abnormalities ofhorizontal clinical, · ducing isolated lateral rectus weakness spare the abducens...

Journal ofNeurology, Neurosurgery, and Psychiatry 1990;53:194-199

Abnormalities of horizontal gaze. Clinical,oculographic and magnetic resonance imagingfindings. I Abducens palsy

AM Bronstein, Joan Morris, G Du Boulay, M A Gresty, P Rudge

Medical ResearchCouncil Neuro-Otology Unit, Instituteof NeurologyA M BronsteinM A GrestyP RudgeInstitute ofNeurologyG du BoulayComputing andStatistics Unit,Institute of Neurology,National Hospital,Queen Square,LondonJ MorrisCorrespondence to:Dr A M Bronstein, MRCNeuro-Otology Unit,Institute of Neurology,National Hospital, QueenSquare, London, WC1N3GB, United Kingdom.Received 24 May 1989 and inrevised form4 September 1989.Accepted 18 September 1989

AbstractFifty one patients with abnormalities ofhorizontal gaze were studied with mag-netic imaging of the brain (MRI) and eyemovement recordings to identify the lociof lesions responsible for isolated ab-ducens palsy, conjugate gaze palsy anddifferent types of internuclear ophthal-moplegias. The lesions responsible for aparticular disorder were identified byoverlapping enlarged drawings of theindividual scans at comparable brain-stem levels and identifying the areaswhere the abnormal MRI signals inter-sected. A statistical procedure was dev-ised to exclude the possibility that theareas of overlap occurred by chance. Inthis paper, the findings in the group ofpatients with VI nerve palsy are reportedsince the location of their lesions could bepredicted from known anatomy, so vali-dating the procedure. The results wereindependently obtained with the overlap-ping technique and the statistical pro-cedure and showed that the lesions werelocated in a region corresponding to theposterior part of the abducens fasciculus.This confirms that central lesions pro-ducing isolated lateral rectus weaknessspare the abducens nuclei. The agree-ment between the procedures used andearlier clinical and experimental resultssuggest that the method we describe canbe applied to locate the site of lesions onMRI scans in other groups of patientswith more complex gaze disorders.

Most clinical studies relating abnormalities ofhorizontal gaze to CNS topography have beenbased on occasional case reports of brain stemstrokes and tumours in which the lesions werevisualised, either on CT scan or at necropsy.Frequently with these studies there wereinadequate descriptions of the eye movementabnormality, either because of the severity ofthe underlying clinical condition or becausethese data were collected retrospectively and,more importantly, the radiological techniquesused were inadequate for imaging brainstemlesions.'

In this and the following paper we report thefindings in 51 patients with disorders ofhorizontal gaze in whom detailed clinical andoculo-graphic studies were made, togetherwith the abnormalities found on magnetic

resonance imaging (MRI) scans of the brain.The advantages of MRI over other imagingmethods of the brain stem and its value foridentifying small brainstem abnormalities arewell recognised and could be important indemyelinating disease, where eye movementdisorders are frequent but in which the detec-tability of brainstem lesions by CT scanning islow.25 However, the use of MRI to correlateabnormal function to CNS topography islimited by the presence of multiple or "silent"areas of abnormal MRI signal, a problemparticularly important in demyelinating dis-ease. For this reason it became clear that tolocalise lesions responsible for a particular eyemovement disorder, it was necessary to developa method capable of obtaining reliable clinical/MRI topographic correlations.

Horizontal gaze defects can be divided intoconjugate and disconjugate. The formerinclude unilateral or bilateral gaze palsies andthe latter include failure of abduction (VIpalsy) or adduction (usually internuclear oph-thalmoplegia). In this paper we will describethe findings in a group ofpatients with VI palsyof central origin. This group was deliberatelychosen as a starting point because of its relativeclinical and anatomical simplicity enabling usto illustrate the technical procedure used andits statistical validation. The findings in themore complex types of gaze defects will bepresented in the companion paper.

