969 STIR MR Imaging of the Orbit Sott W. tlas 1 Robert I. Grossman David. Hakney Herbert I. Goldberg Larissa T. ilaniuk Robert. Zimmerman Fifteen patients with CT-doumented orbital lesions were evaluated with MR imaging at 1.5 T with both onventional spin-eho (SE) and short inversion time inversion reovery (STIR) sequenes. Fat signal was reliably nulled at inversion times of approximately 120-200 mse in all ases, thereby allowing lear detetion of all retrobulbar lesions and normal strutures on STIR images as markedly hyperintense relative to fat. ll lesions were also learly depited on SE images; in fat, short repetition time/short eho time SE sequenes were at least as useful as STIR images for illustrating anatomi strutures and mass lesions, and in a muh shorter sanning time. Separation of opti nerve from periopti subarahnoid spae was lear on SE images, but often diffiult or impossible on STIR images owing to the relatively high intensity of normal opti nerves on STIR images. The synergism of relaxation prolongation with STIR atually resulted in loss of information, as any ability to separate the effets of T1 from T2 on signal intensity was impossible when STIR was the sole pulse sequene. We believe that more information is obtained with standard SE sequenes than with STIR sequenes, and therefore SE remains the method of hoie for orbital MR imaging. Orbital lesions have been evaluated by many investigators with spin-eho (SE) MR imaging (1-9]; and with the more reent implementation of high-resolution surfae-oil tehniques [10, 11], signifiant advanes have been made in MR imaging of the orbit. The orbit is somewhat unique in neuroradiology, in that the bakground stroma in this area is omposed mainly of fat, whih has a harateristi appearane on SE images. It has been proposed that short inversion time inversion reovery (STIR) would be useful in evaluating orbital disease [12, 13], sine one an seletively suppress signal from a tissue (suh as fat) based on its T1 by hoosing the appropriate inversion time (TI) with this inversion reovery (IR) tehnique [14] and thereby highlight retrobulbar lesions, the vast majority of whih are mainly water-ontaining. Furthermore, prolongation of T1 and T2, relaxation behavior harateristi of most diseases, is synergisti with STIR. These relaxation harateristis oppose eah other in their ontribution to signal intensity in SE imaging [15]. We evaluated the usefulness of STIR MR in depiting orbital disease in 15 patients and ompared these images with SE images in these ases. This artile appears in the September/Otober 1988 issue of JNR and the November 1988 issue of JR. Reeived September 16, 1987; aepted after revision February 18, 1988. Presented at the annual meeting of the merian Soiety of Neuroradiology, New York City, May 1987. 1 ll authors: Department of Radiology, Hospital of the University of Pennsylvania, 3400 Sprue St., Philadelphia, P 19104. ddress reprint requests to S. W. tlas. JNR 9:969-974, September/Otober 1988 0195-6108/88/0905-0969 merian Soiety of Neuroradiology Subjets and Methods Fifteen patients, ranging in age from 11 weeks to 66 years, with orbital lesions doumented by CT were evaluated with MR imaging on a 1.5-T system: ll patients underwent both SE and STIR imaging, and images were obtained in the same plane in eah patient with these tehniques. SE sequenes were performed with both short repetition time (TR)/short eho time (TE) sequenes, 600/20/1 (TRITE/exitations), and long TR/multieho TE sequenes, 2500/20-30, 80/2. STIR sequenes used TR s of 800-2500 mse and Tis of 120-200 mse. The Tis were seleted for the purpose of suppressing signal from fat (13). ll images were displayed with magnitude reonstrution. The appearane of lesions, as well as normal retrobulbar strutures, was ompared by using these two tehniques. * General Eletri Medial Systems, Milwaukee, WI.
