High-Resolution MR of the Intraparotid Facial Nerve and Parotid DuctTheodora Dailiana, Donald Chakeres, Petra Schmalbrock, Phil Williams, and Anthony AletrasPURPOSE: To describe a high-resolution MR imaging technique that depicts the complex anatomyof the region of the parotid gland, focusing on the intraparotid components of the facial nerve andparotid duct. METHODS: High-resolution T1-weighted images of the parotid gland were acquiredwith a prototype three-dimensional Fourier transform gradient-echo sequence that permits a veryshort echo time (4.2 milliseconds) by using a modified phase-encoded time-reduced acquisitionscheme. The sequences were obtained at 1.5 T with a head and neck coil. Postprocessed multi-planar, curvedandvolumetricimageswereobtained. Themost clinicallyuseful imageswereacquired at parameters of 30/4.2 (TR/TEeff) a flip angle of 30, a field of view of 18 to 20 cm, amatrix of 512 288 or 512 256, an axial plane, 60 images, no gaps, and a section thickness of1.5mm. Eighteenhealthysubjectswereexamined. Thepositionof thefacial nervewithintheparotid gland was determined by identifying the facial nerve in the stylomastoid foramen and thenfollowingitonsequentialsectionsthroughtheparotidgland.Curvedreformationswereusedtoconfirmthe visibility of the nerve. A similar technique was used for the parotid duct. RESULTS: Theimage contrast obtained was similar to that of standard spin-echo T1-weighted images. The parotidgland showed intermediate signal intensity while the fat spaces showed high signal signal intensity.The vessels had variable signal intensity depending on saturation. The cerebrospinal fluid, nerves,muscles, and ducts had lower signal intensity. In all 18 subjects, the facial nerve from the brainstem to the parotid gland, and the parotid duct from the mouth to the hilus of the gland were seenbilaterally. The proximal intraparotid facial nerve to the level of the retromandibular vein was seenin72%ofthesubjectsandthemainintraparotidductswereseenin66%ofthesubjects.CON-CLUSION: High-resolutionMRimagingofferssimultaneousdisplayof most of theimportantstructures in the region of the parotid gland, including the intraparotid duct and facial nerve.Index terms: Salivary glands, anatomy; Salivary glands, magnetic resonanceAJNR Am J Neuroradiol 18:165172, January 1997With the development of microsurgical tech-niquesfortheheadandneckarea, improvedimaging, which allows three-dimensional viewsof small structures, is becoming more importantfor successful diagnosis and treatment. Wepresent a high-resolution imaging technique forexaminingtheregionoftheparotidglandandadjacent structures that can portray the normalanatomy in detail and show the exact relation-ships of structures, which, although traditionallydiscrete in pictures of anatomic specimens,have not been displayed clearly on routine mag-neticresonance(MR)imagingstudies(1, 2).Theregionof theparotidglandischallengingbecause of its complex anatomy and the poten-tial serious surgical complications associatedwithinjuriestothefacial nerve, parotidduct,andvessels. Inthepast, surgical explorationwas necessary to evaluate accurately the rela-tionship of the abnormalities to the structures ofthis area.Our goal was to develop a clinically feasiblehigh-resolution MR imaging technique thatwould routinely portray the anatomy of the pa-rotid region, including the intraparotid facialnerve and salivary ducts (3 to 5).Received June 2, 1995; accepted after revision July 5, 1996.SupportedinpartbyresearchgrantR29DC01646fromtheNationalInstitute for Deafness and Other Communication Disorders, National Insti-tutes of Health, Bethesda, Md.From the Department of Radiology, The Ohio State University Collegeof Medicine, Ohio State University Hospital, S-209 Rhodes Hall, 410 W10thAve, Columbus, OH 43210. Address reprint requests to Donald Chakeres,MD.AJNR 18:165172, Jan 1997 0195-6108/97/18010165 American Society of Neuroradiology165Materials and