- Email: [email protected]
Magnetic Resonance Imaging Versus Computed Tomography of Leukocoric Eyes and Use of In Vitro Proton Magnetic Resonance Spectroscopy of Retinoblastoma
Magneti c Resonance Imaging Versus Comput ed Tomogra phy of Leukocoric Eyes and Use of In Vitro Proton Magneti c Resonance Spectroscopy of Retinoblastoma
MAHMOOD F. MAFEE, MD, 1 MORTON F. GOLDBERG, MD/ STEVEN B. COHEN, MD/ 1 EFSTATHIOS D. GOTSIS, PhD, 1 MARC SAFRAN, MD, 2 LAVANYA CHEKURI, 1 BAHRAM RAOFI, MD Abstract: To evaluate the usefulness of magnetic resonance imaging (MRI) in the evaluation of leukocoric eyes, the authors studied 28 patients with either leukocoria or intraocular mass with a 1.5-tesla (T) MRI imager. Retinoblastomas were reliably distinguished from Coats' disease, toxocariasis, and persistent hyperplastic primary vitreous on the basis of MRI findings. Calcification cannot be reliably detected on MRI scans. Lesions elevated less than 4 mm may not be detected reliably by MRI at this time. Computed tomography (CT) can detect calcification with a high degree of accuracy. Retinoblastomas appeared as moderately hyperintense masses on T1- and proton-weighted MRis. They be came hypointense in T2 -weighted MRis. This MRI characteristic is similar to that of uveal melanoma. Intraocular calcification in children especially younger than 3 years of age is highly suggestive of retinoblastoma. In the diagnosis of reti noblastoma, MRI is not as specific as CT because of its lack of sensitivity in detecting calcification. However, MRI, because of its superior contrast resolution, offers more information in the differentiation of pathologic intraocular conditions responsible for leukocoria. The authors also describe their preliminary work of in vitro proton magnetic resonance spectroscopy of eyes with retinoblastoma and an eye with uveal melanoma in an 18-year-old black woman . Ophthalmology
96:965-976, 1989
Originally received: October 10, 1988. Revision accepted: February 24, 1989. 1
Department of Radiology and Magnetic Resonance Center, University of Illinois College of Medicine, Chicago. 2 UIC Eye Center, Eye and Ear Infirmary, University of Illinois College of Medicine, Chicago. Presented at the American Academy of Ophthalmology Annual Meeting, Las Vegas, October 1988.
Supported by core grant 1792 from the National Eye Institute, Bethesda, Maryland, and by an unrestricted research grant from Research to Prevent Blindness, Inc, New York , New York.
Reprint requests to Mahmood F. Malee, MD, 840 South Wood St, Magnetic Resonance Center, M/C 711 , University of Illinois at Chicago, Chicago, IL 60612.
965
OPHTHALMOLOGY
•
JULY 1989
•
VOLUME 96
•
NUMBER 7
F'ig I. Retinoblastoma. A, CT scan shows a mass (arrows) with multiple calcifications (arrowheads). B, T 1-weighted MR1 scan showing a relatively hyperintense mass (arrows). The tumor extends into the optic disc; how ever, the optic nerve is intact as confirmed on Figures IC and D. C, proton-weighted MRI scan showing a relatively hyperintense mass (ar rows). Notice that the calcification seen on Figure lA is faintly seen here as an area of hypointensity. D, Trweighted MRI scan shows that the retinoblastoma is hypointense (arrows). Notice hypointensity of calcifi cation (arrowheads) which is easily detected on CT scan in Figure I A.
During examination of a child with leukocoria, the major diagnostic considerations are retinoblastoma, per sistent hyperplastic primary vitreous (PHPV), Coats' dis ease, retinopathy of prematurity (ROP), congenital cat aract, toxocariasis, chronic retinal detachment associated with retrolental fibrosis, retinal astrocytoma, choroidal hemangioma with previous hemorrhage, and a variety of other nonspecific causes of leukocoria. I-s Retinoblastoma is the most common intraocular ma lignancy of infants and children, occurring in approxi mately 1 in 18,000 live births. 6•7 The detection and clinical differentiation of retinoblastoma from simulating lesions may be difficult. 8 The manner of intraocular and extra ocular extension, patterns of metastasis and recurrence,
966
ocular complications, and associated malignancies in pa tients with retinoblastoma make the diagnosis of retino blastoma one of the most challenging problems of pedi atric ophthalmology and radiology. Although ophthal moscopic recognition of retinoblastoma is often reliable, imaging modalities should be used on all patients sus pected of harboring a retinoblastoma in order to detect any gross retrobulbar spread, intracranial metastasis, and secondary tumors. Of the imaging techniques available, ultrasonography, computed tomography (CT), and mag netic resonance imaging (MRI) are the most useful to incorporate into the clinical examination of patients with leukocoria. 1-4·9 • 10 We have evaluated the use of MRI and CT studies in 27 patients with retinoblastoma and sim
MAFEE et al
•
MAGNETIC RESONANCE IMAGING
Fig 2. Retinoblastoma. A, T 1-weighted oblique sagittal MRI scan shows slight increased intensity of the globe. Notice hyperintensity of subretinal exudate (arrows). B, Trweighted axial MRI scan shows markedly hy pointense mass, filling most of the left globe. The subretinal fluid remains hyperintense (arrows). The tumor calcification appears as very hypoin tense image (arrowhead). Notice that calcification cannot be recognized on T 1-weighted (Fig 2A) image which has been obtained along the optic nerve and which has included the calcification.
ulating lesions to evaluate the incidence of calcification detected by MRI and CT in retinoblastoma, to delineate the diagnostic accuracy of MRI and CT in differentiating retinoblastoma from other simulating lesions, and to de termine the potential accuracy of MRI in diagnosing ret robulbar and intracranial spread. We also describe our preliminary work on in vitro proton magnetic resonance spectroscopy (MRS) of retinoblastoma and an eye with uveal melanoma.