970 TLS ET L. JNR:9, September/Otober 1988 The signal intensity (SI) within a pi xel aquired with IR an be determined from the equation [15], SI = k x N (H) (e- TE {T2)[1-2 (e-t1{t') + e- TR{T' ], (1) where k = a onstant expressing system gain, N(H) is the mobile proton density, and T1 and T2 are the longitudinal and transverse relaxation times, respetively. The TI at the "null point" (i.e., the value of TI for whih SI equals zero for the tissue with a speifi T1) is given by the equation [15], Tl nu " = - T1 x In['h (1 + e-tr{t' )]. (2) For TR» T1, Tl nu" is given by the approximation [13], Tl nu" = T1 x In 2. (3) With magnitude image reonstrution tehniques (when TI is seleted for the purpose of nulling signal from fat), tissues that have reovered longitudinal magnetization either more or less than fat will have a positive signal intensity (and fat will demonstrate very low signal). The Tl nu" for fat was determined either from approximation from equation 2, based on a T1 of 200-250 mse for orbital fat [16] or from diret observation of fat signal peak nulling on the signal profile obtained during the pres an with hanges in TI [17]. Short TR sans were aquired in 2 min 51 se. Long TR sans were obtained in 10 min 30 se. STIR sans were obtained in times ranging from 3 min 40 se to 10 min 30 se, depending on the TR. Results ll lesions were learly delineated by both SE and STIR tehniques. Short TRITE SE images demonstrated all lesions, as well as normal retrobulbar anatomy, learly depiting them against a high-intensity bakground of retrobulbar fat (Figs. 1-3). Clear definition of periopti CSF separate from the opti nerve was usually possible with both short TRfTE and long TRITE sans (Figs. 1, 2, and 4). On Long TRITE images, lesions were portrayed against the bakground of low-intensity orbital fat (Figs. 1-4). Normal extraoular musles were only minimally hyperintense or isointense relative to fat on this sequene. On SE sequenes, the well-known misregistration bands of high intensity and low intensity at fat-water interfaes [18] were noted along the frequeny-enoding axis (Fig. 1). These artifats, although often fairly prominent on SE images, were not problemati in any ase. Fat signal was onsistently suppressed on STIR imaging at Tis of 120-200 mse in these 15 patients, thereby depiting all other normal retrobulbar strutures as relatively hyperintense. Changes in TR alter signal intensities in STIR, sine Tl nu" is different if TR is not suffiiently long (equation 2). Inreased TR also inreases overall signal. Varying TE also hanges signal intensities somewhat, sine a longer TE adds ontrast on the basis of T2 differenes (i.e., lesions with long T2 will have a higher signal intensity with a long TE). Relatively small hanges in TI resulted in markedly different ontrast relationships (Fig. 5). Therefore, on STIR images, extraoular musles and opti nerves were markedly hyperintense relative to fat. It was usually diffiult and often virtually impossible to separate opti nerve from surrounding CSF beause of the high intensity of normal opti nerve (Figs. 1-4). The highintensity band at one edge of fat-water interfaes from hemial-shift misregistration [18] was eliminated when signal from fat was nulled (Fig. 1). region of signal void was identifiable on STIR images at all fat-water interfaes (Figs. 3C, 4, and Fig. 1.-lntraonal avernous hemangioma., Coronal SE image, 600/20., Coronal SE image, 2500/80. C, Coronal STIR image, 2500/140/20. Well-irumsribed mass (1) is well delineated between opti nerve (2) and medial retus musle (3) on all sequenes. Mass is isointense relative to musle on short TR/TE () and markedly hyperintense on long TRITE () images. Mass is hyperintense relative to fat, but approximately isointense relative to musle, on STIR image (C). Note hemial-shift artifats (arrowheads) on SE images ( and ), whih are absent on STIR image (C). lso, relative hyperintensity of opti nerve (2) on STIR image is not easily distinguishable from peri opti CSF.
