Note: Descriptions are shown in the official language in which they were submitted.
20~78~7
FES METHOD IMPROVBMENTS
The present invention relate~ to FES method im~xovement~. I~
one aspect, the pre~ent invention ~elate~ ~o a no~el
Rrinciple ~or r~coxdi~ and using natur~1 sensory nexve
a~tivity from pe~ipheral nerve~ ln a system ~or functiona~
electrical stlmula~ion (FE~) wherein the artifac~ ~ignal
generated by muscle stimulation i~ minimize~. In another
aspect, the invention relates to the utilization of a
specific slip-related ~ignal ~or providing a secure grip o~
a~ object throu~h FES os a p~r~ical~y or com~letely paralyzed
muscle which i8 involved in holding the object.
GE~ERAL BA~KGRo~D
It i6 po6~ible to 6timulate paxalysed muscles electrically
and in this way make paralysed limb~ perform functional
lS moveme~ts. T~is tec~nique, called functional electri~al
stimulation (FES), has been known for de~ade~ and se~eral
research grou~s worldwide are currently in~olved in
developing the techniqus. The research i~ dire~ted more and
more towardfi the development of implantable sy~tem~ tha~ use
alo~ed-loop co~trol, since moat pr~viou~ system~ ~ave
su~f~ed from practical and co~metlcal problems becau6e of
the external equipment a~d ~he dif~iculk control of
stlmulated m~le~. To de~elop ~uch 6ystems, it i~ ne~essary
to develop ~nsorg that are ~uitable ~or long-term
imp~an~tion. Some necessary fe~ture~ of such ~enso~s a~e
biocompatibili~y, reliability, durab~lity and small size.
~ry faw artificial sensor~ have these ~eatures.
For a ha~d-grasp prosthesis or foot-drop pxosthesis, nerve
cuf f elec~rode~ might be implanted on lndi~idual nerves in
30 the hand or foot. ~h~3 aign~ expected to tell when a~
object }~as co~ltact with the ski~ (for example in the fins~ers,
the nerve ~ig~al response when the object slips, and in the
foot a~ impact between the sole a~d the ground surface). The
~i~n~ n thus in a hand-grasp prosthesis be used f or i . e .
2~7~7
updating the stimulation intensity if a ~rasped object ~tarts
to 61ip and also to detect the minimum stimulation intensity
necessary for holding an object, and in the ~oot the ner~e
signal can be u~ed to activate the ankle dorsiflexor muscle
at the appropriate time du~ing ~ step-cycle.
Human ~kin, joints and mu~cle~ a~e equipped wi~h numerous
se~sors t~at e~able us to sense our ~urrounding~ ~nd the
st~te of our body. If this info~m~tio~ can be reliably
recorded, it i9 pos~ible to use these _~O~ Lnn~ ~ to
provide feedback si~nals to an F~S system. The technic~l
problem~ of recording the information have been ~ub~tantlal
because of the small size o~ the nerve fibre~ and the low
amplitude of the ~i~n~ hat i~ to be recorded compared to
the noi~e introduced b~ ~timulation o~ ~u~ ne~r~y.
US Pate~t 4,750,499 by J.A. Ho~fer de~cribed an FES Yystem
for p~tially re~toring th~ motor ~un~ion of ~ per~on h~ving
p~r~lyzed mucles, said method comprising implantin~ a ~orce
sensor ~ompxi~ing a nerve electrode for ~ensing electrical
signals pri~ily f~o~ ~ech~noreceptors ~s~ociated with a
2~ ~eripheral sen~ory nerve that supp].ie6 glabrous ~kin of the
pex~on ha~ing the paralyzed mu6cle~;, 6ensing electrical
sign~ls ~ia ~aid force sen~or, proclucing an alectrical
c~n~ol ~ign~l ~or actl~atin~ a mu~cle stimulator i~ response
to the electrical ~ignal~ sen6ed by ~aid ~erve el~ctrode, ~nd
~5 6tlmulating the p~r~lyzed ~usc~le~ in ~ccor~n~e with s~id
contxol ~i~nal. In a practical implementation of this ~ystem
artifact6 ~uch ~ stimulation arti~acts and mus~}e respon~e~
du~ to the ~timulation will, however, be supe~imposed on the
electro~euro~ram (ENG) ~ignal. If the nerve cuff electro~e
must be located in clo~e pro~imity to ~timulated mu6cle, a~
may often be the c~e, th~ EN~ ~ignal will most l~kely be so
disturbed ~hat the proposed sy~tem can not function. ~ ~
In an ~t~empt to xeduce the noi~e introduced by 6timulation ~ :
of muscles nearby, Knaflitz ~nd ~exletti ~ ) have
developed ~ de~ice which eature~ the following: a) a
;:: . :
:. . . : , :
. . .
20g78~37
~'hybxid" outpu~ stage, op~ical isol~tion of both ~he
~timulation output stage and o~ the inp~t ampllfie~ stage,
monophasic or bipha~ timulation output, artif~ot
~uppres~ion obtai~ed by slew rate limiting in the isolated
~tage and signal bl~nking i~ the ~round re~erred stage, and
~i~gle and double differential de~ection o~ the myoelectrlc
sig~al.~ Howe~er, this hae not been a useful approach as the
~witche~ in mo~ ~a~es created more noi~e than ~ras removed
when applied to the high gain ~10000~) ampl~fiers used in
theix ~tudy.
DESt::RIPTION OF ~IE INVEN~IO~
A main aspect of the pre~ent invention relate~ to a method
for a~ lea~t partially restoring the motor fun~tion of a
partially or completel~ p~ralyz~d mu~cle, said method
comprising implanting a nerve electrode for ~e~ing
electrical signals from ~ ~en~ory ner~e which innervate~ a
part of the body which i6 physiologically rel~ted ~o the
partially or completely paralyzed muscle, stimulating ~he
paralyzed muscle by means of a mus~le stimulator, ~en~in;T by
mean~ of said ~er~e electrode an electrical signal from the
sensory nerve, producing a furthe~ control ~ign~l for
reac~ivati~ the muscle ~tl~ulator depende~t on the
electrical signal sensed after the expiration of a
predetermined period of time after the stimula~ion o~ the
muscle, and rest~mulating the p~ralyzed muscle in response ~o
~aid furth~r control ~ignal.
An importan~ ~eature of ~he present inve~tion is that dur~ng
predetermined period o~ time after the ~timulation o~ the
muscle or muscle~, esRentially no electrical ~ignal i8 sensed
~rom the sensory nerve, i.e. the further control signal for
reactivatin~ the muscle ~timulator i~ dependent on~y on the
electrical signal sensed after the expiration of a pre~
determined period of time ~fter the stim~lation o~ the muscle
or ffl~scles~ The objective of thi~ is to a~oid or minimi~e th~
artifact cont~mination ~y the electromyo~ram (EMG) on the EN~
2~978~7
signal. Examples on how this can be aco~pli~hed are described
in more detail in the followiny.
T~e invention ~an al~o be de~cribed a~ a FES ~ystem for at
least partially restoring the rnotor function of a hum~n
5 having ~t least one partially or ~ompletely paralyze~ muscle,
~aid system compr~sing a stimulator mean~3 ~or stimulating the
p~ral~r2;ed m~scle or mu6cles ~ ~n impla:ntable ner~re electrode
for s~n~ing elec~rical ~ignals from a sen~ory nerve whi~h
innerv~te~ a part o~ the bod~ which i~ phy~iologicRlly
10 ~elated to the partially or c:ompletely paralyzed nuscle or
muscle~, and
control means re~ponsive to ~he electrical s~gnal~ sens~d
~ro~n said se~sory ner~e a~ter the expiration of a pre-
detennined p~riod of time after the ~tir~ulation of 'che mu~cle
or mu~cle~ ~or produci~g a ~urthe~ control signal for re-
activating said stimulator mean~.
In a presently preferred embodiment, the invention relate~ to
a FES ~ystem as described above w~e~ei~ said con~rol m~ans
20 compri~es means for amplifying, band-pasE~-filtering, and
optionally reatifying and bln-integrating the sensed electri-
cal signal and means for producing said further control ~ig-
nal in response to ~aid bin-integrat~d electxical sig~al. As
an example, the bin-inte~ration in ~he FBS sys~em according
to the inve~tion can be performed ~ mean~ of an integrator
having an adjustable integration period, ~aid integrator
being synchronized wi~h ~he s~imulator means.
More than on~ par~ially ox completely paraly~ed muscle can be
stimulalted by the method ox ~ES ~ystem o~ the invention, In a
particular e~bodiment whi~h i6 described in more detail in
Example 1, the recordin~ cu~ was placed on the tibial ner~e
~nd the ~o~r plan~Ar~lexor muscles were s~imul~ed in turn by
the co~p~ter. T~e present ~n~ention thu~ ~lso provides ~
method ~or at lea~t partially re~tori~ the motor ~unction of :
~evexal paxtially or compleSely p~ralyzed muscles, said
method comprising impl~nting a nerve electrode for senslng
~,. . .,: i ,
2~7~7
electrical sig~ om a sensory nerve which innervate~ a
part of the body which i~ ph~siologic~lly related to the
parti~lly or completely par~lyzed mu~les, stimul~ting one of
the paralyzed musc~e~ by means o~ a musc~e ~timulator,
5 ~en~ing by means of s~id nerve electrode an electric~l signal
from the ~n~oxy nerve, producing a further con~rol ~ign~l
~o~ ~eactivating ~nother muscle ~timulator ~ependent on ~e
electrical 6ig~al sensed after ~he expiratlon of a pre-
determi~ed period of ~ime ~ft~r the stimulation of ~hè
muscle, and restimulatin~ another paralyzed mu~cle ~n
response to said further control ~isnal. In an alternative
em~odiment, ~everal partially or completely paralyzed muscles
may be stimulated 6imultaneou31~, the further control ~i~nal
for reactiva~in~ the muscle gtimulator or mus~le stimulators
bein~ depe~dent on the electrical ~ignal sen~ed after the
expiration of a predete~mined period of tim~ after the first
stim~l~tion of the mu6cle6. In further embodiments, the
method may compri~e more than one nerve electrode coupled to
one or ~everal mu6cle ~timulator~, i.e. the method may com-
prise the ~ombination o~ two or moxe ~S system~ o~ theinvention workin~ to~ether i~dependently or integrated e.g.
by mean of a computer in an appropriate way.
A ~uitable nerve electrode for se~ing electrical signal~
from a se~o~y nerve can be e.g. a cuff electrode implanted
25 around a peripheral se~ory nerve. In such a cuff elect~ode,
which may be a split cuff or a 6piral cuff elec:trode, the
~tability of th~ re~orded ~ig~al i.~ s~able enough to be used
i~ a FES sys~em. Also othe~ ~uitable nerve electrodes such as
intrafa~cicular ox in~neural electrode ~Hoffer and ~aug-
land, 1~92) m~ be used in the meth~d or FEs system ~ ~heinvention.
