C

Spocatdoads(eteg

mcT3

Int. J. Radiation Oncology Biol. Phys., Vol. 68, No. 5, pp. 1310–1319, 2007Copyright © 2007 Elsevier Inc.

Printed in the USA. All rights reserved0360-3016/07/$–see front matter

doi:10.1016/j.ijrobp.2007.01.062

LINICAL INVESTIGATION Head and Neck

3D MR SIALOGRAPHY AS A TOOL TO INVESTIGATE RADIATION-INDUCEDXEROSTOMIA: FEASIBILITY STUDY

ELEFTHERIA ASTREINIDOU, PH.D.,* JUDITH M. ROESINK, M.D., PH.D.,* CORNELIS P. J. RAAIJMAKERS, PH.D.,*LAMBERTUS W. BARTELS, PH.D.,† THEO D. WITKAMP, M.D.,† JAN J. W. LAGENDIJK, PH.D.,*

AND CHRIS H. J. TERHAARD, M.D., PH.D.*

*Department of Radiation Oncology, and †Image Sciences Institute, Department of Radiology, University Medical Center Utrecht,Utrecht, The Netherlands

Purpose: To evaluate whether magnetic-resonance (MR) sialography can be used to investigate radiation-induced xerostomia. Preradiotherapy (pre-RT) and postradiotherapy (post-RT) MR sialographic images of themajor salivary ducts (parotid and submandibular) were compared.Methods and Materials: Magnetic-resonance sialography was performed pre-RT, and 6 weeks and 6 monthspost-RT on 9 patients with T1-4N0-2M0 naso- or oropharyngeal tumors, on a 1.5-T MR scanner. Patients werepositioned in the scanner, using a radiotherapy immobilization mask. Image registration of the MR sialogramspre- and post-RT with each other and with the CT and consequently the dose distribution was performed. Acategorical scoring system was used to compare the visibility of ducts pre-RT and post-RT.Results: Good-quality MR sialographic images were obtained, and image registration was successful in all cases.The visibility score of the parotid ducts and submandibular ducts was reduced at 6 weeks post-RT, which meansthat the full trajectory of the salivary ducts, from the intraglandular space to the mouth cavity, was only partiallyvisualized. For some of the parotid ducts, the visibility score improved at 6 months post-RT, but not for thesubmandibular ducts. The mean dose for the parotid glands was 35 Gy (1 standard deviation [SD] 3 Gy), and forthe submandibular glands it was 62 Gy (SD, 8 Gy).Conclusion: Three-dimensional MR sialography is a promising approach for investigating xerostomia, becauseradiation-induced changes to the saliva content of the ducts can be visualized. © 2007 Elsevier Inc.

3D MR sialography, Xerostomia, Salivary glands, IMRT, Head-and-neck radiotherapy.

pnr

wsstsawiasS

in

(

INTRODUCTION

alivary dysfunction is a prevalent side effect suffered byatients receiving head-and-neck radiotherapy (RT). Lackf saliva production results in impaired quality of life, andan be permanent (1–6). The elucidation of the exact mech-nism of radiation damage to the salivary glands has beenhe subject of many studies (7). It was suggested thatamage to the acinar and ductal systems is the main causef salivary-gland dysfunction in humans after RT (8). It waslso suggested that edematous stenosis of the major salivaryucts takes place progressively during irradiation, leading toymptoms of acute sialadenitis with an excretory obstacle9). In these studies, either changes in saliva and serumlectrolyte levels were investigated, or salivary-gland scin-igraphy with 99mTc-pertechnetate was performed. How-ver, neither these nor other studies systematically investi-ated the salivary-duct system in humans after RT. In this

Reprint requests to: Eleftheria Astreinidou, Ph.D., Depart-ent of Radiation Oncology, University Medical Center Utre-

ht, Huispost Q00.118, Heidelberglaan 100, 3584 CX Utrecht,he Netherlands. Tel: (�31) 30-2503035; Fax: (�31)

0-2581226; E-mail: E.Astreinidou@umcutrecht.nl A

1310

ilot study, we aimed to do that, by using a noninvasive,ovel imaging approach, magnetic-resonance (MR) sialog-aphy, for the first time.

Magnetic-resonance sialography is an imaging techniquehereby static fluids or slowly flowing fluids in the body,

uch as saliva, are imaged as high-signal-intensity brighttructures against a dark background (10). In other words,he salivary ducts can be imaged because they containaliva. Saliva itself is the contrast medium. This propertylone makes MR sialography an attractive method to usehen studying hyposalivation, as in cases of radiation-

nduced xerostomia. Magnetic-resonance sialography waslready proven to be a useful imaging technique for otheralivary-gland and duct disorders such as sialolithiasis andjögren’s syndrome (11–15).The conventional method of visualizing the ductal system

s by X-ray sialography. This two-dimensional (2D) tech-ique requires injection of contrast medium and manual

This research was supported by the Dutch Cancer SocietyGrant number UU 2001–2468).

