Hyrène Project - French National Research Agency 2011-2014

HYRENE is a fundamental research project aiming at the development of innovative technologies : hybrid systems connecting artificial and biological neural networks. 

One goal of this project is to couple a whole organ (mouse spinal cord) with a hardware networks in order to restore the organ functional activity after a lesion. Further perspectives are the development of smart “neuroelectronic” interfaces for functional rehabilitation. Population aging all around the world raises a societal issue due to the associated increase in neurodegenerative diseases. 
One therapeutic approach to treat resulting functional deficiencies is to propose neural prosthesis based on neuro-electronic implants. In recent years, technological advances in the field of micro- and nano-electronics has led to the development of new instrumentation tools for the exploration of the central nervous system, making use of dedicated interfaces between microelectronics and live neural networks. This field of research has strongly developed since 2000, especially with the emergence of brain-machine interfaces. These interfaces, which are now tested in humans, process brain signals recorded with microelectrode arrays to turn them into command signals for the control of external devices (robotic arms, computers…). 
However, to date, such interfaces remain mainly monodirectional, with no information delivered back to the network. The current challenge is to achieve bidirectional neuro-electronic interfaces, establishing a true dynamic communication between live neural networks and electronic systems. Especially, electronic systems connected to neural networks with existing technologies do not include embedded intelligence. 
The technical approach defined for HYRENE is to couple live large-scale neural networks and artificial neural networks embedded in analog and mixed integrated electronics and endowed with adaptive capabilities (synaptic plasticity). This hybrid coupling will use dedicated microelectrode arrays to record and electrically stimulate live neural networks, with a specific emphasis on stimulation localization. The system including the artificial and living neural networks will form a closed loop with a regulated feedback. The artificial neural networks will implement conductance-based neuron and synapse models, controlled by plasticity rules like STDP (spike-timing dependent plasticity). 
Dedicated integrated electronics will be designed to implement the communication channels between the living and artificial networks: signal conditioning for the biological signals (from living to artificial) and adapted coding of the artificial neurons events (from artificial to living). In this project, integration between physics (electronic engineering, Microsystems) and biology (integrative neuroscience) is mandatory. 

All partners rely on their large experience in multi-disciplinary collaborative projects at national and international levels. This project is expected to generate scientific advances:

• in the field of information science: by the design of embedded self-organized artificial neural networks, able to communicate in real time with entire biological networks;
• in the field of life science : a tool to develop and test efficient strategies for spinal cord rehabilitation.

Partners: ESIEE, Université Victor Segalen Bordeaux 2

Brainbow Project - EU FET Young explorers 2012-2015

Enhancing recovery of cognitive and motor functions after localized brain injuries which disrupt connections between brain and body is widely recognized as a priority in healthcare. Nowadays, neurological diseases implying severe motor impairment are among the most common causes of adult-onset disability. Millions of people worldwide are affected by paralysis, and this number is likely to increase in coming years, because of the rapidly ageing population. Current assistive technology is still limited since only a minority of survivors with hemiparesis is able to achieve independence in simple activities of daily living. The frequent lack of complete recovery makes a desirable goal the development of novel neurobiological or neurotechnological strategies for brain repair.
Over the last decade Brain-Machine Interfaces (BMIs) and generally neuro-prostheses (Nicolelis, 2003; Hochberg et al., 2006; Nicolelis & Lebedev, 2009; Hochberg et al., 2012) have been object of extensive research and may represent a valid treatment for such disabilities. The development of these devices has and will hopefully have a profound social impact on the quality of life. Nevertheless, modern neural interfaces are mainly devoted to restore lost motor functions, because of injuries at the level of the spinal cord (Collinger et al., 2012; van den Brand et al., 2012), or recover sensorial capabilities, e.g. through artificial retinal or cochlear implants (Chader et al., 2009). However, the majority of motor disabilities are caused by brain diseases, such as stroke and traumatic brain injury - TBI - (33%) and not by spinal cord injury (23%).
Only very recently scientific interest has been devoted to in vivo cognitive neural prostheses. The first ever hippocampal prosthesis improving memory function in behaving rats has been presented in recent papers (Berger et al., 2011; Berger et al., 2012). Lately the same group tested a similar device in primate prefrontal cortex aimed at restoring impaired cognitive functions (Hampson et al., 2012; Opris et al., 2012).
The realization of such prostheses implies that we know how to interact with neuronal cell assemblies, taking into account the intrinsic spontaneous activation of neuronal networks and understanding how to drive them into a desired state in order to produce a specific behaviour. The long-term goal of replacing damaged brain areas with artificial devices requires the development of neural network models to be fed with the recorded electrophysiological patterns to yield the correct brain stimulation aimed at recovering the desired functions. All these issues are extremely difficult to investigate in vivo, due to the inherent complexity and low controllability of the system. On the other hand, we believe that important insights (e.g. structure-dynamics relationship, neural coding) might be gained by using in vitro systems of increasing architectural complexity, which can be easily and wholly accessed, monitored, manipulated, and thus modelled.
This topic is extremely up-to-date and represents one of the most important challenges over the next years in terms of clinical impacts and translational medicine. This is demonstrated, not only by the literature over the past years, but also by new US funding programmes in this specific direction (see e.g. DARPA website: http://www.darpa.mil/default.aspx, programmes REPAIR and REMIND). In particular, the group of T. Berger (University of California at Irvine, CA, USA) published several papers in 2013 regarding the development of hippocampal prosthesis (both on in vitro and in vivo experimental models) for memory enhancement (Deadwyler et al., 2013; Hampson et al., 2013; Hsiao et al., 2013). On December 2013, another very interesting paper came out from the group led by R. Nudo, very active in clinical studies related to stroke and TBI (Guggenmos et al., 2013). In this paper the very first example of a unidirectional ‘neural bridge’ aimed at promoting functional connection between two motor areas (i.e. the premotor cortex and the sensory cortex) in a rat model of TBI was demonstrated. Our project BRAIN BOW is exactly along the same line, but with the goal to make even a step forward with respect to these studies: design a chip for network replacement able to operate in a closed-loop fashion. The preliminary results demonstrating that a biological network and an artificial one are able to communicate and influence, in a bi-directional way, their intrinsic dynamics, constitute one of the most promising results of the second year and represent the basis for the activities of the third and final year.
Read more...