In ~ presently pre~erre~ ~mbodime~t, the method according to :~
t~e in~en~on i~ ~ method wh~rein the electrical si~nal i8
ampli~ied, band-pass-fil~ered, and optionally rectified and
bin-int~grated when producing 6aid further control ~i~nal.
- ~ . . : ., .: .
;.... . .. . .. . ... . . . . .. .. .
2~g7~7
The bin-inte~r~ n can be performed b~ mean~ o~ an inte-
grator having an adjustable integration period, said lnte-
grato~ being syn~hroni~ed with the ~uscle ~timul2tor or
~uscle stimulator~. The sensed electrical signal i~ generally
S inte~rated i~to a single ~alue to produce the ~urther control
slg~al .
Within the ~cope of the prese~t invention are al~o msthods
w~erein the EMG artiact i~ minimized by other ~ethod~ such
as subtra~ting the EM~ nal recorded ~y other, ne~r~y e~ec-
0 trodes ~rom the ENG sign~l prior to the proce~ing b~ themethod of the i~ention. By the use of ~ch fuxther methods
for artifact suppre~Rion, the predetermined time period
during which essenti~lly no ele~t~ ignal i8 6en~ed ~rom
the æen60ry nerve can be mi~i~ized.
~n a presently pre~erred embodime~t, the p~rt of the body
innervated by the eensory ner~ compriæe~ a skin area.
However, it i~ contemplated that the method of the i~ve~tlo~
will alæo be u~e~ul when t~e part of the body innervated by
the ~en~ory ner~e compri~es ne~ve~ f~om proprioceptors æuch
as muæcu~ar spindles, Golgi tendon organ~ or join~ recep~or~.
The m~cle ~timul~tion ~a~ be ~ompli~h~d by any suitable
muscle ~timulator. ~or a re~lew, ~ee e . ~ nes a~d Creery,
1~90. ~lso other stimulator~ with appropriate ~eatures can be
uæed in the method of the invention. .
6enerally, for hu~an application~ it i8 pre~ently pr~ferred
that the ætimulation signals should ~e genera~ed ~ ~
~regue~cy o~ about 5-50 E~z, more preferably about 10-20 Hz.
The ~cle re~timul~tion provided by the further stimulation
~on~ol ~i~nal can be ~axied e.g. by ad~us~ing the amplitude
of the stimulation pul~e~. Al~o~ ~he pulse wid~h o~ the
~timul~ion pulse~ c~n be ~aried. Ini ~he presently pre$erred
embodiment~, the pul~e width of the ~ti~ulatio~ipulses i3 i~
the magnitude of about 100 - 300 microseconds, su~h as ~50 -
~0 micros~conds, e.y. 200 microsecond~. Al~o the fre~ue~cy
,''` ` :` : . ' `: ~ ' ' '`. :
: .,:,., : .. : ' ' .. ` :, ` ~
','`', , ' , ` ` . ' ' ' ' ` ` '
:: , ' : , ` ~: : ' ..
: , ~'' ' ~ ' . " "' ' ' '.:
~-: : ' , , :
':" '' ': `
`,, ` : ` `
,
2~9~7
of the ~ti~ul~tion o~ the mu3cle or muscle~ ma~ be varied~
~owe~ex, i~ a simplified embodiment of the method of the
in~ention, the frequency of the stimulatiôn of the m~scle or
mu~cles is sub~tantially con~tant.
Based upon th~ information in the present ~pecifi~ation a~d
claim~ will be within the skill of th~ pe~son skilled in
~he art to determine a suitable prede~ermined period of time
after the ~timulation of the mu~cle du~ing whi~h e~entiall~
no electrical ~ign~l should be ~ensed from the sensory nerve.
0 Thl~ pe~iod will of course ~ary depending on the diagnosi~ of
the patient, the particular paralyzed m~cle or muscles which
are ~timulated, the distance and orientation ~e~ween the
nerve electrade and the mu6cle stimulator, ~he ~ype o* the
~timulation el~ctrode~, the maximum current neede~ for
stimulation e~c. ~enerally, the period i~ cont~pla~ed to ~e
about 3-10 ms or in certain circumstance~ ~ven 1~8 or more
after the stimulation of the ~uscle. ~s a starting poin~, it
is proposed that the time period i~ set to be about 5-7 ms.
Example 1 describes a cat model ~.o validate reduction of
~0 artifact~ by th~ m~thod of the inv~3ntion, and ~xample ~
de~cribes a ~imple F~S system ~or correction o~ footdrop in a
hemiplegic spasti~ male.
The techni~ue used in the present lnvention i~ ~hown ~hema-
tically in Figur~ lA. A n~rve electrode ~uch ~ a tripolar
cu~f elect~ode, i~ implanted on a ~e~30xy nerve t~at innex-
~ates the part of the body of interest, e.g. skin, joint(s)
ox mu~cle~) or a combination the~eof. ~ dif~ere~tial
amplifier (AMP), with high common-mode reJection and a gain
between o.I million and 1 million, a~plifies the ~ignal
recorded from the ~en~er and the two connected
end-ele~trodea. The amplifier has ~ bandwidth from ~00 Hz to
10 kHz, comparabl~ to the bandwidth of the nerve ~i~nal. Thi~
~ignal i~ then rectified in a ful~-wave act~ve rectifier
(R~iI). When nearby mus~les are ~timulated, this signal
una~oidab}y cont~i~s large peaks o~ stimulation a~ti~acts ~if
~ . , .
^ 8 2097~7
sur~ace ~timulation is ~sed, t~e ~timulation pul~e~ will
t~pically have an amplitude of 100 v, whereas the recorded
ne~v~ ~ignal is ln the order of 1 ~V~.
After ampIif~cation of the signal from the nerve cu~f ele~- :
trode, further proces~ing i5 ne~e~sary to remove noise ~nd
artifact~, produce a ~ignal t~at ~epresen~s the over~ll
acti~ity in the ~er~e, and change that siynal into a cont~ol
~igmal to the ~timulator. An ~xample of a ~ircuit to do thi~
h~ been implemented for a portable ~oot-drop corre~tion
system, ~ased on analog components.
The circuit consi6ts of the followin~ parts as shown in the
block diagxam (Fi~ure ls):
HP-filter
Thi~ high-pass filter is m~inly nece~ary ~ecause of ~ossible
offset voltages ~rom the ampli~ier ~ha~ supplies the cir~uit
with information from the electrode. However, ~in~e ~ere
mi~ht be some ~nintended muscle acti~ity in the recorded
~ignal, the cutof~ ~re~uency has been set high, to
approximately 600 Hz.
~0 Active rectifier
Af~er the hig~-pass ~ilter, the ~ignal is now without DC
off~et and i~ then full-wave rectified, a~ the first stage of
producing an envelope of the nerve ~ al. The recti~ier is
acti~e, i.e. it doe~ not lose inform~tio~ beca~se o~ ~ diode-
~oltage drop for negati~e ~alues, as would a p~iverecti:Eier.
Bin-inte~.r~tor with timing circuit
In ~he pre~ent embodiment the bin-integ~ator is the key part
o~ the arSi~act remo~al sche~e. It ~on~ists of an int~grator
30 that can be re~et, as controlled by an external
IB2376B~ /AS/lg93o6o4 ::
. '-:: . "'~' ' : : :`~, : : ,
::;' , :: ' : ,
:,:: , :
2~7~;~7
synchroni~ation signal and the timer circuit ~eing part o~
the integ~at~r. When a muscle i~ electrically stimulated, the
stimulator gen~s a sy~chroniza~ion signal to the bin-
integrator. Thi~ nal is high for t~pically 100 ~8 duri~y
each stimulus. D~rin~ thi6 time, ~he i~p~ to the integrator
ia di~conne~ed by U4~ (see Figure lC), re~ulting in the
value on the output of the i~tegrator being kept constant.
When the synchronization signal end~, it ~ta~ts a timer (U3A
on the circuit diagram) (Fi~ure lC) tha~ short-circuit~ the
capacitor i~ the integrator (C3). The tim~r i8 ad~u~table up
to 27 ms. During thi~ p~riod, that i~ adjustable in du~ation
b~ P~, the integrator i~ reset, and thus ignores the signal.
The timer is norm~lly adjusted to time ou~ ri~ht a~ter stimu
lation arti~cts that contaminate the si~nal have ~ied out.
Then the integrator ~tart8 to integrate during a perlod which
the ~ignal contain8 only neu~al i~formation wi~hout artifacts
until a new synchronization ~ign~l appears from the ~timula-
tor. ~uring the time the integrator outpu~ i8 kept con~tant,
the switches U4~ ~nd U4D co~ec~ it to the hold-circuit made
up by C4 and U2C, ~hereby ~oring the fin~l value of the
inte~ration. The ~equence o~ the~e values is ~s~ally re~exred
to as the recti~ied, bin-integrate~l ENG, or jus~ E~G, and
i8 a direct mea~ure of the overall activity ~n the nerve.
~ndpa~s filter
25 It is nece~sary to process further the RE~-ENG ~o produce a
cc~ntrol signal for the stimulator. In the present em~odiment,
the fir~t step is band-pa~ filtering to produce a ~ooth
signal that woul~ reflect overall change~ in the nerve
activity rather than ~he absolute ~ctivity. ~his is done by
ca~cadin~ tw~ pa~Bive first order ~ilte~ - a lQwpass section
(R10 ~nd ~7) with ~ ~utof~ f~equency of ~pproxim~tely 20 Hz,
and a highpa6s ~ection (~5 and R~0) ~et at approximately
1 H~. An inverting amplifler with gain = lo then ~ollows
~ecause o~ loss o~ signal amplitude in ~he passi~e ~ilte~s.
~7~1.~1M~1~3~
: ~ . : : . :::: . ~ : : . :
: :. , - : : . . :
lo 2897~7
Acti~e rec~ifier
~sually the astest change~ in nexve ac~i~it~ are increases
(as when the heel ~trikes the floor), but sometimes ~
decrease could gi~e ~he most p~ominent peak in ~he highpass
fil~ered ox differentiated nerve activity. To increa~e
detection reliability, both the positi~e an~ negative peak
are used for detection by rectifying the si~nal firet. ~ain,
the rectifier i~ a~tive in order no~ to lose infoxma~ion
because o~ diode-~oltaye drops.
Comparator
After highpas~ filtering and rectification, heel-s~ik~ o~n
now be detected by simple thresh~ld detectio~.