SUBJECTS AND METHODS Our study group consisted of 28 patients: 18 with ret inoblastoma, 3 with Coats' disease, 3 with PHPV (possibly caused by Norrie's disease and Warburg syndrome), 1 with toxocariasis, I with mesoectodermal leiomyoma, 1 with myelinated nerve fibers, and 1 with malignant uveal melanoma. The patients ranged in age from 7 months to
Fig 3. Retinoblastoma. A, T 1-weighted MRI scan shows slight hyper intensity of the entire right globe (curved arrow). Notice normal apical and intracanalicular optic nerve (arrow). B, proton-weighted MRI scan obtained I year after enucleation shows recurrent tumor along optic nerve (hollow arrows) as compared with normal left optic nerve (arrow). Notice involvement of suprasellar cistern (curved arrow) as compared with that of I year ago (Fig 3A). C, proton-weighted coronal MRI scan showing marked tumor in the suprasellar cistern (black arrows). Notice normal internal carotid artery within the cavernous sinus (white arrow).
967
OPHTHALMOLOGY
•
JULY 1989
•
VOLUME 96
•
NUMBER 7
Fig 4. Coats' disease. A, proton-weighted MRI scans show increased intensity of the left globe. B, Tz-weighted MRI scans show hyperintensity of the left globe. Notice the leaves of the detached retina (arrows). The hyperintensity of left globe in the proton-weighted MRI scan is due to subretinal exudate and lipoproteinaceous fluid in the subretinal space.
Fig 5. Toxocariasis. Top. proton-weighted MRI scan shows a hyperintense mass (arrow) in the right globe. Bottom. T 2-weighted scan shows that the mass is hyperintense (isointense to vitreous).
968
18 years. All patients with retinoblastoma had typical ophthalmoscopic findings consistent with retinoblastoma. All the Coats' disease patients except one had typical ophthalmoscopic findings consistent with Coats' disease. Ophthalmoscopic findings in the other patients were con sistent with the respective clinical diagnosis, except in the patient with mesoectodermalleiomyoma who was initially diagnosed to have a ciliary mass of unknown type. The imaging studies were performed, in the majority of cases, using chloral hydrate for sedation (50 mg/kg of body weight) or general anesthesia. Magnetic resonance imaging studies, except in two patients, were performed with a 1.5-tesla (T) Signa unit (General Electric, Milwau kee, WI) with a 3- or 5-mm thick section with 0.6 to 1.5 mm intersection space. Single echo spin-echo (SE) pulse sequences were obtained with a repetition time (TR) of 300 to 800 msec and an echo time (TE) of 20 to 25 msec (TR/TE = 300-800/20-25 msec). Multiple SE pulse se quences were obtained with a TR of 1500 to 2500 msec and aTE of20 to 100 msec. The early echo (20-25 msec) in this sequence yielded proton-weighted (PW) images, and the later echo (30-1 00 msec) yielded T rweighted im ages. 11 The single echo SE sequence yielded T 1-weighted images. 11 The studies were most often performed using head coil with one excitation, a 256 X 256-matrix, and a 20- to 24 cm field of view. We routinely obtain the MRI ofthe eye
MAFEE et al
•
MAGNETIC RESONANCE IMAGING
for 10 weeks Fig 6. Retinopathy of prematurity. A 7-month-old premature boy with a history of fetal respiratory distress necessitating intubation normal. On entirely be to thought was and age of months 2 was he when specialist and hood oxygenation for 2 weeks. He was seen by a retinal in this case were diagnoses differential The identifiable. not were retinas The opacified. were vitreous the of thirds two anterior the examination, globes with bilateral ROP, bilateral PHPV, and autosomal dominant vitreoretinopathy. A, proton-weighted MRI scan shows hyperintensity of the (curved arrow). detachment retinal shows scan MRI -weighted T B, arrow). (curved 2 detachments retinal funnel-shaped and (arrows) masses retrolental to a fibrotic related probably (arrows), mass retrolental Notice The hyperintensity of the globes is related to subretinal fluid and old hemorrhage. hemorrhage. layered recent a to related (arrowhead) hypointensity of area an also scar. Notice
and head. A surface coil 7.6 em in diameter (Medical Advance Corporation, Milwaukee, WI) and an experi mental butterfly-type surface coil (General Electric, Mil waukee, WI) were used with a few patients to improve spatial resolution. For surface coil study, a field view of 12 or 16 em was used. With the surface coil, we use a 3mm section to reduce the problem of partial volume av-
eraging. The thin section has the disadvantage of gener ating less signal (small volume) which may result in a decreased amount of Trweighted information in later echo, particularly in a rapidly decaying T 2-weighted signal. Because of this, and in the light of exaggerated motion artifacts related to inherent increased sensitivity of the surface coil, a retinoblastoma or other intraocular mass
Fig 7. Warburg syndrome. A, T 1-weighted MRI scan shows hyperintense subretinal fluid in both eyes. Notice the detached leaves of the retina (arrowheads). B, T 1-weighted MRI scan obtained 3 months later shows bilateral retinal detachment. Notice fluid level in the right vitreous chamber (arrow). This is believed to be due to serosanguineous fluid in the subhyaloid or subretinal space. Notice a tubular image (curved arrow) suggestive of congenital nonattached retina or Cloquet's canal.
969
OPHTHALMOLOGY
•
JULY 1989
Fig 8. Retinoblastoma. T 1-weighted MRI scan of an excised eye (the entire globe was replaced by tumor) obtained with the CSI magnetic resonance unit demonstrates a hyperintense globe.
may sometimes be better detected on Tz-Weighted images obtained with a 5-mm section and using the head coil instead of a surface coil. In vitro proton MRS and imaging was performed on a horizontal 2-T chemical shift imaging (CSI) unit (Gen eral Electric, Milwaukee, WI). A 6-inch birdcage imaging coil was used for imaging ofthe whole excised eye, whereas a custom-made 8-mm diameter solenoid coil was used for spectroscopy of freshly excised tumors (in the range of 300 mg). In the case of the uveal melanoma, no water suppression was necessary, whereas in the case of the ret inoblastomas water suppression was accomplished by the binomial 1-3-3-1 sequence, with a protocol similar to that reported for in vivo studies. 12- 14 The CT studies were performed with a General Electric 9800 CT scanner. To determine the presence of calcifi cation, 1.5-mm contiguous plain axial (transverse) scans were obtained through the orbits. Because ofthe superior contrast resolution ofMRI scans and to avoid unnecessary radiation as well as to contain costs, we did not perform contrast CT studies of the orbit and head in many of the patients.
