Abstract
We propose an
implantable autonomous stimulation unit for Functional Electrical Stimulation (
1.
INTRODUCTION
Implanted
Distributed
This paper
focuses on the DSU. We give some design principles as well as experimental
results, obtained with our (hybrid) prototype, to show the performances of this
new stimulation device.
2.
METHODS
2.1.
DSU Design
The architectural
design underlying the SENIS concept leads to embed a complex system within each
distributed
·
An analogue sub-system, in charge of the generation
of the electrical stimulus (Figure 1), which delivers precise calibrated
stimulation pulses to specific multi-polar electrode [4][6].
·
A micro-machine in charge of the execution of the
stimulation sequence. It runs micro-programs using a dedicated reduced
instruction set, and drives the analogue subsystem by calibrating the current
pulse (waveform, amplitude, duration) to be applied to the electrode. Any error
detection puts the stimulation unit in a safe mode regarding the physical
system under control (no more stimulation generated while the DSU has not been
rearmed).
·
A protocol interpreter in charge of all the
communication aspects, allowing to get the following services: download/upload
of programs, start/stop/rearm the stimulation, configure the network parameters
(of a DSU), get the status of a DSU, get an application acknowledgment, notify
an event (error detection for instance), etc.
Figure 1: Stimulation and discharge phases
The medical
context involves very stringent constraints in terms of safety and
performances. These criteria have been considered at every stage of the design,
for all constitutive parts. On the digital part, micro-machine and protocol
interpreter have been thus designed using Petri Nets formalism (PN).
2.2.
DSU modeling and circuit synthesis
Due to the
complexity of the DSU embedded system, we use Petri nets for the design, at a
high level of abstraction, of the digital part. Its formalism (theoretical
basis) and associated tools ease the description and verification (analysis)
phases. The PN based model of the micro-machine is given figure 2. It
corresponds to an “instruction interpreter engine”, composed of 3 parts: (a)
execution of MIT or MT instructions (cf. section 2.3), (b) execution of
Figure 2 : Micro-machine PN based model
An analysis of this formal model allows to extract some proofs [5]. In
order to ensure that the behavior of the programmable electronic device (FPGA)
will be exactly that specified on the model (on a functional point of view), we
must be able to directly program (construct) the
programmable electronic device, from this model. So, we proposed an approach
for the automatic code generation (in VHDL, this langage being dedicated to
FPGA programming), allowing the hardware implementation to be directly
performed from Petri net based models. We also developed the corresponding
software environment, called HILECOP[2].
2.3.
DSU Programming (µ-programs)
The DSU embeds a
micro-machine in charge of
the execution of
a µ-program which represents the stimulation sequence to be performed. The
instruction set is reduced to 3 instructions:
·
MIT instruction; it is a stimulation instruction
used to specify the calibration of the stimulation pulse in terms of selection
of active cathodes and current distribution between them, and amplitude (20 µA
grade shared between active cathodes, 5 mA max.) and duration (1 µs grade[3],
512 µs max.) of the pulse.
·
MT instruction; it is a “neutral” instruction used
to specify the duration (1 µs grade3, 65536 µs max.) of the neutral
phase (I=0) between two consecutive stimulation pulses. During this phase a
safe discharge is done.
·
LOOP instruction; it is dedicated to the repetition
of the sequence of stimulation (i.e. sequence of MIT and/or MT
instructions). The number of loops is
given as a parameter of this instruction (endless loop is possible).
With this very reduced
set of instructions, it is possible to program complex stimulation sequences. A
first prototype has been developped and we have carried out some simulations
and experimental validations.
3. RESULTS
Two kinds of
results are exposed, some in vitro experiments (i.e. the stimulation
device being not “connected” to a nerve, cf. section 3.1) and others performed in
vivo on acute rabbit at the SMI Aalborg, in june 2004 (cf. section 3.2).
3.1
In vitro experiments
The experimental setup is shown on figure 3.
Figure 3: experimental setup
The operator can program the
stimulation sequence and download it into the DSU. Then this micro-program can
be started; the sequence is executed by the micro-machine till it ends or it is
stopped by the operator. An example of simulation sequence is given on figure
4.
Figure 4: example of simple stimulation
The micro-program
corresponding to this stimulator output trace is a sequence of 6 instructions:
-
One MIT instruction; the first pulse is generated on
cathodes 3 and 4 with a current distribution of respectively 2/3 (866µA), 1/3
(433µA). Its duration is 2 ms.
-
One MT instruction; this neutral phase, during which
is done the discharged, has a duration of 20 ms.
-
Three consecutive MIT instructions, that compose the
second pulse. This pulse, also generated on cathodes 3 and 4 with a current
distribution of respectively 1/3 and 2/3, has so 3 phases of 1 ms duration
each, with differents amplitudes (220µA, 433µA, 646µA for cathode 3, for
instance).
-
LOOP twice.
3.2
Experimental results on acute rabbit
The stimulation
device has been used to study the neural fibers time recruitment and the
corresponding muscle force response, of the rabbit’s gastrocnemius muscle, when
varying:
- the pulse train.
Figure 5 shows the force response for a pulse each 35 ms (blue curve) and each
50 ms (red curve).
Figure 5: force responses (left) of the gastrocnemius muscle for different pulse train frequencies (right)
- the pulse amplitude. Figure 6 shows the force response (due to a different recruitment) when varying the pulse amplitude from 0 to 105 µA.
Figure 6: force response (left) of the gastrocnemius muscle for increasing pulse amplitude (left).
4. DISCUSSION AND CONCLUSIONS
We have shortly
presented an hybrid prototype of a new implantable autonomous stimulation unit
for
References
[1] G. E. Loeb, R. A. Peck, W. H. Moore, K. Hood, "BIONtm system for distributed neural prosthetic interfaces", Medical Engineering & Physics, vol. 23, pp 9-18, 2001.
[2] C. Veraart, W. M. Grill, J. T. Mortimer, "Selective Control of Muscle Activation with a Multipolar Nerve Cuff Electrode", IEEE Trans. On Biomedical Engineering, vol. 40, no 7, pp 640-653, July 1993.
[3] J. D. Sweeney, N. R. Crawford, T. A. Brandon, "Neuromuscular stimulation selectivity of multiple-contact nerve cuff electrode arrays", Medical and Biological Engineering and Computing, vol. 33, pp 418-425, 1995.
[4] J. D. Techer, G. Cathebras, S. Bernard, D. Guiraud, Y. Bertrand, “High Voltage Circuit Design for Medical Application”, South European Test Seminar, Austria, february 2003.
[5] R. David, H. Alla, “Discrete, Continuous and Hybrid Petri Nets”, Springer Verlag 2005.
[6] S. Bernard, J. D. Techer, G.
Cathébras, Y. Bertrand and D. Guiraud. “Electrical Performances of a New
Multipolar Micro-Stimulator”. Annual Conference of the International FES
Society,
Acknowledgements
Authors would like to thank Ken Yoshida, Hassan El Makssoud, Christine Azevedo and Michel Bénichou.