Timer
In the present example, ~ timer that ~tart~ at heel-~trike
and time~ out a~ter a time that corresponds to the duration
o~ the stance ph~se h~ been emplo~ed. In this way the
inform~tio~ obtained when She heel hits ~he floor has been
used. The timer control~ a tranæistor that rwitches the
st~mul~tor on and off (whi~h iB a commercially available
20 unit). The stimulator is turned on when the timer ru~s out
(estimat~d time of foot lift~ ~nd off when the heel hits the
floor Iwhich then initiate~ a new c~cle). The timer i~ :
adjus~able up to 2.7 ~. This system might ~e modified to use
the inormation obtained when the heel i8 lifted fxo~ the
floor.
T~e analo~ c~rcu~t described abo~e i~ ba~ed on a digital
implement~tion, uising al di~ital 6ignal proces60r tDSP) (TMS
320C35) placed on a plug-in ~rd in a I~M-~ompa~i~le peF~-ona
compute~ The sircuit wa~ thus implemented as a~ ~nalog
30 circui~ as at tha~ time it di~ not seem practi~ally possible ::~
to make a ~m~ll DSP ~ard ~or a porta~le ~nit. It i~
contempl~ted, however, that in the near ~uture it will be
18~37~iEII.001/AS/1493Q~; 04
,: ~ : ::: : ` : ~: , . .
:.: ,. : :
: ` " :
2 ~ 7
pos~ible ~o ad~pt a~ existing portable DSP æ~tem that will
be able to perform t~e same function and other functlon~
well.
The circuit presen~ed ~n Fi~ure lC is implemented using ~tan-
dard analog componentS. However, the size can be dra~t~cally
reduced by using 6urface mou~ted componentS (SMD), as the
present inventor~ have done with the amplifiers th~t have
been ~onstruc~ed. Furthermore, there seem to ~e no m~jo~
problems in implementin~ a digital version on a custom-
desi~ned chip (ASIC) that would include bo~h the ampli~ler~or t~e neural ~i~nal~, the electronics ~escribe~ here for
control of a s~imulator, and th~ s~imulator itself. This ha~
been done by others for more 8~ mple stimulators (~ee e.g.
~gnes and Creery), which has made it po~ible to mount all
the Compo~e~ts inside a biocompati~le case and im~lant thiS
in the body. ~ .
In ~ummary, ~o remove the ar~i~act~ described above, t~e
recti~ied ~ignal i~ thu~ bin-integrated in su~h a w~y th~t : ~:
the noise-free signal in-betwee~ stimuli i8 integrat~d into a
~in~le value by means of a ~witched integrator and
sample-and-hold circuit con~rolled by the ~timulation or
alte~native an~log or digital methods a~ described above.
Co~a~ina~io~ by s~imulation axti~acts wa~ ~educed by
~ampling the ENG only durin~ periods in-bet~een artifacts.
The resul~lng 3ignal, w~ich correspo~d~ to the smoot~ed
en~elope v~ She nerve signal, or in oSher words, the o~erall
a~tivity in the ner~e is then processe~ (~OGIC), after which
it can be used as a feedback ~iqn~l for the ~timul~tor :
~STIM).
Anoth~r method f~r suppre~sion o~ a stimulation artif~ct
could be to blank i~ out by ~i~her di~onneoting or shor~-
cir~uiting the recording electrode during and for ~ome time
after the sti~ulu~ pxi~ciple thi~ would wo~k if ~oi~ele~s
~witches we~e available. Howe~ex, ln e~perime~ts pe~foxmed by
the present inventors with currently a~ailable switches,
1~376~.~SIASI~3~
: . : . : : . :
. , :.,: ,: , . . : : . :, : ~
., . :: , , , , - , : . , : . . , :. ,
. ... .
lZ 2 ~ 5 7
these proved to be una~ceptable becaus~ o~ ad~itional
switching no~e. The metho~ of the invention only ~equires
tha~ the ~irst stage of amplification has a short
recovery-ti~e a~ter eYe~tual SatUration. It ia con~empl~ted
tha~ a recovexy-time in the range o~ 0-~ m~ will ~e
pre~erred, ~ut reco~e~y-times up to about 5-7 ms or even more
will be acc~pt~ble - depending on the ~equency o~ ~u~cle
stimulation.
Due to ~he larger distanceY that can be obtained between
stimu~ation electrode~ and a nerve cuff electrode it can be
expected that in human FES systems, th~ arti~act problem~ ma~
be smaller than thoYe presented i~ the cat model ~see Example
1~. On the other hand, human muscles can ~e larger than ca~
muscle6, and the numb~r of mu~cle~ ~i~ulated in a F~S system
o~ the invention may be much larger than fou~, which may in-
crease the number of artifacts. Ho~ever, as long as there i
at least 10 ~ between any two ~timuli, it should be posstble
to ~ample noise-free ENG. This ca~ be obtained if group~ of
~us~les axe stimulated ~i~ulta~eou~ly rather than ~n a random
or time-distributed manner. In hu~n applications, it i~ also
likely that the stimulation frequency will be lower than the
25 Hz per muscle uYed fo~ the cat expe~iment~, thereby
leaving longer artifact-free periocls between the consee~tive
~timuli.
25 Among several restorative application~ that can be envi~ion- :
ed, ~wo, in par~ r, ~o~ld in prin~iple be implemented
readily~ con~rol of ha~d fun~ion i~ ~uadriple~ic or hemi-
plegic per~on~, and control o~ stance and g~it in paraplegic
or hemiple~i~ pe~on3.
Elec~rical 6timulation of the pe~oneal ner~e, used for cor-
reGtion o a drop-~oo~ ha~ ~ecome an ~ bli~hed therapeutic -~
a~d fu~tion~l method. The ~timulation i6 applied during the
swing ph~se of the ~ffected leg and prevent~ drop-~oot, so
the patient walks faster and more securel~ The use of the
~5 natural tactile inform~tion recorded f~o~ ner~e~ suppl~ing
~ fi~ A~1~3~
. ~ . : ., . ~ ::
.
,........ .. , , I ~ ~ :
. ~ . .. :
, . . . . .
,
. :
' '' : .
13 2~7~7
the foo~ can make it po~sib~e for ~he pa~ient to walk without
wires running from an exte~nal heel cont~ct sen~or to the
stimul~tor. using the distin~t peak th~t appear~ i~ the sur
ENG ~i~nal at heel - contact and the method of the invention,
it h~ ~e~n possible to cont~ol a drop-foot ~timulator dur~g
w~lking (see Ex~mple 2). The di~in~ neural peaks are
contemplated to relate to p~essure changes on the skin and
a~sociated stre~che~ of the ~kin mechanorece~o~ in the ~kin
area ~hat ~ innerv~ted by the ~ural ~erve.
10 The hum~n fingertip has ~ great capabi~ity to detect slips
between an objeat a~d the surface of the skin. ~urface fea-
ture~ protruding a few micrometers from the 3ur~Ace of an
object can be di~criminated whe~ ~troked alon~ the surfa~e of
the skin a~d small sli~s invol~i~g onl~ ~ p~rt of the ~kin :~
surface in contact with ~n object held in a precision grip
can be d~tected ~Johansson and We~tling, 1~87). T~i~
capabili~y ori~inate~ from the large number of low threshold
mechanore~eptors in the ~kin of the fingertips (as many a~
241 units/cm2 (Johansson and Vallbo, 1979)). The infGrmation
~rom these receptors is importa~t for the control of
preci6ion grip; if the ~kln i~ anae~thetized, the ~b~e~t
be~ome~ ina~pable o~ ~dequately adjus~ing the grip ~orce to
the weight ~nd surface structure o~ an objec~ (Joh~n~on ~nd
Westling, 1984).
During p~ecision ~asks, normal subjects u~ually produ~ g~ip
force~ ~re~y greater than the minimum force required to hold
a~ objec~ (Johans~on and Westlin~ 87), which is determ~ned
by both the weight of the object and the fric~ional prop-
er~ies of the sur~ace in cont~ct with the f inger8. If the
~0 grip force is insufficient and thQ object ~tarts to 81ip,
mo~t low-~hre~hold merchanorecep~ive ~ni~ re~pond with ~harp
bursts of ~cti~it~. Slips may happen if the grip force
ch~nge~; a~ ~ oonseqlle~e of ~h~ngin~ joint angles or f~ti~e,
if the weigh~ of ~he object increases (e.g. a coffee c~p ~hat
gets filled) or if the frictional coefficient decreases (e.g.
caused by per6piration). In normal human subjects it has been
18:Z376i31.0011AS/19~3~
: ' : :: : ~ : .: :
: . , ~.::: :.:: ~ :
- , . , :
,. ' '
14 ~978~7
shown that a ahort-latency spinal reflex of cutaneou6 oxigin
is usually elicited by ~uch a ælip, ~o that withln 80 ms of
the ~tart o~ a slip, th~ grip force increases and the objec~
is held set;:urely again (,Johansson and Westl:Lng, 1987). This
rapid corr~ctive re~pon~e iS automatic, ~nd doe~ not i~olve
conscious particip~tion by th~ 8u~j ect~. The re~lex ~ction i~
quite powerful, to the extent tha~ sub; ects often cann~t
voluntarily relea~e their grip slowl~ ~o let an ob~ect fall,
be~ause the reflex tends to interfere ~ohans~o~ a~d
Westling, 1~87).
The cumulatlve activi~y of cutaneous merchanoreceptor~ can bç
reco~ded with a nerve cuff electrode (reviewed by Hoffer,
lggO) and is pre~umed ~o be adeguate ~or feedback control of
a sy~tem for ~un~tional electri~al ~timul~tion (~ES). The
relation ~etween the electroneurogram (ENG) o~tained from a
nerve cuff implanted on a sen~ory ner~e and the ~orce applied
perpendicularly on the sk~n innervated by the ner~ wa~
identified, but contained ~ome inherent ~on-li~earities tha~
made it difficult to use the EN~ for estimation o the
~0 perpendicular skin contact force. The sen~i~ivity of t~e
cutaneou~ mechanoreceptorQ to ~lip~ occur~i~g betwee~ the
skin and an ob~ect ~uggests that the ENG may contain a
different, and possibly ~rery u~eful, kind of infor~nation
other that ju~t information about per~endi~ula~ co~tact
force. The possibility o~ e~tracting 81ip ~elated information
from the ENG recorded with a ~erve cu~f electro~e and its
pos~ible applica~ion in an FES system for hand grasp is a
central feature of the present invention. Wi~h rellable Qllp
in~ormation available, an FES 8y~tem impleme~ting ~
r~ iGial gxipping re~lex~ should enable a par~lysed ha~d
to hold an obje~t with onl~r the neces~ary force and maintain
a good grip ev~n if the muscle~; fatigue, if skin~to-object
~iction ~ ges o~ i~ the weight of ~he obj e~t c:h~n~es . A
di~erent version of a~ H~r~i~icial gripping re~lex" ha3
prevlou~ly been imple~ented ~or the control o~ prosthetic
hand6, where the incipient slip o~ an ob; ec~ bei~ held by ~n
',:, - ~' ~ ` i
:: :
.. . . .
