Section: New Results

Neuroprostheses

Activating the natural parts through neuroprosthetic devices

Distributed Stimulation Unit (DSU)

Participants : Guillaume Souquet, David Andreu, David Guiraud.

We designed and prototype a 12-pole stimulator according to the same concepts than those of the Stim'3D stimulator generation: a distributed stimulation unit [1] . This 12-pole stimulator, called Stim'nD, embeds the same modules as Stim'3D which are: a micro-machine for stimulation profile execution, a 3-layer procotocol-stack based communication module and a reference models based monitoring module. However, the micro-machine is based on another instruction set which allows to control the discharge phase, and the monitoring module is based on different reference models. These reference models monitor the quantity of charge which has been injected, instead of monitoring the stimulus duration.

We developed a specific software environment called SENISManager that allows to remotely manage and control a network of DSUs (see section 5.1.2 ). This environment has been extended in order to allow for controlling the new Stim'nD stimulator. SENISManager has been registered as a new software application at the french APP agency ( IDDN.FR.001.320011.000.-S.P.2009.000.31500). It is already used by partners of the TIME project, taking advantage of our technology (both Stim'3D and Stim'nD stimulators): UAB Barcelone (SP), AAU-SMI Aalborg (DK), IUPUI Indianapolis (USA), MXM Sophia-Antipolis (FR). This software and the stimulator are used to study spatial fascicular and sub-fascicular selectivity, using an intrafascicular multichannel electrode (Fig. 15 ). Acute in-vivo measurements are performed on rat sciatic nerves [19] .

Figure 15. Rat sciatic nerve with a TIME device transversally implanted (X. Navarro, UAB).

Design and prototyping methodology

Participants : Guillaume Souquet, David Andreu.

Prototyping the digital architecture of distributed stimulation units (DSU) is performed on programmable digital electronic components (FPGA). In order to realize and characterize a dedicated circuit, we designed an ASIC version of the digital architecture of a DSU (its layout is shown in Fig. 16 ). The layout of such a complex multi-clock system has been difficult and presents the drawback of not being evolutive. We thus decide to realize the DSU using recent FPGA technology. This technology is flash-based and ultra-low power; corresponding small size dies are available and should comply with small volume and low power active medical implants. This work has been carried out within the NEUROCOM project.

Figure 16. Layout of the DSU ASIC

Analog part of the micro-stimulator

Participants : Laurent De Knyff, Loïc Bourguine, Olivier Potin, Guy Cathébras, Serge Bernard, Fabien Soulier.

The analog part of the micro-stimulator consists of three main blocks:

• a digital-to-analog converter (dac ) for the conversion of the digital signal of the required stimulation wave-form into the effective electrical current,

• an output stage which amplifies and distributes the current from the DAC to the different poles of the electrode,

• a voltage measurement unit which allows us

  1. to check the right generation of the stimulation current,

  2. to detect potential partial or complete open or short between wires or poles,

  3. to assess the inevitable evolution of the nerve-electrode interface in order to adapt the stimulation current properties.

Since the begining of the DEMAR project, several versions of the analog part of the micro-stimulator have been realized. During this last year, we focussed on the characterization of these micro-stimulators and on the design of a new voltage measurement unit.

Concerning the characterization, we developed for the dac and the output stage, different test boards. These boards allow:

• the measurement of the static parameters: integral and differential non-linearity (inl , dnl ), offset and gain of the dac ,

• the dynamic parameters: signal-to-noise ratio (snr ), total harmonic distortion (thd )...

• an evaluation of the quality of the amplification (linearity) and the current distribution to the different electrode poles.

As an illustration, fig. 17 shows the Differential Non-Linearity of one of the dac . Fig. 18 shows a conceptual view of the stability of the virtual electrode defined by the different current ratios between cathodes and anodes of the used multipolar electrode.

Figure 17. Dnl of a fabricated 8-bit dac .

Figure 18. Stability of the virtual electrode (anode is blue and cathode is red).