Conflict of interest: none.Received April 13, 2006, and in revised form Jan 8, 2007.

ccepted for publication Jan 30, 2007.

scTasitqd

pil(ipsfihihfKrsc

MvptdftpmM

wvstmpw

(pt(tp4t

fm(

pdwtsBoptrn

tPitMm1hmphqwarndpt

rc

dr

Fs

1311MR sialography post-RT ● E. ASTREINIDOU et al.

kills for cannulation of the ducts, with a risk of failure inase the opening of the ducts cannot be clearly seen (13).hat might very well be the case after RT, and especiallyfter surgery. Moreover, it can only be performed on onealivary duct at the time, and the salivary-gland tissue itselfs not visible. On the other hand, MR sialography is ahree-dimensional (3D) imaging modality, and conse-uently both pairs of parotid ducts and the submandibularucts can be imaged simultaneously (16).Furthermore, 3D MR sialography has the potential to

rovide spatial information on salivary (dys)function. Thisnformation could be beneficial if a correlation is estab-ished between the 3D dose distribution and the salivarydys)function, because it could result in more efficientntensity-modulated radiation therapy (IMRT) treatmentlanning (17–20). Using conventional RT techniques, thealivary glands are mostly included in full in the irradiationelds, and receive homogeneous doses (17). With IMRT,owever, dose painting is possible, and so it is important tonvestigate whether there is regional radiosensitivity in theuman salivary glands and salivary ducts with regard tounction, as observed by Konings et al. in animals (21–23).onings et al. concluded that after partial irradiation of the

at parotid gland, injury to the major excretory ducts andupply routes for blood and nerves in the irradiated lobeould result in late-function loss in the shielded lobe (22).

In a previous study, we developed and evaluated a 3DR sialography protocol in healthy volunteers, that pro-

ided good-quality 3D images of the submandibular andarotid duct systems in a single scan. Moreover, we showedhat MR sialograms of the same individual, recorded atifferent moments in time, are reproducible (16). In thiseasibility study, patients receiving RT for head-and-neckumors were scanned with this 3D MR sialography protocolre-RT, and 6 weeks and 6 months post-RT. Salivary-floweasurements were performed at the same time intervals asR sialography.The primary goal of the present study was to evaluate

hether MR sialography could detect changes in the sali-ary-duct system post-RT compared with the salivary-ductystem pre-RT, and could consequently be used as a newool to investigate radiation-induced xerostomia. Further-ore, the image registration of the MR sialograms pre- and

ost-RT with each other and with the 3D dose distributionas evaluated.

METHODS AND MATERIALS

Nine patients, treated with primary (7 patients) or postoperative2 patients) irradiation for T1-4N0-2M0 nasopharyngeal or oro-haryngeal cancer, were included in the study. Seven patients werereated with IMRT, and two were treated with 3D-conformal RT3D-CRT) (17). In cases of 3D-CRT planning, the prescribed doseo the planning target volume (PTV) of the primary tumor and theositive lymph nodes was 70 Gy, and to the elective lymph nodes,6–50 Gy. The PTV was determined by adding a 5-mm margin to

he clinical target volumes (CTVs). The dose was delivered in 35 r

ractions, and the dose calculation was performed using a com-ercial treatment-planning system, PLATO RTS, version 2.6

Nucletron BV, Veenendaal, The Netherlands).The IMRT planning was performed using the inverse treatment-

lanning module PLATO-ITP, version 1.1 (Nucletron BV). Theose prescription to the tumor PTV and positive lymph node PTVas 66–69 Gy, and to the elective lymph node PTV, 54 Gy. The

otal dose was delivered in 30 fractions. The dose constraint to thepinal cord and brain was 30 Gy, with a relatively high penalty.ecause of the 5-mm PTV margin, parts of the parotid glandsverlapped with the PTVs. At this overlapping volume, the doserescription of the PTV was used in the optimization process. Tohe rest of the parotid glands, a dose constraint of 20 Gy with aelatively low penalty was given. The submandibular glands wereot included in the optimization process.Magnetic-resonance sialography was performed on a 1.5-T sys-

em (Intera, Philips Medical Systems, Best, The Netherlands).atients were positioned in the MR scanner with the five-point RT

mmobilization mask, in the RT treatment position, by the RTechnologist. A two-element circular surface coil (FLEX L, Philips

edical Systems, Best, The Netherlands; opening diameter, 170m) was placed bilaterally on the RT immobilization mask (Fig.

). There was a twofold reason for doing this: (1) the standardead-and-neck coil does not permit the use of the immobilizationask; and (2) more importantly, the contrast-to-noise ratio of the

arotid duct to fat is higher when using the surface coils than theead coil (16). A 3D water-selective turbo spin echo pulse se-uence, with repetition time/echo time (TR/TE) 6,000 ms/190 msas applied, as described in detail in a previous study (16). The

cquired image resolution was 0.8 � 1.4 � 1.5 mm3, and theeconstructed image resolution was 0.4 � 0.4 � 1.5 mm3. Mag-etic-resonance sialography was performed pre-RT, on the sameay as the planning CT, and was repeated 6 weeks and 6 monthsost-RT. The scan time was 8–10 min, depending on the size ofhe scan volume.