JNR:9, September/Otober 1988 STIR IMGING OF THE ORIT 971 Fig. 2.-lntraonal, partially thrombosed avernous hemangioma., xial SE image, 600/20., Coronal SE image, 600/20. C, Coronal SE image, 2500/80. D, Coronal STIR image, 2500/200/25. Marked inhomogeneity in mass (1) on all sequenes is onsistent with partial thrombosis of avernous hemangioma, whih deviates opti nerve medially. Note lear separation of opti nerve from peri opti CSF on short TR/TE images ( and ) but not on STIR image (D). In this ase, long TR/TE image (C) does not learly distinguish opti nerve from peri opti CSF. 5) [19]. ll lesions were markedly hyperintense relative to fat on STIR sequenes (Figs. 1-4). Intralesional heterogeneity was depited on all SE sequenes as well as on STIR when present in any sequene (Fig. 2). Disussion IR imaging has been applied to various areas of the body for obtaining images with high ontrast, the degree of whih is mainly dependent on the TI and its relationship to the T1 of the tissue of interest [14, 15]. unique harateristi of IR sequenes is the phenomenon of signal nulling. This phenomenon is due to partial reovery of longitudinal magnetization after the initial 180 RF inverting pulse just to the "rossover" point [15], so that the subsequent 90 pulse eliminates all magnetization in the transverse plane just before signal detetion. This results in the depition of all tissue with that partiular T1 as very low signal [14, 15]. The null-pont TI (Tl nuli) is based on the T1 of the tissue from whih one is o interested in nulling the signal (Fig. 6). Tl null an be determined from the signal-intensity equation (equation 1) for IR [15]. If one onsiders situations in whih TR is» T1 than the equation simplifies to equation 3. Therefore, one an approximate Tl null by multiplying 0.69 x T1. (Note that the Tl null will vary with applied field strength, sine T1 hanges with field-strength hanges.) more aurate and individualized assessment of Tl null an also be obtained by diret observation of the signal profile obtained from presanning, whereby the (major) fat peak is separable from the water peak (due to the 3.5-ppm differene in resonant frequenies of fat and water protons) [18, 20]. y varying TI, one an visualize suppression of the fat peak, thereby obtaining the Tl null for eah patient individually [17]. This method of presanning an be performed in about 2 min. STIR sequenes theoretially an be used to highlight all lesions residing in a tissue with a T1 signifiantly different from the T1 of those lesions simply by seleting TI to onur with the quantity, 0.69 x T1 of the bakground normal tissue. This has been applied with some suess to deteting intrahepati
972 TLS ET L. JNR:9, September/Otober 1988 Fig. 3.-lnflammatory orbital pseudotumor., Coronal SE image, 600/20. 8, Coronal SE image, 2500/80. C, Coronal STIR image, 2500/200/20. Large retrobulbar mass (1) is isointense relative to fat on long TR/long TE image (8), harateristi of benign inflammatory pseudotumor as opposed to malignant lesions. STIR image (C) depits mass, however, as markedly hyperintense, indistinguishable from malignant lesions suh as lymphoma. lso note again hyperintense opti nerve (2) and extraoular musles (3) on STIR image (C). oundary artifat of hypointensity is loated at all fat-water interfaes on STIR image. Fig. 4.-0rbitallymphoma., Coronal SE image, 2500/80. 8, Coronal STIR image, 2500/150/20. Large mass (1) in medial aspet of orbit is hyperintense on both long TR/TE SE () and STIR (8) sequenes. Note marked ontrast between hyperintense opti nerve (2) and hyperintense extraoular musles (3) vs markedly hypointense fat (4) on STIR image.
JNR :9, September/Otober 1988 STIR IMGING OF THE ORIT 973 Fig. 5.-Effet on ontrast of hanges in TI., xial STIR image, 2500/200/20., xial STIR image, 2500/250/20. Note total inversion of ontrast relationships in retrobulbar strutures between fat (1) and water'ontaining strutures with 50-mse hange in TI. When TI is seleted so that signal from fat is nulled (), all retrobulbar strutures, inluding opti nerve (2) and extraoular musles (3), are markedly hyperintense. It is impossible to separate periopti CSF from opti nerve on this image. lso note marked hypointensity (boundary artifat) at all fat-water interfaes on both STIR sequenes. 8-51(0) - SI(R) o Tli'1E --+- GO- OTP (S HOTE ST TI ) F T (S HORT T1 ) C HU SCLE (LO NG T1 ) O LES IO N (LON GEST T1 ) T LON G TE. SI INCR ERSES FOR LONG T2 TI SS UES Fig. S.