MethodsHigh-resolution MR images of the parotid gland regionwere acquired with a prototype three-dimensional Fouriertransform(3DFT)gradient-echosequencethat allowsavery short echo time (TE) (Figs 19). In this sequence, TEshorteningwas achievedbyusingamodifiedsteppingscheme for the phase-encoded gradients (both in theplane and along the section-encoding direction) (6). Shortphase-encoded gradients were used for the portion of thedata acquisition that determines the contrast (ie, the cen-tral portion of k-space) and longer gradients were used forthe remaining portions. To optimize the image contrast, 15volunteerswereinitiallystudiedwithvarioustechniques,including gradient-recalled acquisition in the steady state(GRASS)orthespoiledGRASSsequences,withseveraldifferent flip angles (20 to 60) and repetition times (TRs)(33 to 17 milliseconds), and with either prototype bilateralphased-array surface coils (7), a routine head coil (Gen-eral Electric, Milwaukee, Wis), or amovableheadandneck coil (Medical Advances, Wawatosa, Wis). In additiontothesestudies, computersimulationswerecarriedoutusingtheequationsgivenbyMeulenetal (8)toaidtheselectionof optimal parametersandtoprovideamoregeneral pictureof thecontrast behavior of thetwose-quence types at different flip angles and TR. For the com-puter simulations, we used T1/T2 of 2000/1000 for cere-brospinalfluid,650/70forbrain(9),800/35formuscle,and 300/80 for adipose (10).Usingtheoptimal techniqueona1.5-Timagingunit,we studied 18 healthy subjects (23 to 58 years old) whohad no known parotid disease. The optimal technique forroutine clinical imaging was in the axial plane with a mov-able head and neck coil, 3-D GRASS technique with pa-rameters of 30/4.2/1 (TR/effective TE/excitations), a flipangle of 30, a field of view (FOV) of 18 to 20 cm, a matrixof 512 288, a section thickness of 1.5 mm, no gaps, and60sections. Theacquisitiontimewasapproximately9minutes. Multiplanar (orthogonal, oblique, andcurved)and volumetric reconstruction images of the 3-Ddata weremade using the standard reformation software on the im-ager and an image workstation (General Electric Advan-tage Windows) with graphic prescription. The images werecompared with textbook anatomic sections and diagrams.Theextracranial facial nerve(ie, inthestylomastoidforamen)wasseenasalow-signal linear structuresur-rounded by fat (Figs 3 and 4). The most proximal extracra-nial extraparotid segment was seen in all healthy subjects(Fig 3). Other small, low-signal structures were occasion-ally seen near the stylomastoid foramen, but they did notcourse into the gland and were continuous with adjacentvascular structures.The position of the facial nerve within the posterosupe-rior portion of the parotid gland was determined by iden-tifyingthefacial nerveinthestylomastoidforamenandthen following it on sequential sections through the parotidgland(Figs3,4,7,and8).Acurvedreformattedimagealong the nerve was used to follow this identified structure.This reformatted image confirmed the identity and conti-nuity of the facial nerve by showing its characteristiccurved extension from the stylomastoid foramen into theposterosuperior segment of the parotid gland. Occasion-ally, anappropriatebranchingpatternwasseenasthenerve divided into its separate segments.The low-signal-intensity parotidduct surroundedbyhigh-signal-intensity fat was reviewed in a similar fashion.The duct from the mouth to the hilum of the parotid glandFig1. Axial MR image (30/4.2) acquired with a flip angle of30, and FOV of 20 cm, section thickness of 1.5, and a 60-sectionacquisition. The parotid gland appears as a polylobulated, inter-mediate signal structure. The parotid glands posterior margin isdelineated at this level by the high-signal posterior auricular ar-tery. The accessory parotid gland is seen lateral to the massetermuscle, in the course of the parotid duct. The internal and externalcarotid, occipital, and vertebral arteries have high signal