64. Lilly, John C. 1961. "Injury and Excitation by Electric Currents." Chapter 6 in
Electrical Stimulation of the Brain. Daniel E. Sheer, Ed., Univ. of Texas Press for
Hogg Foundation for Mental Health, Austin, Texas. P. 60-64
O
of
Reprinted from ELEcmrcAL STIMULATION
OF THE BRAIN,1961
Printed in United States of Ame~ica
CHAPTER6
Injury and Excitation by Electric Currents
A. 7he BalancedPulse-Pair ~aveform
JOHN
C. LILLY
: : :
·-;
··
CHAPTER
6
Injuryand Excitationby ElectricCurrents
A. ~e BalanceolPulse-Pair ~aveform
JOHN C. LILLY
OUR GOAZ.HASB~N to ~nd an electric current
the proper parametric regions, can have thresholds
waveform with which animals could be stimulated
less than those for excitatory processes. The tech-
through implanted electrodesfor several hours per
day for severalmonths without causingirreversible
nical problem in chronicbrain stimulationis to stay
above the excitatorythresholdand below the injury
changesin thresholdby the passage of current
thresholdin the neuronalsystemunder considera-
through the tissue. We have found one such waveform (Lilly et at., 195~ia, b, c), shown in Fig. 6.1.
There
are certain
ultimate
limitations
in the use of
any electrical waveforms, including the one illustrated
here.
The
search
for
an ideal
waveform
~_+
is
being continued;the ideal one is only approached,
not achieved, by the waveform of Fig. 6.1.
The present chapter treats of the thermal energy
limit and the electrolytic limit which are reached
E
as the neuronalfhr~shold
data are tracedthrough
If
large systematiclchanges of waveform. However,
the discussion
concerns
only chronic
persistent
stimulation--that is, 1 to 3 trains per second for up
to 18hoursper day,5 to 7 daysper weekfor 6 to
100 micr6seconds
12 months. It does not concern acute stimulation
at one session or stimulation for one hour per day
Fro. 6.1. Waveformof stimulatingcurrent: pulse pairs
of current resulting from quasi-differentiation,with pas-
for several weeks. Many waveforms,including 60-
siveelectricalelements,ofa rectangularpulse.Measured
cps.sine-wave
current(Oldsand Milner,1~54),
at 2 percentofthepeak,theduration
ofthepositive
can apparently be used safely for these limited
schedules of stimulation. In our experience they
cannot be used for the intensive, long-term schedule of chronic stimulation. (See Lilly, 19~5~,19S7a
to d, 19~5~a,
b, c, 1959~a,
b; Lilly et at., 19~5~a,
b.)
pulse (upward) is 34 pee., and the duration of the
negativepulse (downward) is 28 ysec. The areas under
the two pulses are equal; thereforethe negativecoulomb
flow is zero for the pulse pair algebraicallysummed
during a time interval of 200 CLsec.from the beginning
of the positive pulse. (From Lilly et at., 1955a.)
IN~URY
tion. This result can be achievedmost easilyby the
Electriccurrent passed through the brain can
properchoiceof waveformsand their timecourses;
cause at least two distinct types of injury: thermal
and less easilyby the choiceof the range of repeti-
and electrolytic.
Thesetwo typesare separableby
tion frequenciesand train durations(Lilly et at.,
the properchoiceof electrical
parameters
and,in
1~5~2).
As has been amplydemonstrated
(Liddell
61
INJURY AND EXCITATION BY ELECTRIC CURRENTS
200 kilocycles) can cause thermal injury by overheating the tissue without excitation and without
electrolytic
injuryif no recti8cation
is allowed
to
occur tin the circuit, and especially at the elec-
3
trodes).
Apparently
theamount
ofelectrolytic
in-
juryisrelatedin greatmeasure
to thetotalquantity
of unidirectional electricity (coulombs) passed
through the tissue in a given interval without
reversals,whereas the amount of thermal injury is
more closely related to the total energy (watts)
passed.
i~ol dr PuLsE-l:~uRnTloN,
~ec 100
100o Boththesetypesofinjurycanbe shownto exist
Fro. 6.2. A typical "strength duration" curve for a
cortical zone using unidirectional rectangular pulses.