-Signal intensity (Sl) vs T1 at null point for fat. Note that rate of reovery of longitudinal magnetization (i.e., Tl), relative to TI, determines signal intensity as ompared with nulled tissue (fat). In addition, as TE is inreased, tissues with longer T2 will have higher signal intensity. Therefore, lesions with long Tl and long T2 will have inreased intensity beause of both these fators. lesions [14]. It has been suggested that STIR might be very useful in the orbits [12, 13], sine most of the retrobulbar ompartment is oupied by fat, whih has a T1 signifiantly different from the T1 of other (water-ontaining) orbital strutures and from the expeted T1 of the vast majority of (waterontaining) orbital lesions. Furthermore, a relatively long TE an also be seleted with STIR, so that prolongation of T1 and prolongation of T2 both ause an inrease in signal intensity. This synergisti effet of T1 and T2 prolongation is opposite to the effet on signal intensity of inreased T1 and T2 in SE sequenes; that is, an inrease in T1 redues signal intensity and an inrease in T2 inreases signal intensity on SE imaging [15]. Sine the great majority of lesions exhibit long T1 and long T2, theoretially STIR would be more sensitive to subtle hanges in relaxation parameters, whih, at least in theory, might not be deteted easily on SE images. Inreases in spin density further augment signal intensity in both STIR and SE imaging. In our series, STIR sequenes did indeed depit lesions with marked onspiuity against a bakground of very-iowintensity orbital fat (Figs. 1-4). In fat, all retrobulbar strutures, inluding extraoular musles and the opti nerve/ sheath omplex, were markedly hyperintense on STIR images. This allowed superb delineation of anatomi relationships behind the globe, and when omparing the long TR/ long TE SE image with the STIR image (on both images, fat is low-intensity), there was no question that STIR depited extraoular musles and opti nerves with greater ontrast than did the long TRfTE SE images (Figs. 1-4). However, short TRfTE SE images also allowed superb anatomi delineation of retrobulbar strutures, differing from STIR only in that the fat bakground was of high rather than low intensity. Furthermore, separation of opti nerve from the surrounding CSF was usually aomplished easily with both short and long TRfTE SE images (Figs. 1, 2, and 4). In our experiene, this separation was diffiult on STIR images, sine the normal opti nerve is relatively hyperintense (Figs. 1-5) when fat signal is suppressed. ll lesions were markedly hyperintense on STIR sequenes. lthough in theory the synergism of prolonged T1 and T2 might be benefiial for deteting subtle retrobulbar hanges, no lesion in our series seen on STIR was not detetable on short TRfTE, long TR/short TE, or long TR/long TE SE images. Other authors, however, have noted the usefulness of STIR for doumenting opti neuritis [13]. In addition, separation of intrinsi relaxation behavior harateristis, attainable (to a ertain extent) from analysis of the three SE images, was virtually impossible from an analysis of STIR images,
974 TLS ET L. JNR :9, September/Otober 1988 sine prolonged T1, prolonged T2, and inreased spin density all inrease signal intensity on STIR. This is partiularly disturbing in the evaluation of orbital pseudotumor, whih has been reported to demonstrate relatively harateristi signalintensity patterns on SE images [21], probably related to the highly fibrous nature of this entity. These lesions are usually isointense or only minimally hyperintense relative to fat on long TRllong TE SE images (Fig. 3), in ontrast to the appearane of linially similar but more ominous diseases, suh as retrobulbar malignanies (Fig. 4), whih are usually markedly hyperintense on long TRllong TE SE images [21]. STIR depits both types of disease as markedly hyperintense, reduing any potential speifiity for MR imaging in distinguishing these lesions. Furthermore, ontrast relationships that use STIR sequenes are highly dependent on aurate flip angles, sine errors in flip angle would alter the null point and would dramatially hange relative signal intensities. One side benefit to suppressing signal from orbital fat is the elimination of the high-intensity band perpendiular to the frequeny-enoding axis seen on SE images at fat-water interfaes due to misregistration (Fig. 1) [18]. Pratially speaking, however, this hemial-shift misregistration artifat is now well known to all those who interpret SE images and is rarely a problem in diagnosis. Other, more rapid methods have been proposed to eliminate possible onfusion from this artifat in orbital imaging [22, 23]. In fat, another artifat is present on STIR images at fat-water interfaes, onsisting of a rim of signal void (Figs. 3C, 48, and 5). This an be asribed to intravoxel signal anellation when longitudinal magnetization of different materials are approximately equal in magnitude but opposite in diretion (i.e., they are 180 out of phase). In this ase, voxels that derive signal ontributions equally from fat and water protons (suh as the situation at fat-water interfaes) have a reovered