intensitycaused by time-of-flight effects.Fig2. Coronal MR image at the level of the external auditorycanal (EAC). The gland has an inverted triangular configuration.Its superior margin is formed by the external auditory canal andtemporal bone. The lateral margin is the pinna and subcutaneoustissues.Medially,theparotidglandbordersthestyloidprocess,internal jugular vein, and posterior belly of the digastric muscle.Other labeled structures include the cochlea, temporalis muscle,superficial temporal vessels, the lateral mass of the first cervicalvertebrae (C1), and the styloid process.166 DAILIANA AJNR: 18, January 1997couldbefollowedonafewaxial images (Fig5). Thecontinuation of the duct into the anteroinferior gland wasfollowed with a curved reconstruction to confirm its posi-tion (Figs 68). The frequency with which the intraparotidfacial nerve and duct were seen was determined.ResultsPredominantlyaxial 3DFTimageswereob-tained(Fig1). Occasionally, coronal (Fig2)and, rarely, sagittal primary data sets were ac-quired. They all provided very good informationabout themorphologyof theexaminedstruc-tures, but the axial images were the most usefulfor reconstructions and routine clinical use.Curved sagittal reconstructions optimallyshowed the course of the facial nerve (Figs 4, 7,and 8). A similar display of the parotid duct anditsramiwasobtainedbyusingcurvedsagittalreconstructions. Volume surface reconstruc-tions could also be used to display the anatomy(Figs 6 and 9). Oblique reconstructions allowedFig3. Noncontiguous axial MR im-agesof theproximal extracranial parotidfacial nerve. AismostsuperiorandDismostinferior. Thefacial nerve(arrow)isseen as a low-signal structure surroundedbyfatinthestylomastoidforamen(A)initsextracranial segment posterior totheparotid gland (B). C shows the facial nerve(low signal just anterior to arrow) just en-tering the posterosuperior segment of theparotidgland. Asmall, low-signal, linearbloodvessel isvisiblejust lateral tothenerve. Note that the vessel does not enterthe gland but courses posteriorly and lat-erallyintheoppositedirectionfromthefacial nerve(C). Dillustratesthetypicalappearance of the facial nerve (arrow) inthe gland, as a low-signal structure extend-ing anteriorly and laterally toward the pos-terior margin of the retromandibular vein.Fig4. Curvedsagittal reconstructionof thedescendingandintraparotid facial nerve is a reformation from multiple axial im-ages (Fig 3). The plan of sections curves along the course of thenerve, producing a sagittal appearance. The mastoid segment isclearly connected to the intraparotid segment. Distal branching ofthefacial nervewasnot commonlyseen, but isvisibleinthisexample.AJNR: 18, January1997 MROFPAROTID 167completevisualizationofthesmallmusclesofthis region.Normal parotid gland tissue showed an inter-mediatesignal intensity(higher thanmuscleandlower thanfat) andacharacteristicfinelobular margin (Figs 13). No capsular or fas-cial division of the deep and superficial portionsof theparotidglandwasseen. Stensensductappearedasalow-signal linearstructuresur-rounded by buccal fat and measuring 2 to 3 mmin diameter (Figs 58). The duct could be seenlateral to the masseter muscle and medial to thezygomaticusmajormuscle,finallypenetratingthe buccinator muscle in a location consistentlyposterior totheanterior facial vein. Theductwasfrequentlyincloseapproximationtothelateralmarginofthemassetermusclejustbe-fore entering the gland. The segment of the ductfrom the oral cavity to the hilum was seen in allcases. The parotid duct was often accompaniedby a small nodular structure of intermediate sig-nal (similartothegland),representinganac-cessory parotid gland (Fig 1).Intraparotid ducts could be seen in 66% of theglands, with a characteristic wide-angle branch-ingpatternof approximately80 (Figs 68)when visible. The intraparotid ducts were seenbilaterally when visible. The largest ducts wereseenatthehilumofthegland,whichisintheanteroinferior