.~. , ,
: .
:. :
, ., : : . : .
~7~
amputee in a pro~thetic hand wa~ measured and u6ed to c~ntrol
the force o~ prehension (Colman and Salisbur~, 1967).
Ou~ e~perimental m~del of a h~nd ~aspang and lifting an
object was the foot of an an~es~hetized cat pressin~ ag~inst
5 an objec~ that would slide down if the force exerted by ~he
~oot was in~u~icien~ to hold the object in place. The fox~e
wa~ p~od~ced ~y stimulation of the plantar~lexor muscle~ vi~
intramuscular electrodes, u~ing a c~mputer controlled FES
s~tem. Slip i~formation extracted fxom ~he tibial ENG wa~
u3ed to compensate for slip occurring in two d~ff~ren~
experimental situations: When mu~cle ac~ivatlon declane~ a~d
the grip force ~ell below ~he minimu~ level re~uired for a
secure grip or when the weight of th~ object increased
suddenly while being held with a constant grip force, ~au8ing
15 the obj ect to move .
The findings made in connection wit~ these experimen~s gave
rise to another important main aspect o~ the prese~t
inventlon. This a~pect of the ~nve~tlon relates to a met~od
~or pro~iding a sec~e ~rip of an objec~ through FES o~ a
partially or completely paralyzed muscle which is involved in
holdin~ the obj ect, compri~ing
implanting a nerve electrode ~or ~e~n~ing electrical signals
~rom a sensory nerve which innerYates a part o~ the bod~
which i~ phy~iologically related to the partially or
completely paraly~d m~Rcle,
detec~ing the s~art of a slip o~ the obje~t from the ENG
si~nal from the nerve electrode,
and immediately upon detection of ~he ~tar~ o~ a slip
producing or modifying a control signal for activating a
muscle ~timulator and ~timulating the paralyæed mu~cle in
respons~ to gai~ control ~ignal.
:
16 ~9~7
~he ~tart o~ the slip ma~r, e . g ., be caused by an ex~xnal
change in the load or wei~ht o~ the object, or ~y an in~ernal
d~sturba~ce such as fa~ , or it may be c~used ~y ~ chans~e
in the frictional coefficien~ of the ~kin, such a~ due to
sweating.
As will ~e expl~ined in greater detail in the following, the
start o~ the slip is pre~erably detected ~utomatically a~ an
event that exceeds a predetermined threshold value, A
6uitable way of doin~ t~ig is whe~e the ENG recorded from the
sensory nerve is processed u~ing analog or digital circuitry,
arld a ~ilte~ed version of the ENG is subtr~cted from the
actu~l electri~al ~ignal sen6ed, thereby removing unrelated
background a~ti~ity ~rom ~he signal, the res~lting ~lcu~ted
signal being ident~fied ~s a signal related to the start of a
slip if it exceeds a predetermined ~alue. The fi~tered
version of the ENG, ~uch as a low pa~s-filtered ~ersio~, is
suitably delayed by a number of samples and then subtracted
from the act~al ~ignal ~en6ed. The actual signal which may or
20 may not represe~t a "start of sl~p-related signal", ~
suitably ~n unfilte~d ~ignal or ~ ~ignal which has been :
subjected to filtering wlth a shorter time constant than the
first-mentioned filtered ver~ion.
Fur~her de~ails concerning thi~ aspect of the invention are
given in below:
Sig~als from cutaneou~ mechanorec~ptor~ were recorded with a
nerv~ cuff electrode implanted on the tibial nerve of cats.
The~e ~lgnals ~an be rel2ted to the perpe~dicular ski~
co~tact ~orce appl~ed on the central footpad. The sig~al~ can
also be recorded during functional electrical stimulation
~ES) and th2ref~re may be u~ed a~ ~eedback ~or an FES s~stem
appropriate for restoring motor ~unctions ln patient6
paraly~ed ~y, for example, a ~pinal cord injury or ~troke.
. .. . . . . , . . .. , .. . ~ , .. .. . . . ~ .. . .. .. . . .. . ~. .. ... . .
17 2~97~7
Example ~ describes how slip-related information wa~ derived
from the cu~aneous electroneurogram (EN¢). This inormation
was used in ~n ~vent-driven contxoller ~or pre~ision grip
that allowed the FES system to compensate ~or unexpected
5 61ips between an object and the skin. In thi~ way an "a~ti~i-
cial gripping reflex~ was implemented that comp~nsated auto-
matlcally for internal change~ tigue) and external pertur-
ba~ion~ lincreased load, changing frlctlonal coefficient).
Thls control acheme proved to be ~o~ust, and is presumed to
be applicable ~or restoration of precision grip in paraly~ed
humanfi using FES.
The info~mation recorded with a nerve cuff elec~rode
implan~ed on a cutaneous nexve can be used reliably for
~eedba~ ~o~t~ol i~ a ~ES sy~tem. This appro~ch could havP
dlrect applications for the restora~ion of preci~ion gr~p ln
spinal cord inju~ed patient~ (e.g. C4-C6 ~uadriple~ic;, where
i~ could ~e u~ed ~o i~plement a~ "artific~al gripping ~eflex'
zimilar to the natural ~ripping re~lex that 1~ prese~t 1
normal h~man~.
The ~ em of the invention reliably detected slips from the
re~orded ENG, in ~ituations where t~he slips occurred either
because of decreasi~g ~rip force or be~au~e in increased load
force. The grip wa~ ~ully regained by i~suing a pul~e doublet
to each of the four muscle~, combined with a 25-30% increase
in ~he duration o~ ~ubsequent pul~es. The reaulting grip
~orce clo~ely re~embled th~t o ~u~ane when ad~usting to a
~udder~ increa~e i~ load force (~oh~nsson and Westling, 198~)
by increasing rapidly to a high value and the ~et~lin~ at a
value ~omewh~t high than b~fore the load ~hange. ~he time
30 ~etween the obje~t fir~t started to move and ~lip detection
was in ~he range o~ 50 to 100 m~. Typically, for ~lip8 cauaed
by ou~ "fatigue" p~r~digm, the object moved less than 3 mm
before it w~ c~ught, and for ~lips caused by a ~udden
increase in load the object moved lesa than 4 mm befoxe it
W~S c~ugh~. The method proved ~o be equally robust in ~11
three c~t~ u~ed in thi~ ~tudy.
. _ . . .. . .
.' : .: -., , ' .'' ': ~ . .'. ., :
'` ' ' ' :'. ' ,' .' ', ' ' ' ' : ,
` 18 2~7~7
Only one pa~a~eter (the threshold v~lue) i~ the slip det~c-
tion algorithm needed initial adjustment at ~ach da~ o~
experiment, although it did ~ot need any ~urther change~ for
~he ~st bf the recording ~e~qion (that usually l~ted 4-6
hour~ he threshold ~a~eter could be increas~d ~y more
t~an 50~ of its optimal value wi~hout ~eriously effecti~g the
detection of 81ips. Once ~et, the thre~hold value ~eeded no
other ~d~ustment durlng the re~t of th~ experimen~ that day.
The increme~t in the ~timulat~on intensi~y after ~lip w~s set
to 25-30~ in the pres~nt experiment~, whic~ proved necessary
and sufficient to ~ecure the ~rip i~ these specific ~itu-
ations. In a clinical FES system, the optimal value of thi~
factor will depend not only on the inst~ntaneou~ recruitment
curre for the stimula~ed muscle~s), but also on the rea~ons
15 for ~lips, e.g~ how the weight of the obje~t ha~ changed, how ~ :
strong a per~urbation was, as well as a number of other
factors, æuch as the frayility of the object. It is antici-
pated that this value ma~ be easil~ ad~ustable by the usex,
to ~it the exact condi~lons of his/her muscles, ~emperament,
and FES system.
In additio~ to e~abli~g an FES sy~tem to compensate for
declining grip for~e or external perturbations, slip informa-
tion may also be used for an initial determination of the
optimal sti~ulation i~tensity, and thus a~oid causing muscle
fatigue ~rom using unnece~sarily strong stimulation. The
sti~ulatio~ ~chem~ shown in Fig. ~, i.e., a ~lowly decreasing
intensity co~bi~ed with slip detec~ion and compensation,
~toma~ic~lly determine~ ~he minim~m ne~ess~xy intensit~ ~o
hold an object and to adjust to Qlowly varying load change6,
~0 or i~ may just be used initially to de~ermine a suitable
in~en~ity that i~ ~hen held constant unl~s and unti~ other
~ occur.
In the intact human, infoxmatio~ from ski~ re~eptors is
es~ential for the regula~ion of force duri~g p~e~isio~ grip.
Visual cues s~ch as ~ize, material a~d surface structure play
: . . . ' ' " ' ' ' . . . ;, ., ' ,, ' ' `: ~ ,
'' ' . " ~ ~'' ` ` , ~ , ' , ' '
~ 19 2~7~
a role in deciding the ini~ial f~rce~ when gripping and
lifting an object, but for the s~eady-state holding pha~e,
the ~oma~o~e~ory i~o~matio~ ~rom the skin effectively
defines the necessary force wi~h no app~rent relation to pre-
lift visual cues. The ra~ult~ repo~ted hexe suggest that an
FES system for precision grip in paralysed human~ could u6e
the same cutaneous afferent in~ormation a~ the intact
organism ~nd implement a slmilar ~ontrol strategy, thereby
adj4sting the ~rip for~e automatically to the weight and
surface ~tru~ture of ~he objeGt in hand in the form of an
"ar~ificial gripping ref l~x " .
In paraly~ed humans, tactile information ari~ing from the
th~mb, index and~or middle fingers could be recorded with
nerve cuff electrodes, either from the palmar ~taneous
branch of t~e med~an nerve p~o~imal to the wri~t, or from
individual internal digital nerve branches in the hand.
Experimen~s with a nerve cu~f implanted on the median ner~e
of a monkey (Milner et al., 1~1) have produced ~G -~ignal~
d~ring perturbation~ of p~ecision grip that closely re~bmble
~0 those repre~ented in the cat experiment~ by the pre~ent
inventor6 sug~esting that the median nerve may be a ~uitable
~ource of the de~ired signal.