longitudinal magnetization that averages to zero. This situation, in fat, is the same as if the T1 of the voxel orresponded to the null point. This has been termed the "opposed magnetization artifat," whih depends on the differene in T1 s between tissues rather than on relative hemial-shift differenes [19]. In onlusion, STIR imaging an be performed to suessfully suppress signal intensity from orbital fat and thereby provide high-ontrast images of the orbit. These images allow exellent anatomi delineation of normal strutures as well as retrobulbar mass lesions. ll lesions in our series were depited as markedly hyperintense on STIR images. However, short TRITE SE sequenes were equally useful in illustrating anatomi strutures and mass lesions, and in a muh shorter sanning time. This derease in sanning time allows highresolution surfae-oil imaging to be performed with less interferene from motion artifats, a frequent problem in orbital MR imaging. Furthermore, separation of opti nerve from periopti subarahnoid CSF was lear on SE images, but often diffiult or impossible when STIR was the sole pulse sequene. lthough the high intensity of hemial-shift artifats was virtually eliminated with STIR, these artifats are well known and easily reognizable on SE images. Therefore, despite the high-ontrast images of the orbit produed by STIR, whih allows exellent anatomi delineation of normal strutures and mass lesions, we believe that more information is obtained with standard SE sequenes, and therefore SE remains the method of hoie for orbital MR imaging. CKNOWLEDGMENT We thank Kara R. Reynolds for assistane in manusript preparation. REFERENCES 1. Hawkes RC, Holland GN, Moore WS, Rizk S, Worthington S, Kean M. NMR imaging in the evaluation of orbital tumors. JNR 1983;4 :254-256 2. Edwards JH, Hyman R, Vaira SJ, et al. 0.6 T magneti resonane imaging of the orbit. JNR 1985;6:253-258 3. Daniels DL, Herfkens RJ, Gager WE, et al. MagnetiC resonane imaging of the opti nerves and hiasm. Radiology 1984;152:79-83 4. Hans JS, enson JE, onstelle CT, lfidi RJ, Kaufman, Levin M. MagnetiC resonane imaging of the orbit: a preliminary experiene. Radiology 1984;150 :755-759 5. ilaniuk LT, Shenk JF, Zimmerman R, et al. Oular and orbital lesions: surfae oil MR imaging. Radiology 1985;156 :669-674 6. Sobel DR, Kelly WM, Kjos O, Char 0, randt-zawadzki M, Norman D. MR imaging of orbital and oular disease. JNR 1985;6:259-264 7. Sullivan J, Harms SE. Surfae-oil MR imaging of orbital neoplasms. JNR 1986;7 :29-34 8. Mafee MF, Peyman G, Grisolano JM, et al. Malignant uveal melanoma and simulating lesions: MR imaging evaluation. Radiology 1986;160: 773-780 9. tlas SW, ilaniuk L T, Zimmerman R, Hakney D, Goldberg HI, Grossman RI. Orbit: initial experiene with surfae oil spin-eho MR imaging at 1.5 T. Radiology 1987;164 :501-509 10. Shenk JF, Hart HR Jr, Foster TH, et al. Improved MR imaging of the orbit at 1.5Twith surfae oils. JNR 1985;6 :193-196 11. Shenk JF, Hart HR, Foster TH, Edelstein W, Hussain M. High resolution magneti resonane imaging using surfae oils. In: Kressel HY, ed. Magneti resonane annual 1986. New York: Raven, 1986 :123-160 12. Smith FW, Parek HS, Forrester J, Redpath TW. MR imaging of the orbit and eye using inversion reovery sequenes. Radiology 1986;161(p):286 13. Johnson G, Miller OM, MaManus 0, et al. STIR sequenes in NMR imaging of the opti nerve. Neuroradio/ogy 1987;29 :238-245 14. ydder GM, Young IR. MR imaging: linial use of the inversion reovery sequene. J Gomput ssist Tomogr 1985;8 :585-587 15. Wehrli FW, MaFall JR, Glover GH, et al. Dependene of nulear magneti resonane (NMR) image ontrast on intrinsi and pulse sequene timing parameters. Radiology 1985;157 :573 16. ottomley P. Foster TH, rsinger RE, pfeifer LM. review of normal tissue hydrogen NMR relaxation times and relaxation mehanisms from 1-100 MHz: dependene on tissue type, NMR frequeny, temperature, speies, exision, and age. Med Phys 1984;11 :425-448 17. Newman R. Operator's pulse sequene data notes. Milwaukee: General Eletri, 1987 18. abok EE, rateman L, Weinreb JC, Horner SO, Nunnally RL. Edge artifats in MR images: hemial shift effet. J Gomput ssist Tomogr 1985;9 : 252-257 19. Hearshen DO, Ellis JH, Carson PL, Shreve P, isen M. oundary effets from opposed magnetization artifat in IR images. Radiology 1986; 160: 543-547 20. Dixon WT. Simple proton spetrosopi imaging. Radiology 1984;153: 189-194 21. tlas SW, Grossman RI, Savino PJ, et al. Surfae oil MR of orbital pseudotumor. JNR 1987;8:141-146 22. Daniels DL, Kneeland J, Shimakawa, et al. MR imaging of the optic nerve and sheath: orreting the hemial shift misregistration effet. JNR 1986;7:249-253 23. tlas SW, Grossman RI, xel L, et al. Orbital lesions: proton spetrosopi phase-dependent ontrast MR imaging. Radiology 1987;164 :510-514