Material and methodsFor convenience much of this section is com-mon to the two papers.

(A) Patients The 51 subjects of the studywere classified on the basis of examination oftheir eye movement signs into VI nerve palsy,gaze palsy and internuclear ophthalmoplegia.If a patient presented a combination of dis-orders, they were classified according to themore prominent, for instance, a patient with aconvergent paralytic squint due to a severe VInerve palsy associated with some slight slow-ness of saccades in the yoked adductor will beclassified as a VI nerve palsy and not as aunilateral gaze palsy.There were seven patients (aged 18-39 years)

with unilateral VI nerve palsy thought to be ofcentral origin because of the presence of con-current long tract or brainstem signs. Horizon-tal diplopia was a prominent symptom in alland each had limitation of abduction, varying

194

on March 27, 2021 by guest. P

rotected by copyright.http://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp.53.3.194 on 1 M

arch 1990. Dow

nloaded from

http://jnnp.bmj.com/

Abnormalities of horizontal gaze. I Abducens palsy

from a few degrees to complete absence. Fivepatients had MS or a single brainstem episodethat was probably demyelinating. One patienthad a haematoma (patient 3) as an unexpectedMRI finding and another had a pontine glioma(patient 1) (See fig 2A). An evaluation of thesignificance of MRI findings for clinical diag-nosis in some of these patients has beenreported elsewhere.26

(B) Ocular-motor assessment All patients hada full neurological and neuro-otological clinicalinvestigation. Eye movements were recordedwith bilateral monocular horizontal DC EOGon paper by means of an ink-jet polygraph(Mingograph). Monocular calibrations wereobtained while the other eye was patched. Inmost patients vertical recordings were alsomade of the more mobile eye.The targets for saccades consisted of a row of

red LEDs subtending 100 intervals at a distanceof 50 cm from the nasum. Saccades were madefrom the centre LED to 300 left and right,firstly 10 stimuli in one direction and then inthe opposite; 200 or 10° were used if eyeexcursion was limited. The stimuli for eyemovement were switching between the centreand the eccentric light. Head movement wasrestricted with a chin rest. The interval be-tween stimuli was approximately one secondbut the main criteria for delivering a newstimulus was that the eyes had attained thetarget, as monitored on the paper trace. Handmeasurements of peak saccadic velocity weremade at a paper speed of 150 mm/s and resolu-tion of I mm = 1°.

In addition, the following eye movementresponses were also recorded: vestibulo-ocularreflex was tested in the dark with the subjectsseated on a motorised Barany chair usinghorizontal angular velocity steps of + /- 40°/sand sinusoidal rotational stimuli with frequen-cies of 0 1 to 0-4 Hz and peak velocity of 40°/s.Vestibulo-ocular reflex suppression was inves-tigated during sinusoidal stimuli by having thesubjects fixate an LED attached to and rotatingwith the chair, placed 50 cm from the nasum inline with primary gaze. Smooth pursuit waselicited with a laser target moving sinusoidallybetween 0 1 to 0 4 Hz, 20° peak, and optokin-etic nystagmus was evoked by a full field drumencircling the patient rotating at + / - 40°/s.Patients were examined for spontaneous nys-tagamus in the light and dark in primary and onright and left gaze. A caloric test was performedaccording to the technique of Fitzgerald andHallpike' with a modification suggested toinvestigate vestibulo-ocular reflex suppres-sion.'

(C) Identification of lesions A 0-5 TeslaPicker superconducting system was used for allthe MRI examinations in this study. Duringthe, period when observations were beingcollected several upgrades in radio frequencytransmitter, frequency synthesiser and gradi-ent and surface receiver coils resulted in pro-gressive improvements in image quality.