~or the restoration of gait and ~tance 1n per~o~ i~paired by
paraplegia or ~troke, cutaneou~ feedback information
2~ originating from mechanoreceptor~ in th2 601e~ of the fe~t
could be recorded with nerve cuff electrodes implanted on the
internal and external plantar branche~ of the po~te~ior
ti~ial nerve, and the ~ural ~e~ve. The ~ibial nerve branches
contain afferent~ ~xom the medial and lateral aspects of the
30 ball of the foot, ~nd the su~l nerve contain3 affexent~ fro~
the ~eel. To avoid me~hanical damage to the ner~es or cuff
electrodes, the ele~trodes wo~ld ~e~t he installed proximal
to the ankle, rather than in the foot.
I~ st~oke patients who ma~ require o~ly electricsl ~timula-
kion o~ ~he pe~oneal nerve to control ankle dorsiflexion
209~ 7
during t~e ~wing phas~, a ~i~gle ~ural or tibial nerve cuff
w~uld ba ~uf~icient to monitor whether or not the ~ected
~oot i5 ~upporting wel~ht. Recording~ frvm the sural nerve in
man (Slnk~2r et al., 1991) hae demon trated a di~tlnct peak
in ~he nerve ~ignal at ~oot contact. In paxtially or oo~pl~t~
par~lçgi~ ~atients, wher~ the coordlnated activation o~ both
legs must be re6tored, three cuf~ recording electrodes in
each leg would provide in~orm~tion on mediolateral and
anteropo~ter~or welgh~ di~tri~ution on each leg, as well a~
on timing of foot contact and inter-limb load dl~txibution,
con~idered e~ential for ~u~cessful restora~lon o~ g~it with
~ES. although ~u~h lnformation ~an be obtained with external
preQsure sensor~ placed inside shoes, inherent problem~ of
calibration, ~echanical and electrical dri~t, lead breakage,
and ~en~itivity to en~ironmental factors like moiature and
temperature, th~t afect external tr~n~ducers, would be
avoided i~ nerve c~$f electr4des were used. Becau~e sati~-
~actory re~toration of gait often cannot b~ achieved u~ing
F~S alone, individual ~ol~t~on~ ~re likely to involve F~S-
ba~ed hybrid sy6tem~ tailored to the epecific pattern of~e~60rimotor def~c~t pre~ented by each patient.
For hand control application~, the tactile feedback recorded
~rom the median nerve or from ~ts ~ranche8 would regulate the
~utput of a portable mulsi~hannel FES ~ystem to control
~e~eral type~ o~ grip by stimulating for~arm and hand muscles
via, e.~., permanently lmplanted ep~my~ial electrode~. To
a~old ri3ks associated wi~h tran~out~neou~ pa~sage of lead~,
it would be pref~rabl~ that both the nerve ~ecoxding and the
musclQ etimulation information would be telemetered acros~
the ~kin. The command ~ignal~ for the FES ~y~tem would be
generated by the user, u~ing unaffected motor functionc. In
initial application~, regu~atory feedback would be
lmplemen~ed to uae slip ln o~mation to automatlcally update
the ~timulation parameter~. To also implement continuou~
~5 regulatory feedback, app~opriate algorithm~ would have to be
avallable to extract momen~-to-moment informa~ion on grip
force or other xele~ant p~rameter, fro~ the ~NG ~ignal.
1~2376BI.~OIIAS119~306 0~
~` 21 2097~7
The principle of the present lnvention may also be useful in
other ~pplications where physiologic~l sig~ls must ~e
sampled during stimulation o~ nearb~ muscles, e.~. in FES
systems USiXl~ the EMG from int~ct or partially int~ct muscle~
or ENG from e~i~ere~t nerves a~ a means to ~o~trol the stimu-
~ator or prothesis.
It i~ evident that ~he two abo~e main aspec~s o~i the
i~vention can be used independently or combined in o~der to
impro~e FES methods.
': ' ` ': '~ ':, :', : .: ` , ' ' , . : ~: ,
2~ 2~97~7
LEGEND ~O F~GURES:
Figure 1.
A) Principle for recordin~ a~d using natural sensory nerve
acti~ity from peripheral nerves in a sy~tem for functional
neuromu~cular stimulation. Figure shows as an example a
~ystem for restor~tion of upper limb fu~ction, but the
principle would be similar for lowe~ limb.
1. Telemetxic coupling between i~pl~nted and ex~ernal
equipment
10 ~. Implanted stimulator
3. Shoulder position ~enso~
4. Stimulation ele~txode~
5. External portable control units and batterie~
AMP: Amplifier for the ne~ral ~i~nal
15 ~BI; ~e~tifier ~ bin-integrator (see Figure lB)
~OGIC: Extr~ction of information and generation of control
signal
STIM: S~i~ula~or
~) Block dia~x~m o~ the artifact ~uppression technique and
met~od to co~t~ol a foo~drop stimu]Lator. The fir~t par~, tha~
corresponds to the RBI unit in ~i~ure lA, i~ where ~he
axtifact~ in the neural signal are ~uppressed. It consists of
the highpass fil~er, rectifier ~nd bin-integrator ~see text
for de~ail6). The second par~ peci~ic to the ~oot-dro~
prosthe~is application, and gene~ates a control signal for
the stimula~or based on the artifa~t free ne~ve ~ign~l. Thi6
~-- part corrsponds to the LOGI~ uni~ in Figure lA, ~nd con~ist~
o~ the band-pass ilter, rectifier, comparator and timer ~see
text for det~il~).
C) ~xample of how the ~y~tem in Fig~re lB can be implemanted
using standard analog elec~ronics. The p~rt~ separated ~y
d~shed lines corr~pond to the blocks in Figure lB. The
: ~ . :, ;;, . .
. .
23 2~97~
circuit here has bee~ used in a port~ble ~tem ~or footdrop
correction. See text for detailed description of the circuit.
Figure Z.
R~w ~i~nal recorded ~xom the ~ibial nerve cuff while nearby
~uscle~ were ti~ulat~d. Four sweeps o~ data are shown
~uperimposed. ~n the top trace, the b~ndwid~h was 65 Hz - 10
kHz, and the EM~ ~olleys were clearly present, wh~re~s in the
bot~om ~race, the signal was high-pass filtered at lOOo Ez
and the EMG pickup was pra~ti~lly removed. ~orizontal bar~
lo show the period~ duxi~g which the ner~e cuff signal w~s
bin-integrated ~d ~mpled by the computer.
AS - amplifier saturation
TO = Temporary tape o~erload due to saturation
Figure 3.
Veri~icatio~ of the a~ti~act removal method, The solid ~xa~es
~how ~he force (top3 and normal rectified, bin-integrat~d E~G
(bottom) during st~mulation of the four cal~ muscles in a cat
model under anaestesia. ~h~ da6hed traces ~how a similar
trial, with the tibial nerv~ blocked distal to the recordin~
cuff.
Figure 4.
~e~tified ~nd bin-inte~ted (RBI) hum~n ~ural ner~e cuf~
reco~ding~ while tapping with a finger on the ~kin within the
innervation area o~ the nerve. In the ~op trace, ~o elec- :
trical sti~lation is applied and the ner~e signal is no~
dis~urbed ~y artifactL~ ~hs middl~ trac~ shows the dist~r~ed ~ .
nerved signal when a 100 Hz stimulation is applied ne~t to
the nerve without ar~ifact suppre~sion. The bottom ~ra~e
shows ~he ~erve sig~l du~i~g e~e ~ame 100 Hz, but wlth
artifact suppression.
... .. .
' ' ' .~ , ' , . . ' .' ' ' ' ' ~ ' '
" ~ ~ " '. ' . ' ' I ~
~ '. ', . ~, , . ~ , ,
24 ~ 7 8 5 7
Figure 5.
Diagxam of elect~odes implanted in cat hindli~b and apparatu~
u~ed to measure ~eural respon~e~ during grip. The cat was
under anesthesi~. The limb was fixed at the ankle malleoli
and knee with atx~a~ic, cupped clamps. Forcas ~ere produced
by F~S via electrodes implanted in the ~our ankl~ dorsiflexor
~u~cles (MG, LG, Sol and Pl; not ~ re shown). ENG acti~ity
in the tibi~l ner~e was recorded wi~h a tripolar nerve cuff
electrode. ~hen the plantarflexor muscles were stimulated the
paw moved (curved arrow until the footp~d pressed against an
object ~hat could slide vertically along a low~riction
hearing. It wa~ covered with fine sandpaper o~ t~e s~rfa~e
contacted by t~e ca~'s footpad, and contained two ~orce
transducers to measuxe vertical (load) force a~d horizontol
(g~ip) f~rce, as well a~ a linear po~ition transducer to
measure vextical position.
`` 2s 2097~7
RE FEREN~ES
P.grles, W. F., MacCreery, r). E~., editor~, "Neural Prosthesis~
ln: dFundamerltal Studies~, Printice Hall, New ~er~ , USA,
l9gO .
CoL~a~. ~.B., I,l. Salisbur5r, A~ ol. Eng. S, pp~ SOS-~ll, 19~7.
tior~o~, A.~, }1, F~rssb~rg, R.S. Johansson, ~:;. ~cs~ Ylsul~ d~ cues ~ ~e progr~nF,
of ~an~p~lla~e fo~ces du~g prec~s~on ~p," ~p. Br~;n Rcs., 83, ~p. 477-482~ 1~91.
Haugl~nd, l~X., ~A. Hoffer, T. Sin~ "Ski~ contactforce i~f~a~o~ i~ elec~D~ ~aphic
sigr ~s ~:co~d by imp~ted ~lerve ~uff clec~dcs," JEEE rransaclzoAs on Biom~. En~.,
(sub~tted concu~ iiy).
Ra~glantl~ ~, J~. Hoffer, ~'Senso~y aervc s~nals ~ecorW by ~lg~nud cuff c~cctrodc~
duAslgfi~on~l e~ ic~l s~a~on of ~imb D:uscks,"lEEE ~oc~o ls on~iorr~d. ~ng.,
tsu~il~;a COD~n~ty),
~cr; JA., "TcchDiqucs to ~udy spinql-co~, p~ripheral n~. ~nd mu~cle activitS~ iA f~ely
m3~ ~a~s," NcuromcJfsod~, Yol. 15. pp. 6~ 1990.
J.A., ~u~land ~ ~nd Li, T~ Obta3~:ng skin ~ontact fos~e info~o~ ~m irt~plantcd
n~e cuff~ecording el~e~od~s, Pro~. IF~E L-ng. fn M~d, d~Biol. 30c, Inr~. ~or~ 928-
g~g, ~8~.
~o~e~, J~., Hau~ltnd~ ~ and Sin~, T. Fun~o~al resrora~on of ~ecision ~ip usl~ slip
~forma~n ob~ed f~ pesiphe~l ne~ve~eco~n~s. ~roc. Ann. ~ tl. Co~. IE~E Eng. in
~fcd ~ Biot. Soc. 13:~96-~g7, lg91.
