pallidal oscillator model, targeting low-cost and multiplierless implementation with relevant and reliable dynamical characteristics on digital neuromorphic platform. High-accuracy neural computation is limited in scale and efficiency by available hardware resources, so there are significant demands for cost-efficient hardware circuits in the large-scale simulations of neuromorphic field. Thus, the feasibility of a digital implementation with lower hardware overhead cost is investigated in this paper. Implementation results on a field-programmable gate array device demonstrate that the presented model can reduce the hardware resource cost significantly compared to the conventional look-up-table-based design. The proposed methology is an essential step towards the real-time implementation of large-scale spiking neural network, and is meaningful for the investigation on the neurodegenerative diseases and its model-based closed-loop control. It can also be applied in the real-time control of the bio-inspired neurorobotics.

Real-time implementation of the coupled neural mass and its application

The neural mass model is a self-oscillation network composed of two neural populations. In this study, we use the field-programmable gate array (FPGA) device to implement the neural mass model and the hardware implementation results are exactly the same as the MATLAB simulation results. The study reveals that dynamical characteristics of the neural population implemented on FPGA can meet the real-time computational requirements. Besides, we propose a control method of the robotic arm based on the oscillation dynamics of the network. For the implementation results of FPGA is real-time, it can be used to realize the robotic control. A closed-loop control system is realized by inputting the error signals of robotic arm into the neural network model and obtaining the feedback signal to arm joint for error elimination. The results show that the control method based on the neural mass model can quickly and effectively eliminate the angle errors.

Cost-efficient FPGA implementation of a biologically plausible dopamine neural network and its application

Abstract Dopamine neurons play an essential role in terms of cognitive coordination and executive functions, which has been investigated in the therapy of multiple psychiatric and neurodegenerative disorders, such as schizophrenia and Parkinsons disease (PD). This paper first explores a series of efficient methods for the hardware implementation of dopamine neuron model aiming to reproduce relevant biological behaviours. In addition, a modified dopamine neuron model based on piecewise linearisation is presented for efficient realisation to reduce the hardware overhead of the original dopamine model and improve the feasibility of the digital design, which is significant for the large-scale network emulation of dopamine system. The accuracy of hardware implementation is validated in terms of dynamical behaviours and bifurcation analyses, and the simulation results including ion channel properties and compensation effect of N-methyl-D-aspartate (NMDA) and γ-Aminobutyric acid (GABA) activation, coincide with the biological dopamine neuron model with a high accuracy. Hardware synthesis and physical implementation on Field Programmable Gate Array (FPGA) illustrate that the proposed model has reliable performance and lower hardware costs compared to original model. These investigations are conducive to construct large FPGA-based network to explore the neurophysiological mechanisms of dopamine system.

Real-Time Neuromorphic System for Large-Scale Conductance-Based Spiking Neural Networks

The investigation of the human intelligence, cognitive systems and functional complexity of human brain is significantly facilitated by high-performance computational platforms. In this paper, we present a real-time digital neuromorphic system for the simulation of large-scale conductance-based spiking neural networks (LaCSNN), which has the advantages of both high biological realism and large network scale. Using this system, a detailed large-scale cortico-basal ganglia-thalamocortical loop is simulated using a scalable 3-D network-on-chip (NoC) topology with six Altera Stratix III field-programmable gate arrays simulate 1 million neurons. Novel router architecture is presented to deal with the communication of multiple data flows in the multinuclei neural network, which has not been solved in previous NoC studies. At the single neuron level, cost-efficient conductance-based neuron models are proposed, resulting in the average utilization of 95% less memory resources and 100% less DSP resources for multiplier-less realization, which is the foundation of the large-scale realization. An analysis of the modified models is conducted, including investigation of bifurcation behaviors and ionic dynamics, demonstrating the required range of dynamics with a more reduced resource cost. The proposed LaCSNN system is shown to outperform the alternative state-of-the-art approaches previously used to implement the large-scale spiking neural network, and enables a broad range of potential applications due to its real-time computational power.

Real-Time Prediction of the Unobserved States in Dopamine Neurons on a Reconfigurable FPGA Platform

Real-time prediction of dynamical characteristics of Dopamine (DA) neurons, including properties in ion channels and membrane potentials, is meaningful and critical for the investigation of the dynamical mechanisms of DA cells and the related psychiatric disorders. However, obtaining the unobserved states of DA neurons is significantly challenging. In this paper, we present a real-time prediction system for DA unobserved states on a reconfigurable field-programmable gate array (FPGA). In the presented system, the unscented Kalman filter (UKF) is implemented into a DA neuron model for dynamics prediction. We present a modular structure to implement the prediction algorithm and a digital topology to compute the roots of matrices in the UKF implementation. Implementation results show that the proposed system provides the real-time computational ability to predict the DA unobserved states with high precision. Although the presented system is aimed at the state prediction of DA cells, it can also be applied into the dynamic-clamping technique in the electrophysiological experiments, the brain-machine interfaces and the neural control engineering works.

FPGA-based hardware simulation of nonlinear autoregressive Volterra model to reconstruct the single neuron spike pattern

This study explores the implementation of the nonlinear autoregressive Volterra (NARV) model using a field programmable gate arrays (FPGAs)-based hardware simulation platform and accomplishes the identification process of the Hodgkin–Huxley (HH) model. First, a physiological detailed single-compartment HH model is applied to generate experiment data sets and the electrical behavior of neurons are described by the membrane potential. Then, based on the injected input current and the output membrane potential, a second-order NARV model is constructed and implemented on FPGA-based simulation platforms. The NARV modeling method is data-driven, requiring no accurate physiological information and the FPGA-based hardware simulation can provide a real time and high-performance platform to deal with the drawbacks of software simulation. Therefore, the proposed method in this paper is capable of handling the nonlinearities and uncertainties in nonlinear neural systems and may help promote the development of clinical treatment devices.

A multi-FPGA embedded system for the emulation of modular small-world network with real time dynamics

The small-world network plays a vital role in brain function investigation and complex network study. Multi-FPGA design for the spiking neural network remains some challenges in hardware simulation and application. In this paper, a novel multi-FPGA design is proposed to implement a modular small-world network which can ensure a high computational speed and a high calculation accuracy. Time-division multiplexing technology is used in the proposed design to cut down the resource cost of the hardware platform. Clock synchronization is ensured to guarantee the biological dynamics of the network. The proposed implementation can be applied in the investigation biologically-inspired intelligence and brain-like intelligence. Besides, the design scheme can be also extended to other kinds of complex networks and be used to study network dynamics of neural networks.