The threshold current (at 60 pulse pairs per second, in
trains 5 seconds long repeated at 30-secondintervals)
is shown as related to the pulse duration, on a log-log
in some other waveforms in the intensity ranges of
interest in brain stimulation. As an example, in the
case of unidirectional rectangular pulses we choose
the threshold for the amount of excitation of, say,
cerebral cortex, to excite movement in the periph-
plot.Beyond
about5 msec.is the"rheobasd'
plateau, ery in the unanesthetized
monkey.For any given
i.e., "constantcurrent"threshold;such long pulses
causerapidelectrolytic
injury.Below0.1msec.is the
pulse durationwe choosea repetitionrate suf~cientlyhigh (50 to 200per second)so that we can
45" constantslope showinga first-powerrelationbetweencurrentand pulseduration;this is the "constant
keep the thresholdcurrent low. As we increasethe
pulse duration, the current at threshold becomes
quantity" region of excitation. In the very short (0.01
lower until we reach rheobase, at about 5 msec.
msec.)durationregionsuchpulsescan causethermal
injuryat threshold,
(Lillyet al., 1952).But as we changepulseduration, the quantity of electricitypassed and the
energy dissipated at each threshold value are also
and Phillips,1951;Lillyet al., 1952;Hess,1954c; changed.
If theelectrodes
arenonpolarizable
(Lilly
Cure and Rasmussen,1954), these parameters also
i~uence excitatorythresholds and possibly deter-
et al., 1952), the resistance is constant with time
and with pulse duration changes; and the energy
mine whichsystemsare excitedin the vicinityof
(watts) per pulse is given as peak current (am-
the electrode.
peres) squared times the resistance (ohms) of the
Our ~rst emphasisis on avoidanceof injury.We
electrode.The amountof electricity(coulombs)
proceedon the basisof accepting--as
threshold passedper pulseis the productof thepulseduraexcitationof any or all systemsat the electrode
tion (seconds)timesthe peak current(amperes).
tip--thevaluesof currentwhichgivea visuallyor
To illustratethepointsinvolved,
somevalueshave
tactuallydetectable
response
or changeofbehavior. been calculatedfor a representative
implanted
(Of course,other methodsof detectioncan give
electrodein a monkey.The resultsare shownin
lower threshold values, and other waveformscan
Figs. 6.2 to 6.5, in all of which pulse repetition
exciteothersystemsat the electrode['hreferentia~·
frequencyis held constantat 60 pulsesper second,
excitation],but these subtletiesare not germane at
and train duration at 5 seconds.
thispoint.) Wearelooking
forthosecharacteristics
Figure6.2showstherelationbetweenthethresh-
of waveforms which cause excitation but do so
old current and the pulse duration. The classical
withoutinjuringbraintissueaftermanyhoursor
daysof stimulation.
rheobaseis difficult
to obtainbecauseof the elec~olyticinjury,whichraisesthe thresholdrapidly;
The two types of injuryare possiblybest illus-
trated by the effectsof two extremety~s of
the valuehere is an extrapolatedestimate.
Figures6.3,6.4,and6.5are calculated
fromthe
waveforms.Unidirectionaldirect current causes
data of Fig.6.2.Figure6.3givesthe thresholdpeak
electrolytic
injury(Horsleyand Clarke,1908)but
-energyper pulse;Fig.6.4,the thresholdquantity
creasedto, zerovaluesufficiently
slowlyto allow
accommodation
to takeplacein the nervoustissue.
pulseduration;and Figi 6.5,the energyplotted
againstthe quantity.In eachof thesegraphsan
little excitation,if increasedfrom, and then deHigh radio-frequency
currents(greaterthan, say,
of electricityper pulse, each plotted againstthe
estimateis given of the thresholdfor the two types
TECHNIQUES
ofinjury:
thermal
inFig.6.$;electrolytic
inFig.
6.4; both in Fig. 6.5. However, it is suspected that
theelectrolytic
injurythreshold
isverymuchlower :r
than this value; the lower limit is yet to be deter-
mined carefully for stimulations lasting many days.
%
The thermal injury threshold is an estimatebased
on the observation of the breakdown of water tin
normal saline) to steam for single pulses at a metal-
lic electrodewhoseresistancewas the sameas the
r,
q
Elaclrolylic
injury
·electrode used in determining the data on these
graphs; the values for tissue damage are probably
lower
than
the
one
estimated
here.
This
threshold
o.l
is a function of the area of metal exposed and the
Ir,
to
too
looo
PULSEDURATION.
msec.
size and ef~ciencyof the heat conductionpath away
Fro. 6.4. Quantityof,electricityper pulseversuspulse
from this electrode interface; howevei, the value is
duration at threshold. For· this electrode, waveform, and
of the approximately correct order of magnitude for
train parameters, the electrolyticinjury threshold is at
practical electrodes. As the area of the metal in con-
about 0.5 microcoulombs per pulse at about l.0-msec.
tact goes down (smaller electrodes) and hence the
resistance goes up, the threshold energy may reach
suchvaluesthatthelocaltemperature
reaches
injurious levels at local threshold values of current.