Before each MR follow-up session, it was verified in the moldoom whether the RT immobilization mask still fit the patient, andonsequently whether the pre-RT position was reproducible.

The scan volume included the parotid glands and the subman-ibular glands bilaterally, and was slightly tilted (5–8°) around theight–left axis. It also included a small tube filled with 25 mL

ig. 1. Positioning of the patient in the magnetic resonance (MR)canner. The surface coils (FLEX L) are placed bilaterally on the

adiotherapy (RT) immobilization mask.

Miiw

dcsswsp

od23pbddetflpSw

twocat

pavrtpitbvdd

md

gwpduc

ilrmt51

ftmmpawanof(pFa(m6d

V

ggofrotritsaf

Fs

1312 I. J. Radiation Oncology ● Biology ● Physics Volume 68, Number 5, 2007

nCl2 · 4H2O solution, 19.2 mL/L, which was taped onto themmobilization mask at the level of the parotid gland. The signalntensity of that solution was used for normalization purposeshen comparing MR sialograms.The analysis of 3D data sets was performed with in-house-

eveloped software that allows visualization of the transversal,oronal, and sagittal views of the data set together. Thus, anyelected point of the data set was viewed in 3D. The trajectory ofubmandibular (Wharton’s) ducts and parotid (Stensen’s) ductsere mainly visible on the transversal planes, and the higher-order,

mall parotid-duct branches were more visible on the sagittallanes (16).

A categorical scoring system was used to evaluate the visibilityf the salivary ducts (Fig. 2). In brief, the parotid ducts wereivided into four parts: two intraglandular (part 1, main duct; part, higher-order, small-duct branches), and two extraglandular (part, segment of Stensen’s duct superficially to the masseter muscle;art 4, last segment of Stensen’s duct that turns 90° to pierce theuccinator muscle before opening to the oral cavity). The subman-ibular ducts were divided into two parts (part 1, intraglandularuct and the posterior bend of Wharton’s duct around the posteriordge of the mylohyoid muscle; part 2, the extraglandular segmenthat runs through the sublingual space and opens to the anterioroor of the mouth near the lingual frenula). The score was 1 if theart was visible, and 0 if it was not. If the whole trajectory oftensen’s duct (including the small, intraglandular duct branches)as visible, the total score was 4 for the parotid ducts.For the submandibular duct, the total score was 2 if the whole

rajectory of Wharton’s duct and the intraglandular part of the ductere visible. Visibility was scored for each MR sialogram by twobservers, and a consensus was reached. This score was used toompare the visibility of ducts post-RT to their visibility pre-RT,nd to quantify indirectly whether all or only part of the ductrajectory was detectable.

The MR sialograms 6 weeks and 6 months post-RT, and thelanning CT, were registered to the MR sialogram pre-RT, usingmutual information algorithm. The registration was inspected

isually using a “linked cursor” approach. Two windows with theegistered data sets were simultaneously opened. When the loca-ion of a point in one data set was selected, a visible mark wasositioned automatically at the corresponding location in the reg-stered data set (24). The registration was considered good whenhere was a successful match of the body contour, base of the skull,rain tissue, and other recognizable structures, such as bloodessels. Afterwards, the matching of the location of the salivaryucts on the registered MR sialograms was checked, and the 3D

ig. 2. Schematic overview of the salivary-duct visibility scoringystem, from Astreinidou et al. (16).

ose distribution was superimposed on the sialogram pre-RT. The s

ean dose to the salivary glands and the mean dose to the salivary-uct parts were also obtained.The diameters of Stensen’s and Wharton’s ducts upon entry to the

lands were recorded. The diameter of the duct was defined as the fullidth of half-maximum (FWHM) of a Gaussian fit of a profile drawnerpendicular to the salivary ducts (at the end of the intraglandulaructal part) on a transversal plane of the data sets. A paired t-test wassed to compare diameters before and after RT. Differences wereonsidered statistically significant when p � 0.05.

Salivary-flow measurements were performed at the same timentervals and often on the same day, immediately after MR sia-ography. Stimulated parotid saliva was collected from both pa-otid glands simultaneously with Lashley cups, and mixed sub-andibular and sublingual saliva was collected from the floor of

he mouth using a pipette. Stimulation was achieved by applying0 �L of a 5% acid solution to the mobile part of the tongue everymin, and saliva collection was carried out for 10 min (25).