Hoffer, ~A. and~laugland, M. Si~ls f~o~ tact;lg sens~rs in ~3abrous s}~n sl~itable for resPring :
motar f~ncdo~s in pa~alyzed human~ VRA~ ~RO~THESFS ~ Re,D~ b~g ~o~or
Fl~ncJion ~er Discose or ~ fsabil~ty, R.B. Stcin, ~EL Pe~ m ~d I~. Popov~c, edi~s. ~ :
Oxf~d U~i~. Psess, pp. 9~-12~, 1992.
Hof~cr, J.A., N. Sugano, (~.E. I,Deb, W.B. M~ks, hq.;l. O'I~ono~an, C.A. Pratt, "C~t ~indlimb
a~o20neumns dus;n~ 10con~ation. II. ~om~ tivîty pattent8t~ 30~ , of ~e~uop~ys. ~al, S7,
~o. 2, 3~p. 53~-553, ~eb. 1987.
. . ,
18~376131.M~/AS/19~3 ~
.: ~ : : . . : . ~ : : . , : . . . : .
:.: . . . :: : . , ,: ::
i . .: - ~: ~ I :: . . .
26 2~g78~7
Joh~ssDa, ~s., A~. 'It~o, "T~ sensibiliy i ~c human band: ~el~e a~d absol~
der~si~ies o~ pes ~f me~hano~ecep~e ur~iTs in gl~brous skin~"J. P~ ol. (Lond~ 2815,
pp. ~83-300, l9~g.
nsson, RS., G. Wes~ng, "~ s of ~labrous s~n ~eprors ~nd sensorlmolor ma~y in
au~oma~c c0~ f pr~isio~ ~p ~heJ~ ng rouE h~ ar n~ore slippay obje~ tas~
Res.~ 36, PF~ ~S0-5~4, lg84.
3Oha~s50n, ~.S., ~. ~es~ing, "Siga~ls in ~d~e ~ffere~ e fingers eliciPng ~p~l~e rno~
~sponses d~g pre~ grip," E;~p. Bra~ ~es, 6~. pp. 141~154, 1~87.
Joh~sso~. RS.. C;. Wes~ing, "Pn~gramm~d ~sd ~ggered ~e~n~ id loa~ ~ge8 d~ g
prcc~sic n ~ip," E~p. Br~in R~s., 71, pp. 7~-86, 1988.
l~ilncr, T~., C~ Dug~s, N. Pic8$d, A. SInltb "~usaneous ~ nt acti~ n the meti~n ~e . . .
dun~g ~aspnng in thcpr~a~c,~aln~cs., 548, pp. æ8 ~41, l~gl.
Si~ ugl~d, M., Haase, J. and ~of~e~, J.A. Whole senso~ ve re~s 1~ hu~
-~n app~icadon f~rne~al p~s~eses. P~oc. Ann. ~ntl. Co~ E; Eng. in ~ed. ~ Bisl. Soc.
900~ 91.
~P1lbO~ A.B., R~. Joha~s~on, '~p~es of sutancous ~cc~anor~ep~r~ he hu~Dan hand
rcla~ oucb sensa~on," H~n Ncuro~o~. 3~ 14, S~ger Verlag, l9~A
Wcs~ling, ~ RS. ~oh~sson, "Rcsponses in glabsous sldn mechanorec~s d~ p~slon
~p fn h~mans," E~p. br~in re~carch, N~. 66, pp. 128-140, 1987.
Zaj~c, E:~.. ~L. ~oung, '~schar~o prop~cs of hindlisllb moto~curons L~ d~brate ca~ during ~ ~ :
~ocomo~o~ ~nduced by mese~ceph2~ic s~mu~a~on", Jour~ of ~europhysiology, vol. 43, pp.
1221-1~35, May 1980~.
. . .
~, : , : ,~
;,: . . , : : . :
--` 2~8~
27
EXAMPLl~S
EXAMPLE 1
MINIMIZING ~ G~AR'rIFAt:~T SIGNAI GENERATEr) BY MUSC~E
~:TIMU~TION I~ A CAT E~CPERI?~ilT
5 ~_
In ~at~ which were surgic~lly ane~thetized, a 30-40 mm long,
2.2 mm I.~. silicone rubber cu~$ with 3 circumfer&ntial
~tainle~ steel wire electrodes (Cooner A~ 631) wa~ impla~ted
on t~e le~t ~ibial ne~v~, 2-4 cm proximal to the ankle joint.
A scia~ic ner~e recording cuff, 20 mm long, 4 mm I.D. with
three stainless steel wire electrodes, was implanted in the
mid-thigh region. Leads from the cuffs and other lmplanted
device~ coursed 0ubcu~aneou~1y to a~ external connector
mounted on the cat'~ back a~ described in ~offer, 1~0. A~ter
:L5 6urgery, catq were ~ive~ analgesics (Acepromazine Male~te ~nd
subcut~neous Morphine, 0.10 mg/kg) for at least ~4 hr.
Recording sessions ~tarted 4-7 day~ after implantation.
Nerve blocking cuffs:
To exclude the participation of motor activity in the ENG
~0 recorded from the tibial nerve during walkin~, in several
cats a blocking cuff 8 mm long, was placed on the tlbi~
nerve, b~tween the ti~ial and sciatic recording ~u~s. ~xonal
condu~tion w~ blocke~d by infusing lidocairle sodium solution
(~), via a catheter that 1~3d to the bloclcing cu:E:E from a :
25 ~ort in the external connector. The conduction block was :
asse~sed from the pr~gre~si~e reduction of the compound
action ~otential xecorded by the s~iatic ner~e cu~f, e~oked
by stimulation o~ the tibial ner~e at the distal r~cording
cu~. Usually, the tibial nerve was completely blocked after
30 20-30 mix~ute~. ~t ~che end o e~L~h expe~im~nt the block wa6
rever~ed wi~ usion of norma~ ~aline solution. . .
'. , . , ' : ': '. '.':'~ ~ ' "' i '. ' ~ " ' ::
2~ 2~97~7
~n o~hex cat3, a tibial nerve blocking cuff wa~ placed dis~al
to the tibial nerve re~ording ~uff in order to identify the
contributions fro~ footpad ~erents, and the pre~ence in the
recording cuff ~i~nal o~ any artifacts cau~ed by the ~timuli
s a~d/o~ compound EMG potent~al~ d~ri~g FES of nearby mu~sley.
ata ~ollection:
~ average once per week during a 1-3 mo~ths period, each cat
w~s anaesthetized with halothane g~s, the left ~oot wa~
shaved, remaining hair was removed with depilatoxy cream, ~nd
~he leg was s~cured at the ~nkle malleoli ~nd knee with two
pair~ of cup~ed holders. In the first series o~ experime~t~,
a servo-controlled motor wa~ u~ed ~o pu~h perpendicularly on
the ~entral footpad with ~ 1 cm di~c~6haped probe, ~pplied
fo~ces were monitored by a ~ries tran~du~ex, The ~osition
and compli~n~e of the motor were electronically re~ulated
with posi~ion, veloci~y and ~or~e f~edback. Control 8igna
were generated with ~n IBM-compatihle l386 computer. ~e
tibial E~G wa8 analog-rectified ancl bin-~ntegrated i~ 1-lo ms
bi~ ak PSI-l). ENG, m~tor position ~nd forc~ d~ta were ~:
digitized on-line (100 Hz/channel) with the same computer.
Removal o~ arti~c~
Two ~teps are used to reduçe the ~mplitude of arti~act~ from
near~ mus~le activity (EMG) in the nerve ~ctivity ~ENG
si~nal): 1) High-p~g~ filtering and 2) ~ynchronization of
6ampling and stimula~ion.
) The freq~ency di~t~ibution~ of ENG and EM~ re~orded by
tripolar nerve cuf~ electrodes are largely ~on-over-
lapping. Most o~ the E~G could th~r&~ore ~e ~extracted~
from the cuff signal by filtering the ~ignal w~th a ~harp
hiyh-pa66 analog filter at 1~00 ~z (Ithaco model 4302,
set at 80 ds/decad2).
~) Bec~use the times of 6timulation were k~ow~ a~d both the
~timulatlon artlfa~t a~d the EMG CAPs were limited in
8~376~ 1~306 P4
- 29 2~978~7
~ime, it was pc~6~ible to reduce th~ arti~ac~ pickup ~ub-
stantially ~ only u~lng the cuff ~i~nal at the end of
the inter-pul~e int~rval, i.e. the sa~pling wa5 locked to
the ~timula~lon, resulting in a ~ampling frequency of
loo Hz.
Figure 2 ~hows 4 superimposed recording~ o~ the cuff signal
durlng FES. Each record wa~ 40 m~ long and thus included 4
~timuli. The four tra~es were ~yn~hronized to the time ~f
stimulation of the soleus muscle. Each tim~ a muscle was
stimul~ted ~abelled 'Stiml), the fi~t event in the ~uff
signal w~ the ~timulation axtifact, ~howing up ~ a narrow
spike that would vary in ampl$tude depending on the
~timulation inten~ity and the ~uscle that wa~ ~timulated.
This was followed by the compound EMG volley fxom the
~timulated mu~cle, ~howing up a3 a ~lower, large ~mplitud0
wave in the cuff ~ignal (la~elled EIN). T~e shape of this ~MG
bur~t depended on the stimulus in~ensity and ~he mu~cle being
~timula~ed, but ~ otherwise very repeatable, The neural
~ignal itself wa~ the hi~er-frequency signal, 5 ~V in
amplitude, onto which the artif~ct~ wer~ ~dded.
Th~ data in Fi~ure ~ were obtai~ed by record$ng the cu~
signal on an FM ~ape re~order while~ the co~pute~ c~ntrolled
the stimulation intensity using external foxce feed~ack, as
de~cribed above. The signal was then replayed and ~ample~ at
a high rate t20 kHz) to produc~ the. ~igure. For large
po~itive arti~act amplitude~, the amplifter satuxated (marked
AS) and for la~e ~egative amplltudes, the t~pe record~r
overloaded (marked TO) ca~ g th~ nal to be zero during
the overload.
The effecti~ of filteri~g ~an be seen by comparing tihe top a~d
bottom tra~es in Figure 2. In the top panel, filtered b~tween
~5 ~z and ~O k~z, the pickup of the EMG volley ~howed up very
clearly. In the hottom panel, the same ~ignal wa~ ur~her
high-pa~ filtered at 1 kHz, and the EM~ çontamination wa~
3S laryely removed.