Brain images were all obtained with thestandard head receiver coil supplied by the

manufacturer. Pictures from multislice imag-ing protocols in the transverse plane wereavailable for all patients. The position of thehead in the scanner typically maintained scanplane orientation at about 5-10° positive to theorbito-meatal line. Eight or 16 transverse10 mm thick slices with a 2 mm intersliceseparation were generally made, but towardsthe end of the study, software improvementsallowed 5 mm contiguous slices to be made. Ina few patients sagittal images were also avail-able.An imaging matrix of 128 by 256 was routin-

ely employed, giving pixel dimensions ofapproximately 2 4 mm by 1 2 mm. Each line ofdata collection was subject to two averages.

Spin-echo (SE) and inversion recovery (IR)sequences were used. The nomenclature for SEand IR scans are SE TR/TE and IR TR/TI,the inversion time TI, and the echo time TE,and are defined in the glossary ofNMR termsof the American College of Radiologists.9 Onthe SE sequences used, lesions appear as areasof increased signal (white) relative to normalbrain. Most patients had SE 2000/60 sequen-ces, some early examinations varied from this,usually in the echo time (SE 2000/40). With IRsequences, usually IR 1500/500, but sometimesIR 2000/500, lesions appear as low signal(black).The images analysed were always those

obtained in close temporal relationship withthe oculographic examination. One observer(GduB) was unaware of the ocular-motorstudies and identified and traced all the areas ofabnormal signal. Doubtful areas were com-pared for confirmation in all the sequences(including IR) obtained from the patient on thesame day and sought in neighbouring slices. Athin, smooth periventricular border of highsignal in SE images, when of even thickness,has not been accepted as good evidence of thepresence of an abnormality. Small areas ofabnormal appearance seen in single slices andin only one sequence had to be very clearlydelineated and far from possible confusion withCSF signal to have been accepted for inclusion.

Selective cuts of the scans were enlarged,normalised in size and drawn onto transparentacetate sheets using a projector. The transverseslice where the VIII nerves, internal auditorymeatus and the labyrinth were visualised wasidentified in all patients. Due to the ascendingcourse of the VIII nerves this.cut represents alow pontine level, and for patients with VIpalsy was the only cut drawn. In cases of gazepalsy, a cut in the region ofthe upper halfofthepons was included and in cases of INO a thirdcut in the mesencephalon, in addition to thepontine slices, was drawn. The scans ofpatients with unilateral clinical signs wereoriented as if they were right sided. Theacetates were traced on a representative draw-ing of the brain stem at the appropriate levelwhich thus hosted all the areas of abnormalsignal from the different patients; for thisprocedure the IV ventricle, in particular itsventral border, the VIII nerves and internalauditory meatus and the brainstem-cerebellarcontours, in that order, were used as anatomical

195



j.com/

J Neurol N


arch 1990. Dow

nloaded from


Bronstein, Morris, Du Boulay, Gresty, Rudge

reference points.In this composite picture the areas where

abnormal signals from different patients in thesame clinically defined group overlapped couldbe seen. Of these, the areas where there was an

overlap of the abnormal signals of at least halfof the patients (OL50) were considered likely tobe the sites of the lesions responsible for theoculomotor signs common to the patient group.In the appendix a statistical procedure is des-cribed to test whether the overlaps could haveoccurred by chance.The structures likely to be involved by

lesions were identified by overlaying theacetates and OL50 drawings on to comparablebrainstem sections to the same scale taken fromneuro-anatomy atlases.'° `* This procedurewas done for each group of patients, using theOL50 drawings. In addition, estimates were

made for each individual subject of the extentto which a structure intersected with an area ofincreased MRI signal as drawn on the acetates.This was rated blindly by an observer on a scaleof + to + + + +. The + signs relate both tothe size of the area of abnormal signal with thatof the anatomical structure investigated, pro-viding an estimate of the percentage of area

involved, and to the location of the lesioncentrally or eccentrically with that structure;thus, a small lesion would be + in the centre ofa larger structure and + - if placed eccen-

trically; a lesion about half the size of a struc-ture would be + + if located centrally and + ifeccentric.