segment lateral andanterior tothe retromandibular vein. The intraparotid facialnervecouldalsobeseenin72%of healthysubjects from the posterosuperior aspect of theglandtoapproximatelythelevel of theretro-mandibular vein (Figs 3, 4, 7, and 8). In thosepatientsinwhomtheintraparotidfacial nervecould be seen, the nerve did not branch early.Theregions of visibilityof theintraparotidfacialnerveandtheintraparotidductswereintwodistinctlyseparatesegmentsofthegland,so that confusion between the two did not occur(Fig8). Theintraparotidconfigurationof theducts and nerves was also completely different.Fig5. Axial MR images through Stensens duct.A, The parotid duct (Stensens duct) is seen as it penetrates the buccinator muscle just lateral to the second molar. The teeth in themaxillary alveolar ridge are seen just medial to the buccinator muscle. The anterior facial vein is between the duct and the muscle.B, The course of the zygomaticus major muscle is seen at a slightly more superior level.Fig6. Surface and section plane volume image of the parotidduct and its intraparotid duct rami is a reconstruction of a 120-sectionacquisition.Theheadisviewedfroma45rightlateralperspective. The anterior surface is along the coronal planethrough the orbits, maxillary sinuses, and mouth. The parotid ductcrosses the face to penetrate the buccinator muscle just superiorto the facial vessels. The lateral cut surface is through the mas-seter muscle and parotid gland, showing the internal ductalbranching pattern. The branches divide at almost 90.168 DAILIANA AJNR: 18, January 1997Fig7. Curved sagittal reconstructions along the facial nerve and the parotid duct.A,Axial sectionshowstheplaneofthecurvedreconstructionthatfollowsthecourseofthefacial nerveposteriorly(descendingthrough the posterosuperior segment of the parotid gland) and the parotid duct and intraparotid ducts anteriorly.B, In this curved sagittal reconstruction, the parotid duct is visible from the oral cavity to the intraparotid ducts. Note the characteristicpattern of the parotid ducts, with almost 90 branches. The intraparotid retromandibular vein appears as a vertical low-signal region asa result of saturation. The facial nerve in the temporal bone and parotid is seen in continuity.Fig8. Curved reconstructions andoriginal axial MR images of the intraparotidfacial nerveandparotidduct. All imagesare from a single 3-D acquisition.A, Curved sagittal reconstruction paral-lel to the left partoid duct (pd). The mainintraparotid duct is visible with smallbranches. The dotted line is parallel to thesection plane of B.B, Original axial image, parallel to theduct,showsitsanteriorandinferiorloca-tioninthehilusofthegland.Theparotidduct is seen exiting the gland anteriorly.C, Curved sagittal reconstruction paral-lel to the facial nerve (fn). The dottedwhite line is parallel to the image plane ofD.D, Original axial section shows the hor-izontal portion of the facial nerve (fn) in theposterosuperior segment of the gland. Thefacial nerveandparotidductareseenintwo distinctly different portions of thegland. Theyhave characteristic appear-ances and cannot be confused.AJNR: 18, January1997 MROFPAROTID 169Insomehealthysubjects, theductscouldbeseen but not the facial nerve, and vice versa.Owingtobloodflowandthelocationoftheimagevolume,thearterieshadhighsignalin-tensity when not saturated (Fig 1). There couldbe extensive saturation and low signal intensityofthevessels.Thesignalofthevenousstruc-tures was predominantly suppressed because ofsaturation, except inentryzonesections(Fig2). Although flow-compensating gradients werenot used, no major flow artifacts were observed.A few flow-related artifacts were sometimes ob-served in the carotid arteries, but smaller arter-ies were not affected at all. The inferior internalcarotid and most of the external carotid arteryanditsbranches(posterior auricular, anteriorfacial, internal maxillary, superficial temporal)as well as the