RESULTS All retinoblastomas had calcification, which was de tected by CT scans. Calcifications as small as 2 mm were 970
•
VOLUME 96
•
NUMBER 7
detected by CT scanning. Calcifications on MRI scans may be seen as varying degrees of hypointensity in all pulse sequences (Fig 1). In contrast to CT scanning, which is highly specific for calcification, MRI may be nonspecific (Fig 2). Retinoblastomas on T 1-weighted and proton weighted MRI scans appeared as areas of moderately high signal (hyperintense) compared with the signal intensity of ~treous (Figs 1B and C, 2A). On T z-weighted MRis, retmoblastomas appeared as areas of moderately (Fig 1D) to markedly (Fig 2B) low signal (hypointense). The non calcified portion of the retinoblastomas was less hypoin tense than partially calcified retinoblastomas on T z weighted MRis. Tumors elevated less than 4 mm could not be definitely identified on MRI scans. One year after enucleation (with no evidence ofoptic nerve involvement on MRI and histologic examination), marked recurrence of tumor along the optic nerve and with extension into the suprasellar region developed in one patient (Fig 3). The intracranial portion of the tumor was isointense to brain in all pulse sequences. None of the patients with Coats' disease had calcifi cation evident on CT scans. The CT and MRI in one of these patients showed total retinal detachment (Fig 4). The subretinal fluid in this case appeared as a hyperintense image in all pulse sequences, similar to other exudative retinal detachments. The eye was removed because ofad vanced disease complicated by glaucoma, and the diag nosis of Coats' disease was confirmed on histologic ex amination. The CT findings of the toxocara granuloma could not be differentiated from noncalcified retinoblas toma. Magnetic resonance imaging, however, showed that the lesion was hyperintense in T 1- and proton-weighted MRis (Fig 5, top) but unlike retinoblastoma, the lesions appeared hyperintense in Tz-weighted MRis (Fig 5, bot tom). Computed tomographic scanning in patients with PHPV (possibly caused by Norrie's disease), ROP, and Warburg syndrome, showed dense vitreous chambers. Magnetic resonance imaging showed marked hyperin tensity of the affected vitreous chamber on T 1- T z-, and proton-weighted MRis (Fig 6). In one patient with War burg syndrome with bilateral suspected PHPV, the MRI scan showed bilateral retinal detachment with hyperin tensity ofsub retinal fluid (Fig 7A) and layered fluid (acute hemorrhage) on follow-up MRis (Fig 7B). In vitro proton MRI using the CSI unit was performed on two excised unfixed eyes with retinoblastoma. The le sions appeared hyperintense in T 1-weighted (Fig 8, 9A) and hypointense in T z-weighted MRis. (In one of these eyes only, T,-weighted MRI scans were obtained.) Reti noblastomas appeared more hyperintense in T 1-weighted MRis ofin vitro (Fig 9A and for comparison Fig 1B; same patient) than the in vivo T 1- and proton-weighted images. The in vitro magnetic resonance spectroscopy was per formed on one of these eyes. This disclosed several res onances that were tentatively assigned to inositol, lactate, N-acetyl aspartate, creatine, and choline (Fig 9B). In vitro MRI of an excised eye with uveal melanoma was performed on the CSI unit. The lesion appeared as hyperintense in T 1-weighted (Fig lOA) and hypointense
MAFEE et al
•
MAGNETIC RESONANCE IMAGING
10
B
a
6
4
2
0
FPM
Fig 9. Retinoblastoma. A, T 1-weighted MRI scan of excised eye (same patient as in Figure I) obtained with the CSI magnetic resonance unit demonstrates a hyperintense mass (arrows). 8, Water-suppressed in vivo proton magnetic resonance spectrum of 300 mg of tumor tissue of the same retinoblastoma tumor obtained with the CSI unit at 2 T. The spectrum was acquired with an 8-mm diameter solenoid coil which nevertheless was not completely filled with the small tumor sample, resulting thus in some loss of homogeneity and of course spectral resolution. The 1-3-3-1 binomial sequence was used for the water suppression, with an interpulse delay of 2.34 msec, for a total of 256 scans. Tentatively, the composite peak centered around 4.1 parts per million may he attributed to inositol, lactate as well as sugars, the peaks at 3.3 and 3.5 parts per million to inositol while the peak approximately 3 parts per million to creatine and phosphocreatine and choline. The peaks at approximately 2.3 to 2.0 parts per million are unassigned, although the peak near 2.0 parts per million may be due to N-acetyl aspartate.
in T z-weighted MRis (Fig I OB), characteristic of melanotic tumors. 15 The in vitro proton magnetic resonance spec trum of this uveal melanoma is shown in Figure I OC, which shows a resonance at approximately 6.9 parts per million that corresponds to melanin and melanin pre.. cursor. 14
DISCUSSION High-resolution, thin-section CT scanning can detect calcification within retinoblastoma with a high degree of accuracy. 1·4 Calcification may be small and single, large and single, multiple and punctate, or several fine-speckled foci. JA The noncalcified mass of retinoblastoma on CT scans may be very difficult to differentiate from simulating lesions. Computed tomography is of value in accurately assessing the extraocular and intracranial spread of reti noblastomas. Since the first report of the entity known as trilateral retinoblastomas, 16• 17 CT has played an important role in establishing the diagnosis. These ectopic tumors are typically located in the pineal or parasellar region.