I~37~x.~ ~306~
2~78~7
The ~timul~tion artifacts were bla~ed out by sampling the
BNG ~nl~ durin~ period~ _ -between artifact~. Periods of
artifacts are normally e~ily located by visuel inspection of
he ~NG-signal since remain~ o~ stimulus-related EMG, direct
stimula~io~ ar~i~act~ and amplifi~r r~covery are ~ynchronized
~o the stimul~tion, which i8 no~ ~e case for the ENG. ~he
bin was defined so that none o~ the~e ~rti~act~ we~e apparent
within the window. Instead, we used a re~ti~ier/integra~or
(B~k RBI-l) that had an adjustable inte~r~tion period a~d was
reset in synchrony wi~h an external clock. Thi~ clock was
supplied by the ~tlmulator, and it was thus pos~ible to
integrate the E~G signal in bin~ that las~ed 3-4 ~ and that
included only data from the la~t pa~ of each inter-stimulus
interval (~o~izontal bars in Figure 2~. The ENG in eac~ o~
15 the~e periods was ~ectified an~ in~egrated, resulting in a
sin~le value ~or each bin that was sampled just before a new
stimulation pulse was elicited.
The ~alidlty of thi~ method of noise or artifa~ suppre~sion
WAS demonstrated by the ~ollowing ~xperiment. ~he muscle~
20 were first stimulated to generate a force profile show~ by
the ~olid trace in Fig~re 3, top panel. Thi6 gavç rise to the
~NG signal shown by the ~olid trac~3 in Pigure 3, bo~tom
panel. Wi~hout changing the setup, conduction in the ne~ve
was th~n blocked with infusion of 0.4 ml lidocaine to the
blocking cuff. A~ter approximately half an hour, the ENG
~ignal recorded from ~he tibi~l nerve was insensitive to
touching a~d squeezing o the ~oot, demonstrati~g that
afferent conduction wa~ completely blo~ked. The -~timulation
was then repeated and a similar force profile was generated,
(shown by the dashed line in Figure 3, top pan~l). The
sampled cuf f ~ignal was ~ow redu~ed to the ~lat dashed line
in ~igure 3, bottom p~nel, i.e. containing ~either E~G, nor
s~imulation artifacts or EMG. Since ~he only difference w~s
the blocking o~ the ~ex~e distal to the reco~ding cu~f, thi~
experiment showed that for the unblocked ner~e, the sampling
method removed all introduGed artifacts and only ~ampled ENG
~c~i~it~.
IU376EX.0011AS1199306 04
31 2~78~7
EXAMPLE ~
FOOTDROP PROSTHESI~ ~N A H~MIP~E~IC SPASTIC MALE
For the footdrop prosthe~i~, the neural ~ignal from the sural
nerve can be u~ed ~or detection of foot-contact.
S Hi~hpa~s-filtering at 1 Hz followed by rectiEicat~on and
threshold ~omparison reliably detects when ~he hebl touche~
the floor during walking. Thi~ h~s been u~ed ~o ~witch on and
off a commercially available peroneal stimulato~ lKDC 200~A)
a~d in ~his way, replaci~g the usual heel-switch ~at mu~t be
mounted in the shoe with the ~atural sensors in the 6kin of
the ~oot.
In the present example, the sural nerve in a 35 year old
hemiple~ic ~pastic male subject with a drop foo~ was
instrumentecl with a ~ripolar whole nerve cuff electrode
15 approximately 7 cm proximal and 3 cm pos~erior of the ~atsral
malleolus of the right ank~e joint. A cli~ical examination
had xevealed tha~ the ~atient had an Achille~ tendon con-
tracture and tremor ~round the ankle j oint. ~he subject gave
his con~ent and the ~tudy wan approved by the Local Ethioal
Committee.
Surgical procedure
The surgery was performe~ during lo~al ~naesthesia. To
pre~ent compres~ion-neuropathy a~oci~t~d wlth post-surg~cal
oedema, it was ensured that the inner dlameter o~ the cuff ~.
wa~ more tha~ 3~ la~ger ~han the ner~e diame~ex. The th~ee
Teflon-coated multi~t~nd ~tainless steel l~ad out wixes
~ooner Wire Company, ~5A) from the ~uff elec~rode we~e put
through ~he skin approxir~a~ely 25 ~m abo~re ~he la~eral
malleolu~3, The rlerve cu~ was placed so that the nerre was
30 neither pulled xlor tor~ed by 'che wires. Thls make~ the
long-tenn prognosis of a nerve prepa~a'cion excellent.
ls237~;~lAsll~3o6 o~
32 2~7~S7
Nerve cuff electrode confiquration
The nerve cuff recording ~lectrode consisted o~ an insulating
cuff (silioone tubing) cont~ining thr~e circumferential metal
electrodey (~lexible 40-str~nd stainle~s ~teel wir~, ~e~lon-
co~ted~, placed around pa~t of the sural nerve. The de~ign,fabxica~ion and surgical implan~atlon of nerve cuff r~cording
electrodes have been reviewed in detail elsewhere (~o~er,
lg90). The in~ulating cuff ~erve6 to re~ol~e the sm~ll act1on
currents gener~ted by nerve fibres, by constraining the
current ~low within a long, narrow resi~ti~e path. In this
applic~tion, the ~uff wa~ 3 cm lon~ ~nd had an inner di~met~r
of 2.5 mm.
eural ~e~ g~Sion_and s~imulation
The leads ~rom the implanted nerve cuff electrode ~ere con-
nected to a dif~erential ENG ampli~ier with a high common
mode rejection. The ~r~n~cutaneous ~timulat~on was m~de by
means of a reference electrode abo~e the tibiali~ a~terior
muscle and an ac~lve ele~trode above ~he common perone~l
nerve just distal to th~ br~nching off of the ~uperficial
~O peroneal nerve.
The neural amplifierrwas ba~ery-~upplied a~d optlcally
isolated ~rom the mai~s ~o increase the common mode rejection
and to reduce the risk to the gubj~ct. To further reduce
noise pick-ups, an external ref~rence electrode wa~ placed
b~twee~ the stimulation elec~ode~ And the nerve ~uff
electxode. T~e ~eural signa~ wa~ fourth-order b~nd pass
filtered from 0.7 to 10 kH~ (Kron-Hite, model 3750). This
would red~ce remai~i~g pick-up of 50 Hz from the m~ins, if
any, and the EMG from ~eigh~ouring mus~les to a negligible
level. ~he bandwidth of th~ record~d neural ~ignal ran~ed
~ro~ 0.2 to 3.0 kHz.
1~376EX.WI/AS/1~306 W
`~ 33 2~97~7
and wlthou~ removal o~ arti~act~ ~xom_ ~he nerve ~iqnal
The ~ural nerve activit~ in a human Eubject was r~cord~d
while tapping with a ~inger on the skin w$thi~ the inner
vation ~rea o~ ~he nexve. Fig. 4 (top) ~hows the ampli~ed,
recti~ied and integ~ted (RBI) nerve si~nal without an
el~ctrical Qtimulation. A ~lear peak in the ~erve signal was
observed when the finger touched the skin. Fig. 4 (middle)
show~ the RBI sural nerve activity when an electric~l ~timu-
lation every lO ms (100 Hz) was applied ~ex~ to the nerve.
The electrical 6timulation eli~ited an a~tion pote~ial in .
the nerve r~cordings approximately 2 m~ after the st~mulation
~i~h a duratio~ of approximately 2 ms (not ~hown). Fig. 4
~middle) shows how thi~ unwa~ted ~ignal incr~ased the
xecorded activity making it ~ery difficult to detect thenerve re~pon~es caused by the finger-tapping. Applying the
~rtif~ct suppression techni~ue by samplin~ the nerve signal
in a window from 5 to 1~ ms after each stimulation, the art~-
facts ~ere removed and the ~erve ~esponses to the finger-
tapping were again reco~nized in the cuff electrode recor-
dings tfig. 4 (bottom)). The ~ecti~ied and integrated nerve
signal in each of the sampled periods after a stimulatio~
re~ulted in ~ single ~alue ~or each bin that was ~ampled ~us~
be~ore a new stimulation pul~e was elicited. The artifact ::
~5 suppression is made exactly as desGribed for the cat data in
Exampl~
BXANPLE 3
Slip-detection and ~ompensation in a eat model of human
position grip
The hindlimb of anaestheti~ed cat~ was used as an
experime~t~l ~odel for the paralysed human limb, with the
central footpad as a mod~1 gla~rou~ skin. Three cat6 (4-~ kg)
were chroni~lly implan~ed using asep~ic technig~es,
~ :
l R~37~ A~ g93 o6 o~l ~
' j : ' ' ' ' .................................... : ' '
' " . '~ ' ", ~ . , , ~ , , ' :
, ~4 2~7~7
~ollowing the procedure~ and guidelines described in ~xample
1.
Bipolar intramus~ular stimulation elec~rodes were im~lanted
in each o~ ~our ankle extensor muscles: Medi~l a~d lateral
ga~trocnemius (~G, LG), ~oleu6 (Sol~ an~ plant~ris (Pl). The
electrodes ~onsisted of two Teflon coated, 40-~tr~nd ~tain-
les~ ~teel wire~ (Cooner 634), with the e~d~ dein~ulated for
15 mm, and ~ere in~ert~d in the muscle diagon~lly to the
fibres, about 2 cm ~part in the proximal part of the muscle
(see Fig. 5). Bipolar ~timulati~n electrodes were used,
rather than monopolar with respect to a common ~ro~nd,
because of the higher 6electivity attained~
A tripolar nerve recording cu~ ~30 mm long, 2.2 mm I.D.) was
implanted on the tibial nerve, distal to the mu~cular
branches; and 4-5 ~m above the ankle. A~ this level, the
tibial nerve contains mainly Affere~t fibre~, mo~tly from the
plantar Yur~açe of the foot and the ~ootpads, and the cuff
could ~e implan~ed without obstxuc~ing the blood ~upply to
~he nerve or causing mechanical damage to t~e nerve.
The cats recovered for at least 3 days ater ~urgery ~efore
the first experiment was performed, in order for the
implanted devlce~ to sta~ilize wlthin the l~g. At the
begi~ning of each recording session, ~he cat was
a~ae~thetized with an intr~enous injection of Thiopentothal
(8-10 mg/kg) throu~h a catheter implanted in a superficial
jugular vein, intubated and maintained under anae~thesia with
Halothane in a mixture of nitrou~ o~ide and oxygen.
To remove sen~oxr affe~ent ~ontribution~ from hair receptors
in ~he skin surrounding the centxal footpad, prior to ea~h
experiment the ~oot was sha~ed ~nd treated with ~epila~ory
cream, followed by ~ thorough wash and application of
moisturizing cre~m.