internal and external jugular veinanditstributaries(superficial temporal, retro-mandibular, anterior facial) were followedthroughouttheircourse, andtheirrelationshipwith the adjacent structures was displayed.Small arterial branchesnear thestylomastoidforamenwereoccasionallyseen. Thesewerenot confused with the intraparotid facial nerve,because they coursed from anterior to posteriorastheywent posteriorly, whichisoppositetothecourseof thefacial nerve. Thesevesselsalso did not enter the parotid gland.The muscles showed low to intermediate sig-nal intensity(higher thanthefluidandlowerthanthegland) (Figs 1, 2, and5). Obliquereconstructions allowed routine visualization ofthecompletecourseof thesternocleidomas-toid,posteriorbellyofthedigastric,styloglos-sus,stylohyoid,stylopharyngeous,medialandlateral pterygoid, masseter, temporalis, bucci-nator, and zygomaticus major muscles.DiscussionA complete examination of the parotid glandregion has required a combination of sialogra-phy, angiography, and either computed tomog-raphy (CT) or MR imaging. After multiple stud-ies, theradiologist must fusetheinformationobtained from the separate imaging techniques.Even with multiple examinations, many impor-tant structures are not clearly delineated, suchasthepositionof thefacial nerveortherela-tionshipof thevarious structures toonean-other. Because of relatively rapid saturation andslowflowof thesesmall vessels, it hasbeendifficult to follow then over long distances or tocreateroutineMRangiographicimages. Theangiographic information provided by the tech-nique described here is important, since the de-tailed anatomy of the venous and arterial struc-tures is seen along with the relationship of thesestructures to the other parotid components.Sincethedataareina3-Dformat, surfaceand volume reconstructions are possible for di-rect comparisonwiththefindingsat physicalexaminationandfor useinpossiblesurgicalinterventions(Figs6and9). Becauseof theversatility of this type of examination, it may, infact, belessexpensivethanacombinationofother procedures.Several technical factors need to be opti-mized in order to produce high-quality images.The phased-array coils generate the highestsignal-to-noiseratioperipherally, but becauseof the signal drop-off they are difficult to post-process. They are also less versatile for exam-iningdeepstructuresandaremoredifficulttoset up. The movable volume head and neck coilhasasignal-to-noiseadvantageof upto50%compared with the head coil. This difference insignal-to-noiseratioisimportantfordepictingthe many small structures. The neck coil couldbecenteredlower over themandibleandtheface, making it suitable for most studies.Inourstudy, optimal resultswereachievedwith an FOV of 20 cm, a 512 288 matrix, aFig9. Surfaceandsectionvolumeviewofabenignparotidtumor is a volume reconstruction from an axial image acquisition.Theleft sideof thefaceisviewedfromaleft anterior obliqueangle. Arectangular sectionhasbeenremoved, exposingthesagittal and coronal planes. The surface of the pinna and externalauditory canal (EAC) are well seen. This 55-year-old patient hada benign mixed tumor, which is lower in signal intensity than thenormal gland. It is seen as an oval mass just inferior to the pinna,producing a bulge on the lateral skin surface. Other labeled struc-tures include the masseter muscle and the normal parotid gland.170 DAILIANA AJNR: 18, January 19971.5-mm section thickness, and 3-D volume ac-quisition of 60 sections using the volume headand neck coil. This FOV and matrix size corre-sponds to an in-plane pixel resolution of 0.39 0.69mm2.Theapparent(effective)resolutioninthephase-encodingdirectionisfurther in-creased, since the data are interpolatedtoapixel resolutionof 0.390.39mm2(byzerofillingtherawtimedomaindatabeforeapplying the Fourier transform along the phase-encoding direction).We attempted to increase the spatial resolu-tion further by reducing the FOV, increasing thematrix size, or decreasing