Intraocular calcification in children, especially younger than 3 years of age, is highly suggestive of retinoblas toma.18-20 We have seen, however, calcifications in a child younger than 1 year of age and who had bilateral mi crophthalmos with colobomatous cysts. We share our experience with that of others: 1·19 namely, none of the simulating lesions, including PHPV, Coats' disease, toxo cariasis, ROP, chronic retinal detachment, retinal astro cytoma, choroidal osteoma, or optic nerve head drusen tends to contain calcification in the children (:::;;3 years of age) in whom retinoblastoma is usually diagnosed. In children older than 3 years of age, some of the simulating lesions, including retinal astrocytoma, ROP, toxocariasis, and optic nerve head drusen, can produce calcification. Calcifications that are easily detected by CT were difficult to recognize on MRI scans. Dense calcification was seen as an area of hypointensity in T 1-, Tz-, and proton weighted MRis (Fig 1C, D). In some cases, calcifications were not detected on MRI scans. In the diagnosis of ret inoblastoma, MRI is not as specific as CT scanning be cause of its lack of sensitivity in detecting calcification. However, the MRI appearance of retinoblastoma may be 971
OPHTHALMOLOGY
•
JULY 1989
•
VOLUME 96
•
NUMBER 7
Fig 10. Uveal melanoma in an 18-year-old black woman. A, T 1-weighted MRI scan of excised eye obtained with CSI magnetic resonance unit demonstrates a hyperintense mass (arrow). B, the mass (arrow) appears hypointense in Trweighted MRI scan. C, proton spectrum ofthe excised eye showing a large peak at approximately 6.9 parts per million, which may be assigned to aromatic amino acids and melanin. 14•25 The resonance at 4.77 parts per million is from water.
c 10
8
6
4
2
0
PPM
specific enough to differentiate retinoblastoma from sim ulating lesions. In all retinoblastomas (>3 mm), we ob served a mass that displayed MRI characteristics oflesions with relatively short T 1- and short Trrelaxation times. All retinoblastomas were seen as mildly to moderately hyperintense lesions on T 1- and proton-weighted MRI scans (Figs 1B, C; 2A; 3A). They became hypointense in T 2-weighted MRI scans (Figs 1D, 2B). This is very similar to MRI characteristics of uveal melanoma (Fig 11 ). In this study, none of the patients with PHPV or Coats' dis972
ease and one patient with toxocariasis demonstrated sim ilar MRI characteristics to those of retinoblastoma. Both CT and MRI have some advantages and disad vantages. Computed tomography is the study of choice for detection of calcification. Calcification as small as 2 mm can be detected by high-resolution thin section CT scanning. Magnetic resonance imaging has superior con trast resolution to CT scanning (Fig 2B). The phenomenon of partial volume averaging is more problematic in MRI than CT scanning, because with CT it is possible to obtain
MAFEE et al
•
MAGNETIC RESONANCE IMAGING
thinner sections (1-1.5 mm) than MRI (3 mm). In the study of eyes with suspected retinoblastoma, we recom mend plain CT ofthe eyes and MRI of the eyes and head. The optic nerve involvement, recurrence (Fig 3B, C), extracranial and intracranial (Fig 3C) involvement can be better evaluated by MRI than CT scanning. Magnetic resonance imaging appears to offer more information in the differentiation of pathologic intraocular conditions responsible for leukocoria (Figs 4- 7) and other lesions such as mesectodermal leiomyoma (Fig I2). Since the MRI characteristics of retinoblastomas in our study were similar to that of malignant uveal melanomas 11 • 15 (Figs I C, D; II), we decided to study with in vitro spectroscopy these two lesions in order to obtain additional information from their spectral analysis. Our preliminary study showed that the spectral appearance of retinoblastoma was completely different (Fig 9B) from that of uveal melanoma (Fig lOC). The major resonances in retinoblastoma were due to phosphocreatine and N acetyl aspartate. On the other hand, the major resonance in uveal melanoma was due to melanin or its precursors aromatic amino acids (tyrosin, phenylalinine). The N acetyl aspartate is a major resonance recorded on the spectra ofthe brain. 12 •14 The presence ofN-acetyl aspartate in our retinoblastoma specimen (Fig 9B) may be due to the fact that retinoblastoma is derived from primitive em bryonal retinal cells (either neuronal retinal cells or pho toreceptor). Although we cannot reach any definite con clusion from our preliminary in vitro spectroscopic ex perience of these two lesions, however, it is possible that further future study may prove that theN-acetyl aspartate and melanin precursors might be considered as a tumor marker for retinoblastoma and uveal melanoma, respec tively. In order to understand the technical jargon of magnetic resonance spectroscopy, a glimpse at the fundamental of magnetic resonance spectroscopy (MRS) can bring us to understand the importance of MRS and the potential use
Fig 11. Uveal melanoma. Proton-weighted MRI (top) and T2-weighted MRI (bottom) scans show a hyperintense mass (arrows) in proton weighted and hypointense mass in T 2-weighted MRis. Notice hyperin tense subretinal exudate (curved arrows).
of in vivo magnetic resonance spectroscopy in the future practice of combined magnetic resonance imaging and in vivo magnetic resonance spectroscopy. HIGH-RESOLUTION IN VITRO PROTON MAGNETIC RESONANCE SPECTROSCOPY OF HUMAN EYES
Since the discovery of nuclear magnetic resonance phenomenon, magnetic resonance spectroscopy has proved to be one of the most powerful tools in analytical chemistry. Magnetic resonance spectroscopy is a form of
Fig 12. Mesectodermal leiomyoma. An 8-year-old boy with a large tumor of the left ciliary body. Proton-weighted MRI scan (A) and Trweighted MRI scan (B) showing a hyperintense mass (arrows) arising from left ciliary body. Notice that its MRI characteristics are quite different from those of retinoblastoma.