RESULTS

Magnetic-resonance sialography was performed success-ully, and was well-accepted by all 9 patients. The reposi-ioning of the RT immobilization mask at 6 weeks and 6onths post-RT was still good, and there was no need toake any new masks. The 3D image registration of the

re-RT MR sialograms with the post-RT MR sialogramsnd the CT data, and consequently the 3D dose distribution,as inspected visually, and it was considered satisfactory in

ll cases. The body contours were in good agreement, ando differences �2–3 voxels were observed when checkingn specific points. The average diameter of the parotid ductsor this group of patients was 2.1 mm (standard deviationSD), 0.5 mm) pre-RT, 2.2 mm (SD, 0.8 mm) at 6 weeksost-RT, and 2.3 mm (SD, 0.5 mm) at 6 months post-RT.or the submandibular ducts, the average diameters pre-RT,t 6 weeks post-RT, and 6 months post-RT, were 2.2 mmSD, 0.5 mm), 2.0 mm (SD, 0.3 mm), and 2.1 mm (SD, 0.4m), respectively. The diameters as measured 6 weeks andmonths post-RT were not statistically different from the

iameters measured before RT (p � 0.05 in all cases).

isibility of the salivary ductsWe assessed the visibility of the salivary ducts of 18 parotid

lands and 16 submandibular glands. Two submandibularlands were excluded because they were in the vicinity of theperated area. One patient was lost from the 6-month post-RTollow-up. In general, the trajectory of the extraglandular pa-otid duct (Stensen’s duct) was visible on the transverse planesf the sialograms. The small parotid-duct branches were de-ectable as small, hyperintense dots in the hypointense sur-ounding of the parotid tissue on the transverse planes. Scroll-ng through the sagittal planes at the level of the parotid glands,he full trajectory of the small ducts as they converged from theecond-order branches to the first-order branches, and eventu-lly to the main intraglandular parotid duct, could be nicelyollowed (Figs. 3a, 4a, 5).

From the 300 salivary-duct parts (both parotid and

ubmandibular) that were evaluated, there was discussion

fag

mfmafedatw

rltw

wdrsmm

vppm1iv7tti

1313MR sialography post-RT ● E. ASTREINIDOU et al.

or only 11 (approximately 3%). Consensus was reachedfter comparison of the pre- and post-RT MR sialo-rams.The visibility score of the parotid ducts was the maxi-um of 4 for all parotid glands, and was constant during the

ollow-up scans for those parotid glands that received aean dose of �20 Gy. The visibility score was reduced for

ll other glands at 6 weeks post-RT, and started to improveor some patients at 6 months post-RT (Fig. 6). An inter-sting observation was that the intraglandular main parotiduct (part 1) was always visible, pre-RT, and post-RT. Theverage dose to that part was 33 Gy (SD, 13 Gy), similar tohe average mean dose to the whole parotid gland, whichas 35 Gy (SD, 13 Gy).The changes in visibility score post-RT were due to the

educed visibility of the other ductal parts, the extraglandu-ar and intraglandular small parotid ducts. The continuity ofhe smaller duct branches to the larger ones was lost. There

Fig. 3. A sagittal plane of the three-dimensional magnethrough the left parotid gland of a patient in whom the msmall duct branches were clearly visible (a) before radiois superimposed on (c) the same sagittal plane and (d) aof the parotid gland. Arrows indicate exactly the same pduct at that location was 16 Gy.

as no indication of regional damage to the small ducts p

ithin the gland. The dose to parts 3 and 4 of the parotiduct was 25 Gy (SD, 17 Gy) and 33 Gy (SD, 22 Gy),espectively. The dose to part 3 was lower because it isituated superficially to the masseter muscle, and that part isostly spared by IMRT, while part 4 is situated moreedially and is closer to the high-dose area.The full trajectory of the submandibular ducts was

isible (maximum score, 2) for all submandibular glandsre-RT. The visibility score remained unchanged in oneatient at post-RT measurements. In that patient, theean dose to the left and right submandibular glands was

5 Gy and 2 Gy, respectively. For all other patients, onlyntraglandular part 1 was visible at 6 weeks post-RT. Theisibility score did not improve at 6 months post-RT (Fig.), as it did for the parotid ducts. The average mean doseo the submandibular glands and ducts was higher thanhat to the parotid glands, at 62 Gy (SD, 8 Gy). They werencluded in the high-dose volume, because no treatment-

nance (3D MR) Sialogram (TR/TE 6,000 ms/190 ms)traglandular parotid main duct (part 1) and a part of they (RT) and (b) 6 weeks post-RT. The dose distributionrse plane. The blue outline in (c) and (d) is the contourof the parotid duct in (a–d). The dose received by the

tic resoain intheraptransveosition

lanning effort was made to spare them.

a(

S

twfltst

g0av

ot

scop0a

whu

1314 I. J. Radiation Oncology ● Biology ● Physics Volume 68, Number 5, 2007

The signal intensity of the whole salivary-gland tissueppeared to be higher post-RT compared to that pre-RTFigs. 6, 7).

aliva measurementsSalivary-flow measurements were successful in all 9 pa-

ients pre-RT and in 8 patients post-RT. One patient, whoas excluded from the analysis, experienced pain on theoor of the mouth during stimulation with the citric acid at

he post-RT follow-up, and the measurement had to betopped. The MR sialography, however, was performed inhat case.