the section thickness.However, with the head coil or the volume neckcoil, theresultingdecreaseinsignal-to-noiseratio (Table) led to unacceptable loss of imagequalityinallcases.Increasedmatrixsizesnotonly lead to decreased signal-to-noise ratio butalso to prohibitive scan times. Somewhat higherspatial resolution can be achieved in peripheralregions with the phased-array coils, although atthe expense of large variations in signal. Exces-sive scan times are a hindrance for acquisitionof124imagesectionswhenusingaTRof30millisecondswithabandwidthto16kHz. Ac-quisition of 120 image sections is possible,however, withaTRof 17millisecondsanda32-kHz bandwidth, which leads to more realisticscan times.In standard gradient-echo sequences, the TEislimitedbythedurationofthephase-encodegradients for submillimeter spatial resolution.Thus, for the spatial resolution used in thisstudy,standardsequencesrequiredaminimalTE of 6 to 8 milliseconds. At this TE and fieldstrength, thesignal of fatandwaterspinsareout of phase, leading to low signal on gradientechoes in all voxels that contain water and lipidcomponents.Thisisaseriousproblemforthehead and neck region that is not a problem inthebrain, sincetherearenofat-water inter-faces. Thenovel phase-encode, time-reducedacquisition technique used in this study permitsthe acquisition of fat/water in-phase imageswith standard gradients for the center ofk-spacewithout theneedof expensivehard-ware upgrades. Since the center of the k-spacelargely controls the contrast, the image appearsto have an effective short TE. In particular, forimaging in the region of the parotid gland, it iscrucial toavoidlow-signal linearartifactsthatresult in out-of-phase TE (6 to 7 milliseconds)of all voxelscontainingbothfat andwater at1.5 T.For the tissues in the head and neck (muscle,neural tissue, adipose, and cerebrospinal fluid),theGRASSsequenceresultsinhigher signal.This is especially pronounced for very short TR;thatis, whilethesignal decreasesonlyslowlywithTRontheGRASSimages, asignificantsignal reductionoccursinspoiledGRASSim-ages acquired with a very short TR (17 millisec-onds). Furthermore, the contrast between mus-cle, neural tissue, and fat does not changesignificantly between spoiled GRASS andGRASS sequences (the main difference is bettergray-white matter differentiation with theExpected signal-to-noise ratio (SNR) based on spatial resolution (for constant TR/TE of 30/4.2, flip angle of 30, and bandwidth of16 kHz)Field of View, cm Matrix Section Thickness No. of Sections SNR Voxel, mm3Scan Time, min20 512 256 1.5 60 1 0.39 0.76* 1.5 8:2020 512 288 1.5 60 0.94 0.39 0.69* 1.5 9:2018 512 256 1.5 60 0.80 0.35 0.7* 1.5 8:2018 512 288 1.5 60 0.76 0.35 0.63* 1.5 9:2020 512 256 1.5 28 0.71 0.39 0.7* 1.5 4:1020 512 256 0.7 60 0.47 0.39 0.7* 0.7 8:2020 512 256 1.5 1241.41 0.39 0.7* 1.5 16:4020 512 256 1.5 1241*f(TR) 0.39 0.7* 1.5 9:3020 512 512 1.5 60 0.71 0.39 0.35 1.5 16:4020 512 384 1.5 60 0.82 0.39 0.52* 1.5 12:2022 512 512 1.5 60 0.86 0.43 0.43 1.5 16:4022 512 384 1.5 60 0.99 0.43 0.57* 1.5 12:2024 512 512 1.5 60 1.02 0.47 0.47 1.5 16:4024 512 384 1.5 60 1.17 0.47 0.63* 1.5 12:20* Interpolated to square smaller pixel size in plane.Assuming TR 30 ms, bandwidth 16 kHz.Assuming TR 17 ms, bandwidth 32 kHz; f(TR) indicates factor for signal loss for shorter TR.AJNR: 18, January1997 MROFPAROTID 171spoiled GRASS technique). Because of this, andsincetheGRASSsignal is, ingeneral, higher,GRASSimagingispreferablefor imagingtheface and neck.The contrast between the different tissues isfairly independent of TR. Only the overall signalchangeswithTR.Thisisunlikethemorecus-tomaryspin-echosequences,inwhichchang-ing the TR leads to significant contrast variation.Sincethecontrastisnotdramaticallyaffectedby the choice of TR, scan time can be shortenedbyusingashortTR.However,shorterTRwillleadtodecreasedsignal-to-noiseratio. Thus,there is a trade-off between short TR and imageresolution. Shorter