We have found (Lilly et al., 1955a) that if a
120-microsecond rectangular pulse is'quasidifferentiated (to a pulse duration of about 30 ~sec.)
and the resulting pulse pair (Fig. 6.1) is used to
stimulate; no detectable injury occurs over many
weeks of stimulation. Apparently this sliort-term
(within 200 rsec.) reversal of curr'ent, with equal
charge in each of the opposite pulses, reverses
processes leading to "electrolytic" injury.
In some experiments we found that if the pulsepair interval was extended to 16 tLsec. (80/sec.)
and the duration of each condenser-discharge type
pulse duration.
Ofpulse was extended to a time constant of 5 pee.,
the threshold rose slowly in the first few hours of
·stimulation (deep, subcortical electrode). We then
changed to the short pulse pair; the threshold
stayed at the final value for six months. Many
experiments in several loci, at many "dosages,"
demonstrated that "electrolytic" injury was being
generated by this longer pulse pair. In other experiments the duration; of each member of the pulse
pair· was shortened to 10 rsec.; at threshold values
of current the threshold rose. rapidly until breakdown of water began to interrupt the current seen
on an osci~loscope:'I3~epeak energy dissipated per
-pulse ~I·as of the order of 2 watts. The thermal
.injury level was. exceeded by these very much
shortened pulses. Thus for the matched opposite
pulse pairs there are two extremes, limiting the
P
_I
range applicable·to brain stimulationwithoutin-
jury: the lower,low current,low energy,high
·quantity pulse pair causing electrolytic injury; the
upper, high current, high energy, minimum quan-
I
0.01
\
I
QI PULS~.O
DURP;TIONl,Omsec.
FIG.6.3. Peakdissipatedenergyversuspulseduration
at threshold. The data are calculated from Fro. 6.2
in this and subsequent figures. For this electrode, wave-
form,trainduration,andtrain repetition
rate, the
tity pulse pair causing thermal injury. .
In other experiments we found that even in the
presence of a 100 per cent rise in threshold resulting
from electrolytic injury (maintained for six ·weeks),
the damage was not detectable under the micro-
scope;onlyverymuchlargerincreases
in threshold
were correlated with visible damage.
EXCITATION
In general,the:thresholdunderdiscussion
here
thermalinjurythresholdis appro~mately
1 wattpeak
is for either demonstrableevokedmovementor
powerat aboutO.Oe-msec.
pulseduration,
some change in what the mqnkey did in resppnse
63
INTURYAND EXCITATION BY ELECTRIC CURRENTS
UNIDIREGTIONAL
RECTANGULAR
PULSES
~ too
Hess
(1954c)
used
a
very
long
duration
'"balanced" waveform, in order to excite all elements
PULSE
AEPETITIOH
FREPUENCY:
60ppl~sc.
w
TRblN WRP~TION: 5 eer every So ~sc.
13
near the electrodein equal amounts.We do not
la~ow how long he stimulated in each session
(hours) nor for how many days or weeks he ob-
~ lo-ll
a
5
1
served thresholds. If these times are comparable to
I
ours, and the thresholds remained constant, the
)
factsin the two casesdisagree,and furtherresearch
DURATION.
msec.
~I·01 ~a
gO
Elecf rolytic
is needed.
6""'
ddd
.bOl
ful. To obtain an observableeffect on behavior
injury
I
I ··I
Theshorter
balanced
pulses
maypossibly
stimu-
late only short chronaxieelements,but this is doubt
requires firing several thousands of elements near
Thermalinjury
theelectrode.
Thelocal~eldstrengths
nearthe
lo
electrode at the currents observed are presumably
THRESHOLD
PEAK ENERGY..watts
Fro. 6.5. A relativelyminimal-injuryparametricregion:
quantity versus peak energy at threshold at various
o.l
I.o
very much above threshold for even small fibers.
Apparently the shell within which small, longchronaxie elements fire is smaller than the shell in
pulsedurations.(Dataare combined
fromFIGS.6.3
and6.4.) Bothofthe-injury
thresholds
fortheparame-
whichthe larger,short-chronaxie
elementsrespond.
With the longerpulses (resemblingHess'scase),
ters previouslygiven are shown. The vertical straightline portion of the curve is "rheobase"and the horizontal
one is "constant quantity" on this plot. It is to be
emphasizedthat for maximumsafety the waveformof
the two shells may be more coincident in space.