ResultsOto-neuro-ophthalmological resultsSaccadic velocities compared with normal datafrom our laboratory'2 are shown in table 1. Asexpected, abduction velocities in the pareticeye were considerably reduced. In cases 1, 5, 6

*Footnote: In the better images the medial and lateral lemniscican be confidently identified and there are often recognisablydiffering regions of signal in IR scans that probably representmedial and lateral vestibular nuclei and VI nerve nuclei as well asthe MLF. The IV nerve nucleus may also possibly be seen. Thisis not true however of poor quality scans and uncertaintyremains about the precise margins of structures. Since the nucleiand tracts of the brainstem cannot be more certainly identifiedthan this by MRI, particularly on SE scans, the structures likelyto be involved by lesions were identified by the indirectprocedure described above.

Table I Saccadic velocities to 30' target steps in patients with VI nerve palsy

Saccade Velocity (°/s)

Patient To right To left

1 RE not recordable* (W) 420LE 320* (SGP) 325*

2 RE 510 510LE 594 290* (W)

3 RE 200* (W) 425LE446 383

4 RE 415 420LE 405 343 (W)

5 RE 363 356* (SGP)LE 500 270* (W)

6 RE 232* (W) 276*LE 373* [SGP) 320*

7 RE 290* (W) 343*LE 355* (SGP) 315*

Normal Controls (n = 25), RE 453 (324-582) 529 (389-669)(mean and 95",, confidence limits) LE 504 (381-627) 470 (340-601)

(W) = direction of action of the clinically weak lateral rectus.* = outside the 95°,0 confidence limits.(SGP) = subclinical gaze palsy, velocity of the adductor yoked to the paretic lateral rectusbelow confidence limits. RE, LE = right eye, left eye.

and 7 there was also slowing of saccades in thedirection of action of other muscles which wasnot evident on the initial clinical examination.Vertical saccadic velocities were normal in allbut case 5 in which they were marginally slow(upwards 260'/s, downwards 300'/s; normallower limit 396'/s and 342'/s respectively,n = 12 subjects age 20-38 years). Six patientshad asymmetrical vestibular responses in theform of directional preponderances of caloricor rotational nystagmus, measured as durationof the response, to the side opposite the lesion.This association between a central VI nervepalsy and a contralateral vestibular directionalpreponderance has not been recognised before.A detailed report and discussion of the ves-tibular findings, however, would not berelevant here and will be made separately.Directionality of other findings were not sys-tematically related to the VI nerve palsy.

Location of the lesions producing VI nerve palsyAll seven patients with VI nerve palsy hadabnormal signals on the MRI at the lowerpontine level and representative individualscans can be seen in fig 1.

Figure 2A shows drawings of the abnormalsignals of the seven cases. By comparing thesewith fig 2B, which illustrates some of thestructures present in this area of the brain, itcan be seen that in six cases the areas ofabnormal signal were anterior, probably affect-ing the intra-brainstem fibres of the VI nerve(abducens fasciculus). In case 6 the abnormalsignal was more posterior, close to the IVventricle, and may have involved the VInucleus to some extent. Table 2 shows theresults of the individual abnormalities of MRIscans. In all patients there was involvement ofthe abducens fasciculus. Three out of the fourpatients with slowness of the adductor yoked tothe paretic VI, constituting a "subclinical gazepalsy", also showed involvement of the VInucleus and/or the nucleus reticularis pontis(NRP) caudalis. Those patients without a sub-clinical gaze palsy did not have VI nucleusinvolvement but one did have slight involve-ment of the NRP caudalis.

Figure 2C shows the area of the pons inwhich abnormal signals from different patientsoverlap. The OL50 (that is, the area withabnormal signal in four or more patients)encroaches upon the abducens fasciculus at thejunction of the posterior and middle thirdportion of the fibres' course within the pons;the central tegmental tract, the facial nervenucleus and the NRP caudalis are also in-volved.

Figure 3 illustrates the final stage of applyingthe statistical procedure described in theappendix. The shaded area in the matrixsuperimposed on the contours of the lowerpons represents the only square where thesignificant level (p < 0 05) was achieved; thatis, the probability that the overlaps in thatsquare have occurred by chance is less than500. The anatomical structures contained inthat square would correspond to the abducensfasciculus and part of the surrounding NRPcaudalis.