973
OPHTHALMOLOGY
•
JULY 1989
spectroscopy in which, as a result of the magnetic prop erties of certain nuclei arising from their inherent axial spin, radiofrequency radiation is absorbed by the nuclei in the presence of a strong static magnetic field. The mag netic resonance spectrum, in fact, is a plot of signal in tensity versus resonance frequency. The appearance of magnetic resonance spectrum depends on the density of the extranuclear electron and molecular environment surrounding a nucleus. The presence of neighboring atoms and the diamagnetic "shielding currents" that are asso ciated with the distribution of circulating electrons around adjacent atoms slightly modifies the strength of local magnetic fields at a particular nucleus and hence modifies the resonance frequency,2 1-24 so that the details of mo lecular structure change the magnetic resonance frequency ofa nucleus in that molecule. For example, the resonance frequency of a proton within a water molecule differs slightly from that of a proton within a fat molecule, and the resonance frequency of a 31 P nucleus bound in an ATP molecule varies slightly from that of a 31 P nucleus within a phosphocreatine molecule. These small but spe cific displacements or shifts in frequency produce separate absorption peaks in a spectrum (chemical shift), which are "fingerprints" of molecular conformations and there fore directly aid in the determination of chemical struc ture.21·22 The chemical shift is measured and listed in parts per million with respect to a reference signal. Typical ref erence compounds, each of which has only a single res onance line, are water, chloroform, cyclohexane, benzene and tetramethyl silane for proton spectra, perfluorocyclo butane and trifluoroacetic acid for 19F, and phosphoric acid for 31 P. The chemical shift is of paramount importance in an alytical chemistry, because it allows the assignment ofdif ferent signal (lines) in a magnetic resonance spectrum to corresponding substances or molecular groups. Proton magnetic resonance spectra of live human tissues have been performed by several authors 12- 14·24 in which the major resonances were due to water and fat protons. In a recent report, Luyten and den Hollande~4 recorded res onances from N-acetyl aspartate and phosphocreatine in the human brain. Later on, Barany et al 12·13 recorded the spectra of human brain, which showed several resonances that were assigned to metabolites N-acetyl aspartate, glu tamine, phosphocreatine and creatine, choline derivatives, and taurine. They demonstrated the presence of urea, various amino acids, and nucleotides in liver cancer. They suggested that the presence of a large nucleotide pool is a characteristic of tumor tissue involved in rapid nucleic acid (RNA, DNA) synthesisY Figure 10 shows the in vitro MRI and the magnetic resonance spectroscopy of an eye of an 18-year-old black woman with uveal mela noma. The lesion is hyperintense in T 1-weighted and hy pointense in T rweighted MRis, characteristic of mela notic neoplasms reported by an in vivo MRI study. 15 The magnetic resonance spectrum (Fig lOC) shows a major resonance centered around 6.9 parts per million, which may be assigned to melanin or its precursors Tyrosin and phenylalanine. 14·25 The in vitro spectnm of retinoblas 974
•
VOLUME 96
•
NUMBER 7
toma (Fig 9B) was different from that of uveal melanoma (Figs 9B, 1OC), whereas both lesions appear to have similar characteristics on MRI. Our in vitro proton magnetic res onance spectroscopy data have to be considered as pre liminary results. Caution must be exercised in interpre tation ofspectra. Because of our interest in in vivo proton magnetic resonance spectroscopy, we also tried to obtain in vivo water-suppressed spectra from two patients with retinoblastoma and two with Coats' disease. The spectra, however, were less than optimal because of technical problems and therefore were excluded from this com munication. We believe, however, that the potential of in vivo proton spectroscopy to differentiate leukocoria should be investigated. Our previous investigation of in vivo proton spectroscopy has shown that the levels of some metabolites (e.g., melanin) may serve as a marker in the diagnosis ofmelanoma.14 The technique ofin vivo section selective water-suppressed proton spectroscopy reported by several authors 12·13 has some limitations. Recently, a method ofspatially resolved spectroscopy was introduced by Luyten and den Hollande~ 4 to obtain signal from a particular volume only (voxel magnetic resonance spec troscopy). This technique avoids contributions of signal from superficial tissues. The stimulated echo acquisition mode sequence introduced by Frahm and collaborators25 is a new technique that has some advantages over spatially resolved spectroscopy for recording metabolites by proton magnetic resonance spectroscopy in humans. Our pre liminary results with in vitro proton spectroscopy to ob tain biochemical information (Figs 9B, 1OC) is encour aging. Early experience with in vivo proton magnetic res onance spectroscopy has shown its potential for obtaining important biochemical information, 12- 14·24 thus enhancing the diagnostic sensitivity of magnetic resonance studies. Future work on combined MRI and in vivo magnetic resonance voxel spectroscopy is needed, with the goals of characterizing abnormal conditions according to their spectral patterns and of identifying tumor and metabolic markers.
ACKNOWLEDGMENT The authors thank Joel Schulman, MD, for providing three cases.
REFERENCES 1. Char DH, Hedges TR Ill, Norman D. Retinoblastoma: CT diagnosis. Ophthalmology 1984; 91 :1347-50. 2. Goldberg MF, Mafee MF. Computed tomography for diagnosis of persistent hyperplastic primary vitreous (PHPV). Ophthalmology 1983; 90:442-51 . 3. Mafee MF, Goldberg MF, Valvassori GE, Capek V. Computed to· mography in the evaluation of patients with persistent hyperplastic primary vitreous (PHPV). Radiology 1982; 145:713-7. 4. Mafee MF, Goldberg MF, Greenwald MF, et al. Retinoblastoma and
MAFEE et al
•
MAGNETIC RESONANCE IMAGING
simulating lesions: role of CT and MR imaging. Radial Clin North Am 1987; 25:667-82. Abramson DH. Retinoblastoma: diagnosis and management. CA 1982; 32:130-40. Shields JA. Diagnosis and Management of Intraocular Tumors. St Louis: CV Mosby 1983; 437. Devesa SS. The incidence of retinoblastoma. Am J Ophthalmol1975; 80:263-5. Char DH. Current concepts in retinoblastoma. Ann Ophthalmol1980; 12:792-804.
and simulating lesions: MR imaging evaluation. Radiology 1986; 160: 773-80. 16. Bader JL, Meadows AT, Zimmerman LE, et al. Bilateral retinoblastoma with ectopic intracranial retinoblastoma: trilateral retinoblastoma. Can cer Genet Cytogenet 1982; 5:203-13. 17. Zimmerman LE, Burns RP, Wankum G, et al. Trilateral retinoblastoma: ectopic intracranial retinoblastoma associated with bilateral retino blastoma. J Pediatr Ophthalmol Strabismus 1982; 19:320-5. 18. Arrigg PG, Hedges TR Ill, Char DH. Computed tomography in the diagnosis of retinoblastoma. Br J Ophthalmol1983; 67:588-91.