Ig~376BX.WI/A~/19930C 04 ~
:
`` 35 2~3~7~7
During an experiment, the cat waE~ supported ~y a heated
sling, with the implanted hindlimb fixed by two pairs of
cupped claTnps pressed around ~he an1~le ~nd knee jOlnt8 (~ig.
5). This allowed the a~kle joint mo~re and did 3lot do ~erious
5 damage to the 3}cin. The ankle and knee angles were 100. The
foot huny Yertically w~exl ~o~ ~imulated. When the ankle
extensor mll~3cles were ~timulated, the ~ootpad p~l~hecl
horizontally again~t a ~e6t object that could s~ide
ve~tically with low fri~ion alo~ two bars, and that would
fall if held by ~he foot.
I~:? analogy to the precisios~ ~rip experimenta by Westling and
Johansson ~19~7), th~ o~ject w~s equi~ed with force aensor~
that mea~ured the horizontal grip force ~nd the vertical lo~d
~orce by means of strai~-gauge force transducers ~Revere,
FT50). T~e vertical position of the object wa~ measured with
a m~ch~nical linear po~ition transducer (W~ters, LRT-S-lOOB)
(Fig. 5). ~ravity caused a constan~ downwards pull of 1.4 ~,
since the object weighed abou~ 140 grams. The ~urface of the
obj ect was 600 grit sandpaper .
~0 The s~imulator we uaed for F~g ~as the same a~ in ~xample 1,
produced rectangular monopha~ic co:n~tant current pul8es with
a fl~ed ~mplitude for ea~h channel. Pulse widths were
independently controlled for each of the fo~r muacles between
û and 255 ~lS, in s~eps of l~s, by a '386 computer via a
parallel po~t. Each muacle was 6timulated at a fixed
frequenc~ o~ 25 Hz, i.e. with interpulae intervals of 40 ms.
To re~uce ~orce rip~le ~au8ed by unfused tetani, the four
muscles were ~timulated se~uentially, so that o~e of the
muscles wa~ ~timulated every lO ma, giving ~n ~ggr~gate
stimulation frequency of 1~ Hz, e.g. the ~me a~ de~cxibed
in E~ample l.
The ENG si~nal xecoxded from the tibial nerve was filtered :~
~lk-lo kHz bandpass), rec~ified ~nd bin-integrated (in one 3
ms bin every 10 ms) be~ore sampling. This procedure ~a~ used :~
3$ to cancel out the pickup o~ artifacts as de~cribed in Example
1~2376EX.0011AS/1~9~06 04
. :: ,: . ` . ` , : : .
~': ~ ~ : .' : , : ` :
~~` 36 2~7~
1, ~nd also allowed the u~e of a l~wer ~amplin~ fre~uenc~
(100 H ) that the ~requency necessary to sample the raw E~G
~20 kH~. The bin-integrator and sampling were ~ynchronized
to the delivery of stimulation pulses by the com~uter. In
khis way the EN~ envelope ~a~ ~mpled at the sa~e $requency
as th~ ~ggregate ~timulation rate (100 ~z). In the following
the term llE~Gll will ~efer to the envelope of the ENG rat~er
that to the raw ENG, since o~ly the envelope w~ ~pled ~or
feedback purpo~e~.
lo Resul t~
To investiga~e slip-related info~matio~ contained in ~he E~
~ign~l, the following initial experiments were done: The four
plantarflexvr mu~cles ~ere stimulated with a train of pul~e~
o~ constant width, thus generating a force that was
approximately const~nt. The pulse width wa~ chosen 80 that
the generated fo~ce was not suf~lcient for the foot to hold
the ob~ect in place. Uhder these condition~, the ob~ect had
to be partially ~upported by the e~perimenter, who ~ould thus
allow it to fall by relea~i~g it~ ~upport in ~m~ll 8teps,
each step cau~in~ a slip betwe~n the pa~ and the object as
de~cri~ed in Hof~er and Haugland ".~92.
The ~harp bursts of ~G activit~ that signalled when ~lip8
occurred were typical ~or all e~perime~ts and f~r all cat~
used in thi6 ~tudy. Slip-~el~ted burs~s were di ti~ct eno~gh
from the backgxound ENG to be detected wi~h great ~u~acy,
and very early in the sl~p phase.
The sharp bur~ts.o~ ~NG activit~ that ~ignalled when ~lips
occurred werf3 ~pical ~or all experime~t.q and for all cats
u~ed in this ~t~d~. Slip-relaSed burst~ we~e dl~in~ enoguh
30 from the ~ack~round E~ to be detected with great accuracy,
and very early in the slip phase.
lR~ nnllA~ nftn~ ~
37 2~97~57
A. Detect1on of sli~
Sinçe ~lip~ were accompanied by ENG bur~t~, they could be
detected comparing the di~eren~ial E~G to ~ th~eshold ~alue.
Simple dif~erentiation, through, calcula~ed a~ the dif~erence
~e~ween the present ~Ç v~l~e and an old ENG ~alue, pro~ed
too noi~y. In~tead~ a "slip detec~ion" ~ignal waa calculated
by subtrac~ing d lo~-pass ~iltered (time con~tant = 0. 28)
ve~ion of ~he F~NG ~l~nal dela~ed by 20 sample~ (20Q~) from
the ~unfilte~ed' E~G, thereby remo~ing the b~ckground
activity from the ENG. The 'unfiltexed~ ENG al~o needed some
~i~tering to reduce no~e, but thi~ was done with a ~orter
time con~tant (0. 07s). The set of time constant~, ~ela~ and
threshold value save the most sensitive and robu~t ~lip
detection were ~ou~d by trial and error.
Implementation of the d~tection algorithm wa~ done in C
(Tur~o-C 2.0, Borla~d), and the filt~rs were i~plemented a~
~irst order auto-regressi~e (AR) ilters, which are computa-
ti~nally ~ery ~imple. ~he algorithm to detect slip is show~
below:
repeat (this loop xu~ a~ 100 Hz)
Update 20 samples o~ old E~G v~l~e~ :
Sample new ENG v~lue
Background~G~BackgroundEN~*a+OldENG* (l.a~
SlipENG-SlipEN¢~b+NewENG*(1-b)
SlipDetect-Slip~NG-B~Ckgr~UndENG
if SlipDetect~Thxe~hold then ~iqnal that ~ ~lip o~curred
e~d
The constants a a~d b were de~ermined from the time constants
di~cussed abo~e (for O=~.2~ a=0.~51 and n,o . 07s--
~
~Y0.867). Within the al~o~thm, the mu~cle stimulation inten-
sitie6 were also determined and pulse~ were ~ ed by the
computer, as de~cribed below.
18Z376~X.~I/AS/19~306 04
38 2097~57
B. Increase of force after slip
A~ ~oon as a slip was detected, the fo~c:e was in~rea~ed as
fast as po~3ible to reesta~ ;h a secure grip of the object
hefore it moved out of reach. This was done by one of the
~ollowing two di~exen~ method~ By introduoing immediate-
ly a single, closely ~paced pair o~ stimula~ion ~ulses (~dou-
blet") to each of the four mu~cle~. ~his i~ a technique thatis also u~ed by the central nervous system i;n ~o~rnal çondi-
tion~ ~nd can cause t~e force to not only increa~e rapidly,
but al~o remain high for a prolonged period afte~ an extra
pul~e. The ~ime be~ween the two pul~e~ was set to 5 ms. 2).
By increasin~ the pulse width ~arkedly for a ~hort period
after the 61ip, in order to recrult ~o~e mo~or units~
Once the gr~p wa.s re-establishad, it w~s maintaimed by incre-
~ing the pul~e width moderately (rel~tive to the original
PW). ~i~h ~his approach, the grip force change~ cloaely
resembled those seen in hum~n~ d~rlng the ad~ustment to
~udden increases in load force (~ohansson and Westling,
1988).
Because the ENG was sampled at 100 H~, a maximum of 10 ms
could elapse b~tween ~he detec~ion of a slip ~om the ~NG and
the response from the controlle~. Since the ENG w~s ~a~pled
jwst before each stimulus pulse wa~ elicited, ~he fastest
response ~o a detected ~lip wa~ determined only by ~he time
it took to do the calculations (le3s th~n 1 ms). The main
delay in the ~lip detection was caused by the low-pass f il-
tering of the EN~ that wa~ necessary to remove ~alse detec-
tions caused by normal variations in the ENG. The time-con-
stant for thiæ ~ilte~ wa~ 70m~, as de~ribed a~ovQ. The delay
30 fxs~m ~he monl~nt the object first started to slide dow~ un~il
the ~lip w~ detected w~R betwee~ 50 to 100 ms, ~ut ~aried
considera~ly.
doublet pulse caused the next ENG sample to contain addi- -
tional stimulus artifact, but this wa~ not a problem, becau~e
t~2376~X,OOl/AS/19~3~ 04
` ~g 20978~7
once the controlle~ was switched i~to "~lip comperlsation
mode", it did no~ req~ire or expe~t ~lid ~ample~ fo~ the
ne~t 300 m~. Thls p~evented e~xoneo~ detec~ions o~ ~lip
during rapid in increase~ in stimulation intensity, which
predictably gave ri~e ~o ph~sic EN~ burst~ ~hat could
resemble t~o~e caused ~y a 81ip.
The stimulation ~lgorithm in~luded an ~utomatic ~hut-o~f o~
stimulation in the ob~ect dropped ou~ or rang~, determined by
monitoring the signal from the vertical position transducer.
lo C. Te~t o~ closed-l~oP_slip com~en~ation controller
Two set~ of expe~iments were done~
1) In experiments that ~ lated progressive "~atigue" in the
stimulaa~ed muscles, the stimulation ~tarted at ~ le~el higher
that the minimum nese~ y to hold the object. The inten~ity
of sti~lation was then decrea~ed at a constant ra~e, until
the ~or~e became in~ufficient to hold the object, whiah then
~t~rted to alip. De~ection of the ~lip triggered the control-
ler to lncrease the intensity and gra~p the object a~
detailed ln the preceding ~ectiOn, i.e~. an llartificial grip-
ping reflexll was ellcl~ed. ~he in~en~ty of ~timulation wa~~hen slowly decreased at the same rate as prior to khe slip.
2) In experime~t~ ~hat ~odeled the response to lncrea~es ln
external load, the st~mulation inte~sity was held constant at
a le~el ~ufficient for the foot to hold the object. After a
fi~ed time, an extra load wa~ dropped o~ the object, causing
the obiect to ~lip and start to ~11. The ~artificial reflex
loop" cau~ed th~ slip information obtained from the neur~l
~ign~l to increase the ~timulation inten-~ity and thu~ o
the ~rip foxce, in orde~ to ~atah the object before it fell~
1a~q7~1:Y nn~A~I1aoqn~n~
:,' ' . , , , ~ ~ , ,., ,,,: ''