196



j.com/

J Neurol N


arch 1990. Dow

nloaded from



Figure 1 Representative transverse SE sequences at the low pontine level of three patients with unilateral abducens palsy. From left to right, patient 3(Right sided VI palsy) with a pontine haematoma and patients 5 (LVI) and 6 (R VI) with acute brain stem episodes probably demyelinating in origin.In this and subsequent figures, the left side of the scan corresponds to right side ofpatients. The arrows point to the abnormal MRI signal.

DiscussionThis study has shown that the lesion respons-

ible for the central VI nerve palsy in our

patients probably lies in the area of the pos-

terior half of the abducens fasciculus. Thisconclusion is based on the results from the

Figure 2 a: Drawings ofthe areas with abnormalsignal on MRI at the lowpontine cut, in the sevenpatients with VI nervepalsy. All lesions weredrawn as if they were rightsided. b: Significantstructures present at thelevel of the low pontinecut.AbdNr: abducens nervefasciculus. ML: mediallemniscus. NRPC: nucleusreticularis pontis caudalis,the lateral and medial partare indicated. MLF:medial longitudinalfasciculus. IV: 4thventricle. NuPp: nucleusprepositus. AbdNu:abducens nucleus. SCP:superior cerebellarpeduncle. VN: vestibularnuclei. ICP: inferiorcerebellar peduncle.FacNu: facial nucleus.FacNr: facial nervefasciculus. CTT: centraltegmental tract. The VIIInerves are also shown nextto the emergence of thefacial nerve.c: Low pontine structuresas in fig 2 (b). Thehatched area is the OL50for the 7 patients with VInerve palsy, that is, thecommon areas withabnormal MRI signal inat least four of thepatients.

0

Pi1:.......... P4 :'& a a P7 :

P2 ----------- P4^ a7

P3 _______o_____ P5 .:. . ^ . . : IV ventr.P3 :°°°°°°°°°o P6 .........-0

identified 0L50 areas and of the independentstatistical procedure described in the appendixwhich are in remarkable agreement (comparefigs 2C and 3).The structure responsible for the generation

of saccades and quick components of nystag-mus has been termed the paramedian pontinereticular formation (PPRF) which, in animals,is located in the nuclei reticularis pontis (NRP)caudalis and oralis.13 Within the PPRF,excitatory "burst" neurons have been identi-fied whose activity is directly related to saccadegeneration. These neurons undergo a suddenincrease of discharge rate ("pulse") which isresponsible for the high velocity of a sac-cade." 15 The pulse is transmitted to theipsilateral abducens nucleus exciting bothabducens motor neurons and abducens inter-neurons.'6 17 The axons of these interneuronsdecussate immediately and ascend in the con-tralateral MLF to innervate the contralateralmedial rectus neurons.'8 19 Thus, lesions of theVI nucleus produce paresis of abduction of theipsilateral eye and of the yoked adduction of thecontralateral eye, that is, an ipsilateral gazepalsy.2122 Thus, it was not unexpected, in thelight of this organisation, that in most of ourcases the lesions with abducens palsy sparedthe VI nucleus. Only one patient (case 6) hadabnormal signal in the area of the VI nucleus;the reason, why this patient did not have a fullgaze palsy are unclear but it may be that themotor neuron in the abducens nucleus is moresensitive to injury than the interneurons.The OL50s extended on to non ocular-motor

neighbouring structures, in particular the facialnerve nucleus and the central tegmental tract.The involvement of the VII is consistent withthe frequent clinical association between VIand VII palsies in brainstem lesions. Theinvolvement of the central tegmental tract maybe at least partly responsible for ataxic featuresin these patients.Comparison of individual lesions showed an

interesting pattern. In those patients withoutassociated subclinical gaze palsy the abducensfasciculus was the structure almost solely