9. Haik BG, Saint Louis L, Smith ME, et al. Magnetic resonance imaging in the evaluation of leukocoria Ophthalmology 1985; 92:1143-52.
19. Katz NNK, Margo CE, Dorwart RH. Computed tomography with his topathologic correlation in children with leukocoria. J Pediatr Ophthal mol Strabismus 1984; 21:50-6. 20. Zimmerman RA, Bilaniuk LT. Computed tomography in the evaluation of patients with bilateral retinoblastomas. J Comput Tomogr 1979; 3: 251-7. 21. Mafee MF, Rasouli F, Spigos DG, et al. Magnetic resonance imaging in the diagnosis of nonsquamous tumors of the head and neck. Oto laryngol Clin North Am 1986; 19:523-36. 22. Gore JC, Emery EW, Orr JS, Doyle FH. Medical nuclear magnetic resonance imaging.!. Physical principles. Invest Radiol1981; 16:269 74. 23. Rosen BR, Brady TJ. Principles of nuclear magnetic resonance for medical application. Semin Nucl Med 1983; 13:308-18. 24. Luyten PR, den Hollander JA. Observation of metabolites in the human brain by MR spectroscopy. Radiology 1986; 161:795-8. 25. Frahm J, Merboldt K, Hanicke W. Localized proton spectroscopy using stimulated echoes. J Magn Resonance. 1987; 72:502-8.
5. 6. 7. 8.
10. Haik BG, Saint Louis L, Smith ME, et al. Computed tomography of the nonrhegmatogenous retinal detachment in the pediatric patient. Ophthalmology 1985; 92:1133-42. 11. Mafee MF, Peyman GA, Peace JH, et al. Magnetic resonance imaging in the evaluation and differentiation of uveal melanoma. Ophthalmology 1987; 94:341-8. 12. Barany M, Spigos DG, Mok E, et al. High resolution proton magnetic resonance spectroscopy of human brain and liver. Magn Resonance Imaging 1987; 5:393-8. 13. Barany M, Langer BG, Glick RP, et al. In vivo H-1 spectroscopy in humans at 1.5T. Radiology 1988; 167:839-44. 14. Malee MF, Puklin J, Barany M, Cohen S, et al. MRI and in vivo proton spectroscopy of lesions of the globe. Semin Ultrasound CT MR 1988; 9:59-71. 15. Mafee MF, Peyman GA, Grisolano J, et al. Malignant uveal melanoma
Discussion by
Barrett G. Haik, MD The diagnosis of retinoblastoma is one of the greatest chal lenges in ophthalmology, because this is one of the few cancers treated without histologic confirmation. To assure appropriate therapy, and to permit maximum ocular salvage and minimum tumor associated mortality, differentiation from a host of benign simulating lesions must be rapidly accomplished. Ophthalmoscopic recognition of retinoblastoma is quite re liable in the classic situation where creamy-pink, single or mul tiple, solid, convex, or pedunculated masses are seen arising from the retina and growing into the vitreous cavity. Associated find ings of intratumor calcification or tumor seeding give additional support to an ophthalmoscopic diagnosis of this disease. Un fortunately, retinoblastoma may assume any of a wide array of appearances depending on the growth pattern and tumor size, making ophthalmoscopic diagnosis much more difficult. There fore, standard ancillary diagnostic studies such as ultrasound and computed tomography (CT) are often critical in establishing a secure diagnosis. The study by Mafee et al was undertaken in order to more clearly define the role of magnetic resonance imaging (MRI} and proton spectroscopy in the evaluation ofpatients with leukocoria. It represents the largest series ofleukocoric patients studied with MRI to date. Their findings support the viewpoint that CT is currently superior to MRI for diagnosis of retinoblastoma, but
From Tulane University School of Medicine, New Orleans. Supported in part by St. Giles Foundation, Brooklyn, New York.
that MRI has significant and dramatic advantages in the detec tion of its extraocular spread. We agree with the authors' findings of CT's superiority in establishing the diagnosis ofretinoblastoma, because it can detect intraocular calcification with greater sensitivity than can MRI. Calcium produces a very weak signal on almost all MRI se quences, because this tissue contains few protons, and these are fixed to the calcium and not easily precessed by radiofrequency stimulus. However, because retinoblastoma is a heterogeneous tumor consisting of soft tissue and calcium in varying propor tions, one can at times see calcium at a hypointense foci within the more hyperintense soft tissue mass. Magnetic resonance imaging may be superior to CT in de tecting subtle scleral invasion with tumors. Sclera, like bone, has a low hydrogen density and, thus, is also hypointense on both T 1- and Tz-accentuated images. Therefore, any disruption of scleral contours by a higher intensity abnormality is evidence of tumor invasion. Although optic nerve, chiasma! and intracranial extension of retinoblastoma may be detected on contrast enhanced CT, these abnormalities are more easily detected with MRI and represent MRI's true value in patient evaluation. Detection is facilitated for two reasons: ( 1) MRI is not subject to beam hardening or volume averaging artifacts associated with dense bone in the orbital apex, optic canal, or chiasmal region, and (2} MRI is exquisitely sensitive to changes in both the hydration and fat content of neural tissue. The diagnosis of pinealoblastoma, as sociated with the hereditary form of retinoblastoma, is similarly accomplished with more ease through MRI.