TR, yielding a lower signal-to-noise ratio, will give good image quality onlyfor lower spatial resolution (Table). Longer TR,resultinginahigher signal-to-noiseratio, willpermit higher spatial resolution. Signal intensityand contrast in gradient-echo images also var-ieswiththeflipangle. DependingonTR, thehighest signal is obtained with flip angles of 10to 25 on spoiled GRASS images and 20 to 40on GRASS images. The best contrast isachieved with flip angles of 30 to 40 onGRASS images. For flip angles larger than thepeak signal flip angle, the contrast between thedifferent tissuetypes does not changemuchwith the flip angle.Disadvantages of thetechniqueincludeitssusceptibility to motion artifacts (a fact that de-mands cooperation of the patient) and the lackof T2-weightedimages, whichcouldbeveryuseful forexaminingtheparotidduct system.Anopenquestionistheeffectivenessofusingpresaturation pulses to reduce fat signal.As far as the internal architecture of the pa-rotid gland is concerned, there has been somecontroversyabouttheblacklinesseeninitsparenchymaandwhattheyactuallyrepresent(branches of the facial nerve, rami of the parotidduct, fascial planes) (11). The first investigatorsdiscussing the issue considered these to bebranchesofthefacial nerve;in1993,astudybasedessentiallyontheimagingof asingleautopsy specimen after injection of contrastmaterial intheparotidductconcludedthatallthe low-signal lines seen inside the parotidgland represented rami of the parotid duct andthat no branch of the facial nerve was seen (12).Images acquired with this current techniquecommonly, but not always, accuratelydepictthe intraparotid facial nerve, usually to just pos-terior of the retromandibular vein. Furthermore,in most cases, low-signal lines seen in thedeeper portionof theparotidglandcouldbeconnected to the main stem of the facial nervebyusingcurvedreconstructions. Theparotidductwasalsofollowedtoitsintraparotidseg-ment andsomeof itsrami wereshown; theywere confined in the more anterior and inferiorportion of the gland (Figs 7 and 8).AcknowledgmentWe thank Linda Chakeres for editing the manuscript.References1. FrelingNJM,MolenaarWM,VermeyA,etal.Malignantparotidtumors: clinical useof MRimagingandhistologiccorrelation.Radiology 1992;185:6916962. Chauduri R,BinghamJB,CrossmanJE,GleesonMJ.Magneticresonance imaging of the parotid gland using the STIR sequence.Clin Otolaryngol 1992;17:2112173. Romell LJ, Mancuso AA, Larkin LH, Rarey KE, Mahan PE, RossMH. Sectional Anatomyof theHeadandNeckwithCorrelativeDiagnosticImaging.Philadelphia,Pa:Febiger;1994:28,30,38,42, 43, 148, 150, 154, 1584. LillieJH,BauerBA. Sectional Anatomyof theHeadandNeck.New York, NY: Oxford University Press; 1994:351345. Ferner H, MonsenH, ed. Pernkopt Atlasof Topographical andAppliedHumanAnatomy. Philadelphia, Pa: Saunders; 1963;1:113120, 125, 140144, 148, 1496. YingK, SchmalbrockP, Clymer BD. Echotimereductionforsubmillimeterresolutionimagingwitha3Dphaseencodetimereduced acquisition method. Magn Reson Med 1995;33:82877. Schmalbrock P, Pruski J, Sun L, Rao A, Monroe JW. Phased arrayRFcoilsforhighresolutionMRI oftheinnerearandbrainstem.J Comput Assist Tomorg 1995;19:8148. MeulenP, GroenJP, Tinus AMC, BruntinkG. Fast fieldechoimaging: an overview and contrast calculations. Magn Reson Im-aging 1988;6:3553689. SchmalbrockP, ChakeresDW, Hacker V, YingK, Clymer BD.Optimization of submillimeter resolution imaging of the inner ear.J Magn Reson Imaging 1992;3:45145910. Chakeres DW, Schmalbrock P. Fundamentals of Magnetic Reso-nance Imaging. Baltimore, Md: William & Wilkins; 1992: 21222711. McGheeRBJr,ChakeresDW,SchmalbrockP,BroganM,Neg-ulescoJA. Theextracranial facial nerve: highresolutionthreedimensional Fourier transform MR imaging. AJNR Am J Neurora-diol 1993;14:46547212. Thibault F, Halimi P, BelyN, et al. Internal architectureof theparotid gland at MR imaging: facial nerve or ductal system? Ra-diology 1993;188:701704172 DAILIANA AJNR: 18, January 1997