The average amount of saliva collected from the parotidlands in 10 min was 1.7 mL (median, 1.2 mL) pre-RT, and.3 mL (median, 0 mL) and 0.3 mL (median, 0 mL) 6 weeksnd 6 months post-RT, respectively. Despite the fact that the

Fig. 4. A sagittal plane of the three-dimensional magnethrough the right parotid gland of a patient in whom theduct branches were clearly visible before radiotherapy (superimposed (c) on the same sagittal plane, and (d) on aof the parotid gland. Arrows indicate exactly the same pduct at that location was 62 Gy.

isibility score of some parotid ducts improved to maximum o

f 4, at 6 months post-RT, no saliva was collected fromhem.

Saliva from the submandibular glands could not be mea-ured separately as from the parotid glands. The salivaollected from the floor of the mouth was assigned to bothf the submandibular glands. The average amount of salivare-RT was 3.4 mL (median, 3.6 mL), and was reduced to.2 mL (median, 0 mL) and 0.1 mL (median, 0 mL) 6 weeksnd 6 months post-RT, respectively.

DISCUSSION

In this prospective study, a 3D MR sialography protocolas applied, for the first time, to patients receiving RT foread-and-neck tumors. The MR sialography was performedsing a heavily T2-weighted sequence that allows imaging

onance (3D MR) sialogram (TR/TE 6,000 ms/190 ms)intraglandular parotid main duct (part 1) and the small

), but not 6 weeks post-RT (b). The dose distribution iserse plane. The blue outline in (c) and (d) is the contourof the parotid duct in (a–d). The dose received by the

tic resmain

RT) (atransvosition

f the salivary ducts because the saliva appears hyperin-

tTtias

ptwvsvvs

sip

as

sptlTedsmn(issc

1315MR sialography post-RT ● E. ASTREINIDOU et al.

ense, and the surrounding tissue apears hypointense (10).he use of the same RT immobilization mask, every time

he patient was scanned, resulted in successful image reg-stration of the MR sialograms pre-RT with those post-RT,nd of the MR sialograms with the planning CT and con-equently with the dose distribution.

The comparison of MR sialographic images of patientsre-RT and post-RT revealed radiation-induced changes inhe visibility of the salivary ducts. The visibility of thehole trajectory of the salivary ducts as described by aisibility scoring system was reduced post-RT in compari-on to that pre-RT. In healthy volunteers, however, theisibility score was reproducible in long-term time inter-als. Thus it can be concluded that the reduced visibilitycore in patients indicated a lack of saliva (hyposalivation).

Mainly, the extraglandular parts of both the parotid andubmandibular ducts were not visible post-RT, while thentraglandular main duct was always visible, pre-and

Fig. 5. An oblique transverse plane of the three-dimensms/190 ms) of a patient, at a level through the base of mducts before radiotherapy (RT), 6 weeks post-RT, and 6tube containing the MnCl2 · 4H2O solution that was useat left are expanded at right. Arrows indicate the positiowas not visible in this patient.

ost-RT (Figs. 3–5, 7). The diameters of ducts, as measured w

t the end of the intraglandular main duct (part 1), wereimilar to those observed in healthy volunteers (16).

Furthermore, the diameters of ducts post-RT were notignificantly different from those pre-RT, for this group ofatients. This finding suggests that there was no dilatation ofhe ducts due to obstruction that could possibly explain theack of saliva in the extraglandular parts post-RT (9, 12).his is in accordance with what was observed by Kashimat al. (26). In that study, a less detailed pattern of the parotiduct, without small-duct filling, was observed using X-rayialography 20 weeks post-RT. Similar observations wereade with MR sialography. The small-duct branches could

ot be visualized for some of the parotid glands post-RTFig. 4). Whether that was because of the higher signalntensity of the parotid tissue itself post-RT, or because themall ducts were damaged (22), cannot be concluded. Iteemed that the signal intensity was homogeneously in-reased in the whole salivary-gland tissue, even if the dose

magnetic resonance (3D MR) sialogram (TR/TE 6,000that includes a segment of parts 3 and 4 of the parotidpost-RT. The hyperintense circular area is a part of the

ignal normalization. The areas within the white framesrt 4 of the parotid ducts. At 6 weeks post-RT, that part

ionalaxilla

monthsd for sn of pa

as not homogeneously distributed.

o(kt(fignftTaisswTfi

atqspstoss

bwtap

pd6tmcdt

wstcisssreiMptt

scw

tdwcwbn

weeks

1316 I. J. Radiation Oncology ● Biology ● Physics Volume 68, Number 5, 2007

The signal intensity of the salivary glands and, in general,f the rest of the tissues appeared to be higher post-RTFigs. 5, 7). This was likely due to edema. Edema is anown radiation-induced tissue reaction, and edematousissue appears hyperintense on T2-weighted MR imaging27). Furthermore, it is possible to follow how, when, andor which tissues these effects decrease in time, by repetitivemaging. This increases the value of the information, andives a four-dimensional character to MR sialography. Theext step would be to investigate the possibility of quanti-ying this information and, if possible, establish a correla-ion between tissue changes and spatial dose distribution.he quantitative comparison between pre-RT MR imagesnd those post-RT is not straightforward. The difficulty liesn the fact that the intensity of the signal in an MR image isensitive to the system receiver and image-reconstructionettings. Therefore, a relative comparison with the help of aater solution or a tissue-like object, with a constant in time

1 and T2 properties, could be an option. This option shouldrst be validated in healthy volunteers.The visibility score of the salivary ducts was produced bysubjective method, but it nevertheless reflects the radia-

ion-induced changes to the salivary system and quantifiesualitative information. In addition, it indirectly providesome functional information. In brief, the first step of salivaroduction takes place in the acinar cells, and the secondtep takes place in the ductal cells. From there, saliva isransferred to the excretory ducts, and then drains into theral cavity. Therefore, damage to one or both of the abovealiva-production steps is reflected in the reduction of thealivary-duct visibility score.