197



j.com/

J Neurol N


arch 1990. Dow

nloaded from


Bronstein, Morris, Du Boulay, Gresty, Rudge

Table 2 Distribution of abnormal MRI signal at the low pontine level in patients witAVI nerve palsy

Sub Clinical NRPC NRPCPatient Gaze Palsy VI Nu VI Fasc Lat Med ML]

1 + + ++++ ++++ ++++ ++2 no - +- - -

3 no - ++++4 no - +++ + + -5 + + +++ + _6 + +++ ++ + + _7 + - -_ -

The + signs under a particular neural structure estimate their degree of intersection witharea of abnormal MRI signal (+ + + + = 1000/; see the end of the material and methosection for a more detailed explanation).

Figure 3 Drawings of theboundaries of the pons atthe levels of the upper andlower pontine cutssuperimposed on the gridsused to apply thestatistical proceduredetailed in the appendix.The hatched area is thegrid square where there aresufficient overlaps of 1abnormal signals in thepatients with VIth palsy toattain significance at the500 level. Compare thisfigure with the OL,0 in fig2 (c). 10

1

Upper pons

Lower pons

1 16

involved whereas in the presence of subclinigaze palsy abnormal signals in the area of tNRP caudalis or the VI nucleus were prese:For example, the abnormal MRI signalpatient 5, who had a subclinical gaze palinvolved the lateral part of the NRP caudalthought to be part of the PPRF. 7 However, 1companion paper shows that cases with cliically evident rather than subclinical unilategaze palsy had lesions involving not only tarea of the lateral NRP caudalis, but areticular structures at higher levels in the pcsuch as the NRP oralis23. These observaticsuggest that the participation of the lateNRP caudalis in saccade generation may notcrucial.We believe that the agreement between 1

findings from OL50s, the statistical analysis athe current knowledge of the anatomical orgaisation of the abducens nerve, validates 1method we have used to localise the areaslesions on MRI scans responsible for a clinisign common to a group of patients. Thisexploited in the following paper23 concernwith more complex gaze disorders for whithe sites of responsible lesions are less westablished.

We are grateful to the Multiple Sclerosis Society of GrBritain and Northern Ireland for generous financial support ato Drs I Ormerod, D Miller, A Kermode and Mr D MacMa,of the NMR Research Group at the National Hospital, QuiSquare for their patient work in the collection oftheMR imag

AppendixMethod to determine if the areas of overlap ofabnormalMRI signalsfrom different patients are

F significant- The following statistical technique has been

developed to test the confidence with which aclinical sign may be related to the site of acommon lesion in a group of patients who havea certain clinical sign in common (for example,VIth nerve palsy) but who, individually, have

an multiple lesions, as it occurs typically in4d8 demyelinating disease. In these circumstances,

the practice is to examine the scans or patho-logical specimens for the anatomical areas oflesion which overlap defining an area which iscommon to the samples from individualpatients. The problem is that overlaps mayoccur by chance and be unrelated to the clinicalsigns the patients have in common. This is ofparticular importance for MRI images inwhich areas of abnormal signal may be clin-ically silent. Accordingly, the method testswhether the number of lesions overlapping in acommon area of brain within a group ofpatients is unlikely to have occurred by chance.

Coding the ResultsThe acetates drawn from the patients' scanswere divided into grids (fig 3). The resolutionof the grid corresponded to the area of thesmallest abnormal signals drawn on the acetate(these in turn were determined by the smallestarea of abnormal signal detectable on the MRIscans). Each square in the grid was coded withO or 1, where: 0 indicates the absence of anabnormal signal; 1 indicates the presence of anabnormal signal. An abnormal signal wasassumed to be present in a square if it covered

cal more than 25% of the square.the

These squares were rearranged into a row.:he The results were formed into a matrix, wheren each row was the results from a differentin patient and each column was a different squaresy, in the grid, that is, the results were organised as'is, follows:thein-ralthelso)nsnsralbe

thendin-theofcalis

Ledichrell

reatandnuseenDes.

square no:patient Ipatient 2patient 3patient 4

X,Y1,l 1,2 1,3 2,1 2,2 2,3 3,10O 0 0 1 1 0 01 1 1 0 0 0 00 0 0 1 0 0 01 1 0 0 0 0 0

NOTATIONLet C = Total number of squares in the gridLet R = Number of patientsLet r = Total number of squares containing

abnormal signals for patient i. i = 1to R

Let Z2 = Total over all possible pairs ofpatients of the number of squareswhere BOTH patients have anabnormal signal.