975
OPHTHALMOLOGY
•
JULY 1989
As this study clearly shows, both CT and MRI play a critical role in the evaluation of patients with leukocoria. While the detection of calcification is a key element in diagnosis, and is best established on CT, it is evident that extraocular spread of the tumor is best detected on MRI. It is thus obvious that both of these diagnostic imaging techniques are powerful adjuncts in the evaluation of patients with leukocoria. The second part of the report presents the results of the first leukocoric eyes studied with in vitro proton magnetic resonance spectroscopy. The very preliminary results from this study suggest that both Coats' disease and retinoblastoma are distinguishable based on spectroscopy. However, as the authors point out, these results should be reviewed with a great deal of caution because this technique is quite complex, only a small number of speci
976
•
VOLUME 96
•
NUMBER 7
mens were studied, and last, there is no established database for comparison. The idea of in vivo magnetic resonance spectroscopy has in trigued physicians since its theoretical conception, because the technique holds the potential to provide the ultimate in diag nostic information-a tumor biochemical profile. It is encour aging to know that this technique is no longer restricted to the laboratory, but is now clinically feasible. I congratulate the authors for their pioneering work and for giving us an insight into the value of magnetic resonance spec troscopy. If in vivo magnetic resonance spectroscopy advances one half as rapidly as magnetic resonance imaging has over the past 5 years, then we are truly on the verge of a noninvasive biopsy.
MAHMOOD F. MAFEE, MD, 1 MORTON F. GOLDBERG, MD/ STEVEN B. COHEN, MD/ 1 EFSTATHIOS D. GOTSIS, PhD, 1 MARC SAFRAN, MD, 2 LAVANYA CHEKURI, 1 BAHRAM RAOFI, MD Abstract: To evaluate the usefulness of magnetic resonance imaging (MRI) in the evaluation of leukocoric eyes, the authors studied 28 patients with either leukocoria or intraocular mass with a 1.5-tesla (T) MRI imager. Retinoblastomas were reliably distinguished from Coats' disease, toxocariasis, and persistent hyperplastic primary vitreous on the basis of MRI findings. Calcification cannot be reliably detected on MRI scans. Lesions elevated less than 4 mm may not be detected reliably by MRI at this time. Computed tomography (CT) can detect calcification with a high degree of accuracy. Retinoblastomas appeared as moderately hyperintense masses on T1- and proton-weighted MRis. They be came hypointense in T2 -weighted MRis. This MRI characteristic is similar to that of uveal melanoma. Intraocular calcification in children especially younger than 3 years of age is highly suggestive of retinoblastoma. In the diagnosis of reti noblastoma, MRI is not as specific as CT because of its lack of sensitivity in detecting calcification. However, MRI, because of its superior contrast resolution, offers more information in the differentiation of pathologic intraocular conditions responsible for leukocoria. The authors also describe their preliminary work of in vitro proton magnetic resonance spectroscopy of eyes with retinoblastoma and an eye with uveal melanoma in an 18-year-old black woman . Ophthalmology
96:965-976, 1989
Originally received: October 10, 1988. Revision accepted: February 24, 1989. 1
Department of Radiology and Magnetic Resonance Center, University of Illinois College of Medicine, Chicago. 2 UIC Eye Center, Eye and Ear Infirmary, University of Illinois College of Medicine, Chicago. Presented at the American Academy of Ophthalmology Annual Meeting, Las Vegas, October 1988.
Supported by core grant 1792 from the National Eye Institute, Bethesda, Maryland, and by an unrestricted research grant from Research to Prevent Blindness, Inc, New York , New York.
Reprint requests to Mahmood F. Malee, MD, 840 South Wood St, Magnetic Resonance Center, M/C 711 , University of Illinois at Chicago, Chicago, IL 60612.
965
OPHTHALMOLOGY
•
JULY 1989
•
VOLUME 96
•
NUMBER 7
F'ig I. Retinoblastoma. A, CT scan shows a mass (arrows) with multiple calcifications (arrowheads). B, T 1-weighted MR1 scan showing a relatively hyperintense mass (arrows). The tumor extends into the optic disc; how ever, the optic nerve is intact as confirmed on Figures IC and D. C, proton-weighted MRI scan showing a relatively hyperintense mass (ar rows). Notice that the calcification seen on Figure lA is faintly seen here as an area of hypointensity. D, Trweighted MRI scan shows that the retinoblastoma is hypointense (arrows). Notice hypointensity of calcifi cation (arrowheads) which is easily detected on CT scan in Figure I A.
During examination of a child with leukocoria, the major diagnostic considerations are retinoblastoma, per sistent hyperplastic primary vitreous (PHPV), Coats' dis ease, retinopathy of prematurity (ROP), congenital cat aract, toxocariasis, chronic retinal detachment associated with retrolental fibrosis, retinal astrocytoma, choroidal hemangioma with previous hemorrhage, and a variety of other nonspecific causes of leukocoria. I-s Retinoblastoma is the most common intraocular ma lignancy of infants and children, occurring in approxi mately 1 in 18,000 live births. 6•7 The detection and clinical differentiation of retinoblastoma from simulating lesions may be difficult. 8 The manner of intraocular and extra ocular extension, patterns of metastasis and recurrence,
966
ocular complications, and associated malignancies in pa tients with retinoblastoma make the diagnosis of retino blastoma one of the most challenging problems of pedi atric ophthalmology and radiology. Although ophthal moscopic recognition of retinoblastoma is often reliable, imaging modalities should be used on all patients sus pected of harboring a retinoblastoma in order to detect any gross retrobulbar spread, intracranial metastasis, and secondary tumors. Of the imaging techniques available, ultrasonography, computed tomography (CT), and mag netic resonance imaging (MRI) are the most useful to incorporate into the clinical examination of patients with leukocoria. 1-4·9 • 10 We have evaluated the use of MRI and CT studies in 27 patients with retinoblastoma and sim
MAFEE et al
•
MAGNETIC RESONANCE IMAGING
Fig 2. Retinoblastoma. A, T 1-weighted oblique sagittal MRI scan shows slight increased intensity of the globe. Notice hyperintensity of subretinal exudate (arrows). B, Trweighted axial MRI scan shows markedly hy pointense mass, filling most of the left globe. The subretinal fluid remains hyperintense (arrows). The tumor calcification appears as very hypoin tense image (arrowhead). Notice that calcification cannot be recognized on T 1-weighted (Fig 2A) image which has been obtained along the optic nerve and which has included the calcification.
ulating lesions to evaluate the incidence of calcification detected by MRI and CT in retinoblastoma, to delineate the diagnostic accuracy of MRI and CT in differentiating retinoblastoma from other simulating lesions, and to de termine the potential accuracy of MRI in diagnosing ret robulbar and intracranial spread. We also describe our preliminary work on in vitro proton magnetic resonance spectroscopy (MRS) of retinoblastoma and an eye with uveal melanoma.