Far more data will be needed to support any correlationetween dose and visibility score. However, the trend thate observed in our data was for a relatively low mean dose

o the salivary glands, which in this group of patients wasbout 20 Gy, no changes in visibility score were observed

Fig. 6. Duct visibility score vs. parotid mean dose at 6

ost-RT. Above 20 Gy, however, a lower score than the g

re-RT score was recorded, which means that the salivary-uct trajectory was partially visible. The increased score atmonths post-RT compared to that at 6 weeks post-RT for

he parotid ducts of some patients suggests that a repairechanism may have started. However, the same cannot be

oncluded for the submandibular glands and ducts. Theose received by the submandibular glands was, on average,wo times higher than that of the parotid glands.

We must point out that the MR sialography in this studyas performed without the use of any contrast medium or

timulation. Consequently, it imaged the salivary-duct sys-em at rest. About 70% of the salivary production at restomes from the submandibular, sublingual, and minor sal-vary glands spread out in the oral cavity. During gustatorytimulation, while drinking and eating, about 60% of thealiva is produced by the parotid glands (9). Lack of un-timulated salivary flow, however, might play an importantole in the subjective feeling of xerostomia that patientsxperience after radiotherapy. Hence, future studies shouldnvestigate whether qualitative and quantitative findings from

R sialographic imaging correlate with the perception of aatient’s dry mouth, as derived from patient-completed quali-y-of-life assessments. Patient self-reported scores should behe main endpoint when evaluating xerostomia (28).

Nevertheless, the standard method of quantifying xero-tomia is with salivary-flow measurements using Lashleyups (25), which is why we used this method in our study asell.Saliva was collected post-RT from those parotid glands

hat received a mean dose �20 Gy, and the full salivary-uct trajectory was still visible in the MR sialograms. Thereere parotid glands, however, from which no saliva was

ollected at 6 months post-RT, despite the fact that thehole duct trajectory could be visualized again. This mighte because the saliva composition was not normal and couldot flow as it did pre-RT, or because the lag phase of those

post-RT (a) and 6 months post-radiotherapy (RT) (b).

lands at that stage was longer than the 10-min measure-

ma(

sdTwoamd

msrlsop

smopspfteAistdmesnR

1317MR sialography post-RT ● E. ASTREINIDOU et al.

ent time. It should also be taken into account that there isvariation in the salivary measurement itself of about 27%

1 SD) in healthy volunteers (29).Based on salivary-flow measurement and the visibility of

ubmandibular ducts, the salivary function of the subman-ibular glands decreased even more at 6 months post-RT.he amount of saliva collected from the floor of the mouthas very little, and that could explain why the full trajectoryf the submandibular ducts was no longer visible. There islso a certain detection and resolution limit in all imagingodalities. We should be aware of this when evaluating

ata and when the salivary ducts are not visible.However, to establish a correlation between salivary-floweasurements (10 min of measurement time) and 3D MR

ialography (8–10 min of scan time), more patient data areequired, and most importantly, the acquisition of MR sia-ographic images should replicate the salivary-flow mea-urements. Preliminary results from a pilot study performedn healthy volunteers at our institute showed that it was

Fig. 7. An oblique transverse plane of the three-dimensionams) of a patient, at a level through the transverse processsubmandibular ducts, before radiotherapy (RT), 6 weeks poat left are expanded at right. Arrows indicate submandibusubmandibular glands; P � parotid glands.

ossible to acquire sialographic images during and after t

alivary stimulation with 50 �L of a 5% acid solution to theobile part of the tongue every 1 min. The signal intensity

f the parotid glands and duct volumes increased by ap-roximately 40% during stimulation, relative to the un-timulated scan. After stimulation, the signal intensity in thearotid ducts returned to baseline, but this was not the caseor the parotid glands (30). The application of such a pro-ocol to patients should be further investigated. Morimotot al. (31) presented qualitatively similar results to those ofstreinidou et al. (30). They acquired 2D MR sialography

mages of a single parotid gland and duct, before and afteringle-shot stimulation with citric acid. They observed aime-dependant signal-intensity change of the main parotiduct that seemed to correlate with the total saliva volume,easured with the spitting method. With this method, how-

ver, the flow from each parotid gland is not measuredeparately, as with the Lashley cups method. That would beecessary when investigating parotid-gland function afterT. Nevertheless, it is clear from the study of Morimoto’s,

etic resonance (3D MR) sialogram (TR/TE 6,000 ms/190atlas (C1) and the mandible that includes a part of the

and 6 months post-RT. The areas within the white framescts. Submandibular ducts were not visible post-RT. S �

l magnof the

st-RT,lar du

hat MR sialography allows functional evaluation of the

slsas

h(ddMpmflafa

iwpt

itscdtin

1

1

1

1318 I. J. Radiation Oncology ● Biology ● Physics Volume 68, Number 5, 2007

alivary glands, and merits further investigation. Particu-arly, 3D MR sialography, in combination with salivarytimulation, has the potential to provide spatial informationbout the salivary-gland function of both the parotid andubmandibular glands bilaterally and simultaneously.