Let Z3 = Total over all possible triplets ofpatients of the number of squareswhere ALL 3 patients have anabnormal signal.

And similarly for Z4, Z5 etc.

198



j.com/

J Neurol N


arch 1990. Dow

nloaded from



Calculation of resultsAssuming that the abnormal signals in eachpatient are equally likely to lie anywhere on thescan the theoretical means (E(Z2, etc)) andvariances of Z2, Z3 etc were derived24 25:

RE(Z,)= I C_

k= I j<k

fig 3 in the companion paper). This agreementis a particular result of the combination ofnumber of patients and abnormal signals andsize of the abnormal signal and brain area

examined. For instance if each patient hadmore abnormal signals they would be more

likely to overlap. Therefore one would requiremore than 500(5 of the patients to have signalsoverlapping in a certain area for this area to besignificant.

Var(Z2)= R rC +Crr(r-l(rC-1) _ rk= Ij<k CC1

Rrer

E(Z3)=k=I j<k i<j

Var(Z3) = R rr+

k= I j<k i<j

rirjrk(ri- 1)(r l)(rk -1) (rirrk)22--- - C4

Similarly for Z4, Z5 etc

Using Chebyshev's Inequality the probabilityof the observed value of Z2 (or greater) occur-

ring is given by:

Prob (Z2 ) = N) S = Var(Z2)/(N - M)2where m = E(Z2)

If the probability of the observed value of Z2 (orgreater) occurring was greater than 0-05, no

further analysis was done, because it meant thatit was reasonable to assume that the abnormalsignals in each patient were equally likely to lieanywhere on the scan. If, however, thisprobability was less than 0-05 the null hypo-thesis of abnormal signals equally likely to lieanywhere was rejected.

Using Chebyshev's Inequality again, for theprobability of n or more patients having an

abnormal signal in at least 1 square to be lessthan 500, n must be large enough to satisfy:Prob (Zn ) 1) < 005

that is, Prob (IZ. - E(Zn)I > 1 - E(Z.))005

that is, Var(Zn)/(1 - E(Zn))2 - 005

The matrix of results was then examined to see

which, if any, squares had abnormal signals inat least n patients. In this study the level ofsignificance was chosen to be 95 q however a

9900 significance level (or any other value)could have been chosen by determining n sothat the probability of n or more patientshaving an abnormal signal in at least 1 squarewas less than 100°.

CommentsIn practice the areas significant at the 95%level of significance corresponded to areas on

the scans where at least 50% of the patients hadabnormal signals overlapping (OL,O) (fig 3, also

I Henn V, Buttner U. Disorder of Horizontal Gaze. In: GLennerstrand, D S Zee, E L Keller, eds. Functional Basisof Ocular Motility Disorders. Oxford: Pergamon Press,1982:239-45.

2 Ormerod IEC, Bronstein A, Rudge P, et al. Magneticresonance imaging in clinically isolated lesions of the brainstem. J Neurol Neurosurg Psychiatry 1986;49:737-43.

3 Bogousslavsky J, Fox AJ, Carey LS, et al. Correlates of brainstem oculomotor disorders in multiple sclerosis. ArchNeurol 1986;43:460-3.

4 Joseph R, Pullicino P, Goldberg CDS, Clifford Rose F.Bilateral Pontine Gaze Palsy. Arch Neurol 1985;42:93-4.

5 Atlas SW, Grossman RI, Savino PJ, et al. InternuclearOphthalmoplegia: MR-Anatomic Correlation. Am JNeuroradiol 1987;8:243-7.