However, it is yet to be demonstrated that these
differences do not average out in the mass of ex-
FIG.
6.1istobeused.Noregion
fortheunidirectionalcitedelements,
rendering
themundetectable
by
waveform
canbe usedlongerthanthe timesgiven(5
sec,every30seconds)
without
causing
electrolytic
injury. It is alsoto be emphasized
that the waveform
of
behavioralcriteria. A point-to-pointcomparison
shouldbe carriedout,in which,if possible,
injury
is avoided.Theremaybe a differentinjurythresh-
Fro. 6.1 can be shortened or extended to give results
old for each of the different elements, analogous
like those in FIGs. 6.2 to 6.5.
to but not identical
Values of the injury
thresholdsvery closeto the onesof the aboveillustration are foundfor suchshortenedand extendedforms
for very long trains applied continuouslyforhours per
day for severalweeks.Thebidirectional
waveform
in
the properparame~icregionextendsthe permissible
total stimulation time of each session by many hours,
to stimulation.
The ~rst threshold
is definitely -
with the digerences
in excita-
tion thresholds (Lilly, 1950b,1952,1954;Lilly and
Cherry, 1954,1955).
EFFECTS
The criteriafor truly long-termfrequentstimulation were developed in the course of a program
of mapping the monkey's brain for elicitable behavioral, emotional, and motor effects of localized
higher than for the minimumevokablelocal elec-
electricalstimulation.In some systems ("start" sys-
·h·ical activity, and the second can be of a magni-
tems), the animals could be trained to start the
tude comparableto that of the electricalresponse
(Lilly, 1958a,c).
stimulus trains repetitively in their own brains
("start capture," Lilly, 1958c; Olds and Milner,
As can be seen from the illustration, the short
1954); in other systems ("stop" systems),they could
pulsesland pulsepairs).havea constantquantity
be trainedto stop trains startedby the apparatus
of electricityper pulse at threshold; these quanti-
or by the observer ("stop capture," Lilly, 1958·c;
ties are of the order of 0.2 to 0.5 microcoulombs
Delgado et al., 19·5427);in still other systems ("alter-
per pulseat 60 pulses(or pulsepairs) per second
nating"systems),theywouldstarta trainof several
in 5-secondtrains for pial electrodes over cortex,
seconds'duration several times and then stop such
and about 0.05 microcoulombs for some deeper
trains several times (Lilly, 1950a, b).
placements.The long pulseshave, temporarily,a
In the "start"systemsthe sessionswouldlast up
constant current threshold (rheobase of the order
to 12 to 16 hours per day at an average train repeti-
of 0.2 to 1.0milliampere)but suchpulsesdestroy
tion rate of at least 2 trains per secondand peak
tissue,
rates up to 18 trains per second, 7 days per week
64
TECHNIQUES
up to 18 months; each train was about 6 to 12 pulse
pairs at a frequency of 60 pulse pairs per second.
The number of trains delivered per day was from
to those for the "stop" systems for the same reasons.
The sexual systems, at least those subsidiary to
erection in the male, show the "alternating" prop-
80,000 to 260,000. In one spot in one monkey the
erty ex~emely clearly.
total
number
of trains
delivered
over
a 6-month
period was approximately 15 million. The number
of pulse pairs delivered was 180 million, or about
Careful quantitative data concerning the spatial
factors in electrical excitation are conspicuously
10 coulombs(0.06microcoulombper pulse pair).
absent in the literature. Many more data and much
In the "stop" systems, with the crescendo ~ain
technique, trains were started by the apparatus
more analysis are needed for evaluation of the resuits of stimulating closely packed systems like the
every 12 secondsand stopped by the animal at his
hypothalamus, intralaminar nuclei, etc. At best,
own chosen threshold. Because of deleterious effects
on the animal as a whole, these sessions were limited
to 3 hours per day (900 trains) for several weeks
(100,000 trains), or to unbroken 48-hour sessions
(18,000 trains) without sleep. Each train had an
localization of function in the central nervous system suffers at present from the lack of a truly
quantitative theory based on what is done in the
laboratory rather than on some neuroanatomical
conjectures. Obviously, the neuroanatomical ~nd-
average of about 300 pulse pairs at threshold. The
stop system loci thus received up to 30 million pulse
pairs or about 1.8 coulombs for 0.06 microcoulomb
ings are essential; they contain the most quantitative data we have to date. But an understanding
of function requires, in addition, a biophysical
per pulse pair.
In the "alternating" systems the ~gures are similar
quantitation
of the geometry of the relations be-
tween excitation and the ~eld set up by the current.