In recent years, diffusion-weighted MR imaging (DW-MRI)as also been employed to investigate salivary-gland function32–34). Diffusion-weighted MRI produces image contrastue to the random movement of water in tissues. The apparentiffusion coefficient (ADC) is a parameter that quantifies DW-RI. The interpretation of ADC values, however, should be

erformed with care, because they are influenced not only byotion due to diffusion, but also due to perfusion and salivaryow (35). Irrespective of the sensitivity of ADC values to thecquisition and analysis of DW-MRI images, the experiencerom preliminary studies (33–35) is that DW-MRI evaluates to

certain degree the water mobility in the salivary glands. That, i

REFEREN

spin-echo sequence. Radiology 2000;216:347–358.

1

1

1

1

1

1

1

2

2

2

2

2

2

n combination with MR sialography, which is a straightfor-ard method to image the saliva in the salivary ducts, couldrovide insights into the mechanism of saliva production andransfer from the acinar cells to the salivary ducts.

CONCLUSIONS

Magnetic-resonance sialography is a noninvasive 3D imag-ng technique wherein saliva itself is imaged along the fullrajectory of the major salivary ducts, from the intraglandularpace to the oral cavity. We demonstrated that MR sialographyan depict radiation-induced changes to the salivary glands anducts post-RT. Registration of those radiation-induced effectso the 3D dose distribution is possible, and with repetitivemaging, they can be followed in time post-RT. This combi-ation makes MR sialography a novel and promising tool for

nvestigating radiation-induced xerostomia.

CES

1. Chambers MS, Garden AS, Kies MS, et al. Radiation-inducedxerostomia in patients with head and neck cancer: Pathogen-esis, impact on quality of life, and management. Head Neck2004;26:796–807.

2. Braam PM, Terhaard CH, Roesink JM, et al. Intensity-mod-ulated radiotherapy significantly reduces xerostomia com-pared with conventional radiotherapy. Int J Radiat Oncol BiolPhys 2006;66:975–980.

3. Hamlet S, Faull J, Klein B, et al. Mastication and swallowingin patients with postirradiation xerostomia. Int J Radiat OncolBiol Phys 1997;37:789–796.

4. Braam PM, Roesink JM, Moerland MA, et al. Long-termparotid gland function after radiotherapy. Int J Radiat OncolBiol Phys 2005;62:659–664.

5. Duncan GG, Epstein JB, Tu D, et al. National Cancer Instituteof Canada Clinical Trials Group. Quality of life, mucositis,and xerostomia from radiotherapy for head and neck cancers:A report from the NCIC CTG HN2 randomized trial of anantimicrobial lozenge to prevent mucositis. Head Neck 2005;27:421–428.

6. Lin A, Kim HM, Terrell JE, et al. Quality of life afterparotid-sparing IMRT for head-and-neck cancer: A prospec-tive longitudinal study. Int J Radiat Oncol Biol Phys 2003;57:61–70.

7. Konings AW, Coppes RP, Vissink A. On the mechanism ofsalivary gland radiosensitivity. Int J Radiat Oncol Biol Phys2005;62:1187–1194.

8. Dreizen S, Brown L, Handler S, et al. Radiation-inducedxerostomia in cancer patients. Effect on salivary and serumelectrolytes. Cancer 1976;38:273–278.

9. De Rossi G, Radioisotopic studies under pathologic condi-tions. In: De Rossi G, editor. Radioisotope study of salivaryglands. Boca Raton, FL: CRC Press; 1987. p. 52–81.

0. Jara H, Barish MA, Yucel EK, et al. MR hydrography: Theoryand practice of static fluid imaging. Am J Roentgenol 1998;170:873–882.

1. Kalinowski M, Heverhagen J, Rehberg E, et al. Comparativestudy of MR sialography and digital subtraction sialographyfor benign salivary gland disorders. Am J Neuroradiol 2002;23:1485–1492.

2. Becker M, Marchal F, Becker C, et al. Sialolithiasis and salivaryductal stenosis: Diagnostic accuracy of MR sialography with athree-dimensional extended-phase conjugate-symmetry rapid

3. Niemela R, Takal R, Paakko E, et al. Ultrasonography ofsalivary glands in primary Sjögren’s syndrome. A comparisonwith magnetic resonance imaging and magnetic resonancesialography of parotid glands. Rheumatology (Oxf) 2004;43:875–879.