Concerning the voltage measurement unit, we designed a new architecture running on various modes according to the objectives of the measurement. Indeed, for simple checking of the real generated current and for estimating the electrode-nerve interface the constraints are completely different in terms of accuracy and power consumption. The final version will be fabricated at the end of November and first validations are planned to 2010. Moreover, the proposed solution allows isolation and conversion to low voltage (3.3 V) of the measured nodes. Because of the nerve-electrode impedance value (1-4 k ) and the expected stimulation current (5 mA maximum) the pole voltages might be higher than 20 volts and the design of an adc (analog-to-digital converter) at this high voltage supply would involve unacceptable silicon area overhead and significant power consumption. Fig. 19 ) gives the overview of the developed system for a sequential measurement on several poles (24 in this example) with only one adc . This work has been carried out within the NEUROCOM project.

Figure 19. Overview of the voltage measurement unit.

External wireless FES system dedicated to clinical rehabilitation

Participants : Mickael Toussaint, David Andreu, Philippe Fraisse.

Even if we focus on implanted FES system, since it is the most restrictive domain, we also work on surface FES architecture and stimulators; external FES system benefits from our concepts and advancements in implantable neuroprostheses. Regarding surface FES architecture, we particularly developed wireless surface stimulation and/or bio-feedback units with our industrial partner Vivaltis. Using this technology, we aim at proposing an adequate solution for Drop Foot Stimulation (DFS) to face remote controllabilty, mobility and comfort issues (see section 6.2.1 ).

The first prototype of wireless surface stimulator that has been manufactured by Vivaltis is shown Fig. 20 . It can provide 2 channels for stimulation or for bio-feedback.

Figure 20. The wireless surface stimulator

Recording from the natural parts through neuroprosthetic devices

ENG electrode design

Participants : Olivier Rossel, Guy Cathébras, Fabien Soulier, Serge Bernard, Christine Azevedo Coste.

The motivation of this work is sensory information extraction from the peripheral nerves in chronic experiment. The propagation of afferent action potentials (ap ) along the axons can be recorded via the electrical activity of the nerve (electroneurogramm, eng ). Unfortunately, this signal appears to be of very low level and even often below the micro-volt.

Moreover, bioelectrical activity makes the in-vivo environment very noisy, the worst noise being the signal generated by muscle activity (electromyogram, or emg) . In the particular case of peripheral nerve sensing, the emg can exceed the eng by order of magnitude of three at least. This parasitic signal will inevitably saturate the high gain amplifier needed to raise the eng to a sufficient level for acquisition. Analog pre-processing must therefore be carried out in order to reject emg -type noise.

Figure 21. Overview of the nerve signal acquisition system.

Based on a previous prototype [10] , we carried out a system-oriented development of an implantable system for ENG recording (fig. 21 ). In order to facilitate the specification of this system, a single axon model have been used to simulate extracellular potential thanks to the Neuron software (fig. 22 ). Then, frequency analysis of simulated signal shows how to use spatial filtering techniques to extract the useful signal before amplification (fig.23 ). Based on simulation results, we have proposed solutions for signal preprocessing and electrode design [34] .

Figure 22. Extracellular potential along a single axon.

Figure 23. ENG spectrum and spatial filter response.

These simulations have shown that the optimal distance between the poles for this type of electrode is about 375 m . This inter-pole distance is much less than the classical distance between rings in multipolar cuff electrodes. This proposed design have been compared to state-of-art electrodes. The first results show better performance in terms of selectivity [31] , [33] .

Implant dependability

Participants : Fanny Le Floch, Guy Cathébras, Serge Bernard, Fabien Soulier.

The fes implanted system may be hazardous for patient and the reliability and dependability of the system must be maximal. Unfortunately, the associated systems are more and more complex and the fact that their development needs very cross-disciplinary experts is not favorable to safety. Moreover, the direct adaptation of the existing dependability techniques from domains such as space or automotive is not suitable. Therefore, we have developed a strategy for risk management at system level for fes medical implant. The idea is to give a uniform framework where all possible hazards are highlighted and associated consequences are minimized.

Figure 24. Level shifter error detection module.

In particular, we focused on one of the most critical part of the fes system: analog micro-circuit which generates the electrical signal to the electrode. As this micro-circuit is the closest to the human tissue, any failure might involve very critical consequences for the patient. For instance, the level shifter has been pointed out as a high sensitive part of the output circuit and has proven sensitivity to voltage variations [32] . We have proposed a detection system as a simple down shifter (fig. 24 ) able to send a warning to the logic control for the implant to switch in a fail-soft mode.