6 Ormerod IEC, Miller DH, McDonald WI, et al. The role ofNMR imaging in the assessment of multiple sclerosis andisolated neurological lesions. Brain 1987;1 10:1579-616.

7 Fitzgerald G, Hallpike CS. Studies in human vestibularfunction: 1. Observations on directional preponderance(Nystagmusbereitschaft) resulting from cerebral lesions.Brain 1942;65:115-37.

8 Hood JD, Korres S. Vestibular suppression in peripheraland central vestibular disorders. Brain 1979;102:785-804.

9 American College of Radiology, Subcommittee on NMRnomenclature and phantom development. Glossary ofNMR terms. Magnetic Resonance in Medicine 1984;1:414-3.

10 Olszewski J, Baxter D. Cytoarchitecture of the human brainstem. Reinhardt, Basel 1954.

11 Haines DE. Neuroanatomy. An atlas of structures, sectionsand systems. Baltimore, Urban and Schwarzenberg; 1987.

12 Bird AC, Leech J. Internuclear ophthalmoplegia Br JOphthal 1976;60:645-51.

13 Goebel HH, Komatsuzaki A, Bender MB, Cohen B. Lesionsof the pontine tegmentum and conjugate gaze paralysis.Arch Neurol 1971;24:431-40.

14 Cohen B, Henn V. Unit activity in the pontine reticularformation associated with eye movements. Brain Res1972;46:403-10.

15 Fuchs AF, Kaneko CRS, Scudder CA. Brainstem control ofsaccadic eye movements. Ann Rev Neurosci 1985;8:307-37.

16 Highstein SM, Maekawa K, Steinacker A, Cohen B. Synap-tic input from the pontine reticular nuclei to abducensmotoneurons and internuclear neurons in the cat. BrainRes 1976;112:162-7.

17 Buttner-Ennever JA, Henn V. An autoradiographic study ofthe pathways from the pontine reticular formationinvolved in horizontal eye movements. Brain Res 1976;108:155-64.

18 Graybiel AM, Hartwieg EA. Some afferent connections ofthe oculomotor complex in the cat: an experimental studywith tracer techniques. Brain Res 1974;81:543-51.

19 Highstein SM, Baker R. Excitatory termination of abducensinternuclear neurons on medial rectus motoneurons:relationship to syndrome of internuclear opthalmoplegia.J Neurophysiol 1976;41:1647-61.

20 Carpenter MB, McMasters RE, Hanna GR. Disturbances ofconjugate horizontal eye movements in the monkey. ArchNeurol 1983;8:231-47.

21 Meienberg 0, Buttner-Ennever JA, Kraus-Ruppert R.Lesion of the abducens nucleus as the cause of a conjugategaze paralysis. In: Lennerstrand G, Zee DS, Keeler EL,eds. Functional Basis of Ocular Motility Disorders.Oxford: Pergamon Press, 1982;527-9.

22 Pierrot-Deseilligny Ch, Goasguen J. Isolated abducensnucleus damage due to histocytosis X. Brain 1984;107:1019-32.

23 Bronstein AW, Rudge P, Gresty MA, Du Boulay G, MorrisJ. Abnormalities of horizontal gaze. Clinical, oculographicand magnetic resonance imaging findings. II Gaze palsyand internuclear ophthalmoplegia. J Neurol NeurosurgPsychiatry 1990;53:200-7.

24 Eicker PJ, Siddiqui MM, Mielke PW. A matrix occupancyproblem. Annals of Mathematical Statistics 1972;43:988-6.

25 Mantel N. Approaches to a health research occupancyproblem. Biometrics 1974;30:355-62.

199



j.com/

J Neurol N


arch 1990. Dow

nloaded from


abnormalities ofhorizontal clinical, · ducing isolated lateral rectus weakness spare the abducens...

Documents