Partners: Istituto Italiano di Tecnologia (Italy), University of Genova (Italy), Tel Aviv University (Israel)

MHANN Project - French National Research Agency P2N 2011-2015

In 2008, researchers at Hewlett-Packard have unveiled a new electronic component, called the memristor. Theorized by Leon Chua in 1971, this electronic device is a non-volatile, non-linear resistor. By applying a voltage, it is possible to vary continuously the resistance of the device, and the device "memorizes" that resistance after the voltage is no longer applied. As such, these memristors offer a wide variety of applications as binary (OFF/ON) or multilevel/analog memories or switches in reconfigurable memories. In addition, memristors intrinsically behave as artificial synapses. Most of the memristor devices proposed up to this day are based on defect-mediated physical effects, for example the electromigration of oxygen ions for the Hewlett-Packard components. To use such devices, the reliability issues due to high operating temperatures, difficulty of a precise control of the switching behavior and potential device deterioration will have to be solved in order to achieve the large endurance and retention times required for operational components.
In 2009, the UMPhi-CNRS and Thales project partners have patented a new component: the “ferroelectric memristors”. This memristor belongs to another class of memristors, called “Electronic effect memories”. The resistance changes are due to purely electronic effects and therefore preserve the materials structure. They are based on a physical concept radically different from the existing solutions: ferroelectricity in tunnel junctions. The resistive switching is based on the intrinsic switching of ferroelectric domains and therefore possesses a fundamental merit over defect-mediated mechanisms to achieve the reliable performance necessary for commercial production.
On the other hand, companies like Intel have insisted that the most important high-performance applications are not scientific computing but the following three categories: Recognition, Mining and Synthesis applications (RMS), the first two categories largely relying on classification, clustering, approximation and optimization algorithms, for which competitive algorithms based on neural networks exist. Due to stringent power consumption constraints, the clock frequency of processors no longer increases or barely, so that a hardware neural network would retain all its performance and power advantage (about two orders of magnitude) compared to the software version run on a processor.
So we can note that there is a convergence of technology, architecture and application trends: hardware artificial neural networks are well suited to tackle an important class of applications while coping with upcoming technology hassles.
Therefore, memristors constitute an ideal and very timely alternative implementation for synapses of hardware artificial neural networks. Memristors would be far denser than current SRAM-based implementations but they would also require far less power since the memristor is a non-volatile memory. Hardware ANNs with an architecture composed of analog circuitry coupled with the aforementioned memristors open the possibility to build high-performance accelerators able to tackle the large computational tasks of RMS applications.
The purpose of this project is to build a medium-scale prototype of such a bio-inspired architecture, by using long-life and nanometric “ferroelectric” memristors. The area, performance and power benefits of this approach will be evaluated to define its interest for embedded systems.
The MHANN project is multi-disciplinary in the sense that it proposes new physical concepts for devices (physics) and aims at integrating them into on-chip bio-inspired architectures (micro-electronics, computer science and architectures).

Partners: UMPhi CNRS-Thales, Thales TRT, INRIA

MIRA Project - French National Research Agency 2015-2019

This project plans to develop a path towards a novel neuromorphic computation paradigm using memristors to process visual information in a highly efficient way and at unprecedented speed. It will make use of a cutting-edge neuromorphic retina sensor that, rather than frames outputs scene-driven, time-encoded visual information at microsecond resolution. The spike output from the camera will feed into an artificial neural network based on memristors and silicon neurons. This approach will provide – for the first time – a way to process visual input at rates up to 100 kHz while at the same time reducing the system’s power consumption. The proposed project is rooted in the emerging field of “neuromorphic engineering”, relying on close interaction of material sciences, neurosciences, mathematics and microelectronics. The MIRA project will contribute new knowledge, techniques and applications to this growing field and the constituent areas.

Partners: UMPhi CNRS-Thales, Institut de la Vision, Chronocam