4. Takagi Y, Sumi M, Sumi T, et al. MR microscopy of theparotid glands in patients with Sjögren’s syndrome: Quantita-tive MR diagnostic criteria. Am J Neuroradiol 2005;26:120.

5. Sartoretti-Schefer S, Kollias S, Wichmann W, et al. 3D T2-weighted fast spin-echo MRI sialography of the parotid gland.Neuroradiology 1999;41:46–51.

6. Astreinidou E, Bartels LW, Raaijmakers CPJ, et al. 3D MRsialography protocol for post-radiotherapy follow-up of thesalivary duct system. J Magn Reson Imag 2006;24:556–562.

7. Astreinidou E, Dehnad H, Terhaard CH, et al. Level II lymphnodes and radiation-induced xerostomia. Int J Radiat OncolBiol Phys 2004;58:124–131.

8. Chao KS, Majhail N, Huang CJ, et al. Intensity-modulatedradiation therapy reduces late salivary toxicity without com-promising tumor control in patients with oropharyngeal car-cinoma: A comparison with conventional techniques. Ra-diother Oncol 2001;61:275–280.

9. Wu Q, Mohan R, Morris M, Lauve A, et al. Simultaneousintegrated boost intensity-modulated radiotherapy for locallyadvanced head-and-neck squamous cell carcinomas. I: Dosi-metric results. Int J Radiat Oncol Biol Phys 2003;56:573–585.

0. Eisbruch A, Kim HM, Terrell JE, et al. Xerostomia and itspredictors following parotid-sparing irradiation of head-and-neck cancer. Int J Radiat Oncol Biol Phys 2001;50:695–704.

1. Konings AW, Faber H, Vissink A, et al. Radioprotectiveeffect of amifostine on parotid gland functioning is regiondependent. Int J Radiat Oncol Biol Phys 2005;63:1584–1591.

2. Konings AW, Faber H, Cotteleer F, et al. Secondary radiationdamage as the main cause for unexpected volume effects: Ahistopathologic study of the parotid gland. Int J Radiat OncolBiol Phys 2006;64:98–105.

3. Konings AW, Cotteleer F, Faber H, et al. Volume effects andregion-dependent radiosensitivity of the parotid gland. Int JRadiat Oncol Biol Phys 2005;62:1090–1095.

4. Ceylan C, van der Heide UA, Bol GH, et al. Assessment ofrigid multi-modality image registration consistency using themultiple sub-volume registration (MSR) method. Phys MedBiol 2005;50:N101–N108.

5. Roesink JM, Moerland MA, Battermann JJ, et al. Quantitative

2

2

2

2

3

3

3

3

3

3

1319MR sialography post-RT ● E. ASTREINIDOU et al.

dose-volume response analysis of changes in parotid glandfunction after radiotherapy in the head-and-neck region. Int JRadiat Oncol Biol Phys 2001;51:938–946.

6. Kashima HK, Kirkham WR, Andrews RJ. Postirradiationsialadenitis: A study of the clinical features, histopathologicchanges and serum enzyme variations following irradiation ofhuman salivary glands. Am J Roentgenol 1965;114:271–291.

7. Nomayr A, Lell M, Sweeney R, et al. MRI appearance ofradiation-induced changes of normal cervical tissues. Eur Ra-diol 2001;11:1807–1817.

8. Meirovitz A, Murdoch-Kinch CA, Schipper M, et al. Gradingxerostomia by physicians or by patients after intensity-modu-lated radiotherapy of head-and-neck cancer. Int J Radiat On-col Biol Phys 2006;66:445–453.

9. Blanco AI, Chao KS, El Naqa I, et al. Dose-volume modelingof salivary function in patients with head-and-neck cancerreceiving radiotherapy. Int J Radiat Oncol Biol Phys 2005;62:1055–1069.

0. Astreinidou E, Roesink JM, Moerland MA, et al. The potentialof using MRI to obtain spatial information about the salivary

gland function. Radiother Oncol 2005;76(Suppl):S145.

1. Morimoto Y, Ono K, Tanaka T, et al. The functionalevaluation of salivary glands using dynamic MR sialogra-phy following citric acid stimulation: A preliminary study.Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2005;100:357–364.

2. Zhang L, Murata Y, Ishida R, et al. Functional evaluation withintravoxel incoherent motion echo-planar MRI in irradiatedsalivary glands: A correlative study with salivary gland scin-tigraphy. J Magn Reson Imag 2001;14:223–229.

3. Thoeny HC, De Keyzer F, Claus FG, et al. Gustatory stimu-lation changes the apparent diffusion coefficient of salivaryglands: Initial experience. Radiology 2005;235:629–634.

4. Habermann CR, Cramer MC, Graessner J, et al. Functionalimaging of parotid glands: Diffusion-weighted echo-planarMRI before and after stimulation. Fortschr Rontgenstr 2004;176:1385–1389.

5. Thoeny HC, De Keyzer F, Boesch C, et al. Diffusion-weighted imaging of the parotid gland: Influence of the choiceof the b-values on the apparent diffusion coefficient value. J

Magn Reson Imag 2004;20:786–790.