SUBJECTS AND METHODS Our study group consisted of 28 patients: 18 with ret inoblastoma, 3 with Coats' disease, 3 with PHPV (possibly caused by Norrie's disease and Warburg syndrome), 1 with toxocariasis, I with mesoectodermal leiomyoma, 1 with myelinated nerve fibers, and 1 with malignant uveal melanoma. The patients ranged in age from 7 months to
Fig 3. Retinoblastoma. A, T 1-weighted MRI scan shows slight hyper intensity of the entire right globe (curved arrow). Notice normal apical and intracanalicular optic nerve (arrow). B, proton-weighted MRI scan obtained I year after enucleation shows recurrent tumor along optic nerve (hollow arrows) as compared with normal left optic nerve (arrow). Notice involvement of suprasellar cistern (curved arrow) as compared with that of I year ago (Fig 3A). C, proton-weighted coronal MRI scan showing marked tumor in the suprasellar cistern (black arrows). Notice normal internal carotid artery within the cavernous sinus (white arrow).
967
OPHTHALMOLOGY
•
JULY 1989
•
VOLUME 96
•
NUMBER 7
Fig 4. Coats' disease. A, proton-weighted MRI scans show increased intensity of the left globe. B, Tz-weighted MRI scans show hyperintensity of the left globe. Notice the leaves of the detached retina (arrows). The hyperintensity of left globe in the proton-weighted MRI scan is due to subretinal exudate and lipoproteinaceous fluid in the subretinal space.
Fig 5. Toxocariasis. Top. proton-weighted MRI scan shows a hyperintense mass (arrow) in the right globe. Bottom. T 2-weighted scan shows that the mass is hyperintense (isointense to vitreous).
968
18 years. All patients with retinoblastoma had typical ophthalmoscopic findings consistent with retinoblastoma. All the Coats' disease patients except one had typical ophthalmoscopic findings consistent with Coats' disease. Ophthalmoscopic findings in the other patients were con sistent with the respective clinical diagnosis, except in the patient with mesoectodermalleiomyoma who was initially diagnosed to have a ciliary mass of unknown type. The imaging studies were performed, in the majority of cases, using chloral hydrate for sedation (50 mg/kg of body weight) or general anesthesia. Magnetic resonance imaging studies, except in two patients, were performed with a 1.5-tesla (T) Signa unit (General Electric, Milwau kee, WI) with a 3- or 5-mm thick section with 0.6 to 1.5 mm intersection space. Single echo spin-echo (SE) pulse sequences were obtained with a repetition time (TR) of 300 to 800 msec and an echo time (TE) of 20 to 25 msec (TR/TE = 300-800/20-25 msec). Multiple SE pulse se quences were obtained with a TR of 1500 to 2500 msec and aTE of20 to 100 msec. The early echo (20-25 msec) in this sequence yielded proton-weighted (PW) images, and the later echo (30-1 00 msec) yielded T rweighted im ages. 11 The single echo SE sequence yielded T 1-weighted images. 11 The studies were most often performed using head coil with one excitation, a 256 X 256-matrix, and a 20- to 24 cm field of view. We routinely obtain the MRI ofthe eye
MAFEE et al
•
MAGNETIC RESONANCE IMAGING
for 10 weeks Fig 6. Retinopathy of prematurity. A 7-month-old premature boy with a history of fetal respiratory distress necessitating intubation normal. On entirely be to thought was and age of months 2 was he when specialist and hood oxygenation for 2 weeks. He was seen by a retinal in this case were diagnoses differential The identifiable. not were retinas The opacified. were vitreous the of thirds two anterior the examination, globes with bilateral ROP, bilateral PHPV, and autosomal dominant vitreoretinopathy. A, proton-weighted MRI scan shows hyperintensity of the (curved arrow). detachment retinal shows scan MRI -weighted T B, arrow). (curved 2 detachments retinal funnel-shaped and (arrows) masses retrolental to a fibrotic related probably (arrows), mass retrolental Notice The hyperintensity of the globes is related to subretinal fluid and old hemorrhage. hemorrhage. layered recent a to related (arrowhead) hypointensity of area an also scar. Notice
and head. A surface coil 7.6 em in diameter (Medical Advance Corporation, Milwaukee, WI) and an experi mental butterfly-type surface coil (General Electric, Mil waukee, WI) were used with a few patients to improve spatial resolution. For surface coil study, a field view of 12 or 16 em was used. With the surface coil, we use a 3mm section to reduce the problem of partial volume av-
eraging. The thin section has the disadvantage of gener ating less signal (small volume) which may result in a decreased amount of Trweighted information in later echo, particularly in a rapidly decaying T 2-weighted signal. Because of this, and in the light of exaggerated motion artifacts related to inherent increased sensitivity of the surface coil, a retinoblastoma or other intraocular mass
Fig 7. Warburg syndrome. A, T 1-weighted MRI scan shows hyperintense subretinal fluid in both eyes. Notice the detached leaves of the retina (arrowheads). B, T 1-weighted MRI scan obtained 3 months later shows bilateral retinal detachment. Notice fluid level in the right vitreous chamber (arrow). This is believed to be due to serosanguineous fluid in the subhyaloid or subretinal space. Notice a tubular image (curved arrow) suggestive of congenital nonattached retina or Cloquet's canal.
969
OPHTHALMOLOGY
•
JULY 1989
Fig 8. Retinoblastoma. T 1-weighted MRI scan of an excised eye (the entire globe was replaced by tumor) obtained with the CSI magnetic resonance unit demonstrates a hyperintense globe.
may sometimes be better detected on Tz-Weighted images obtained with a 5-mm section and using the head coil instead of a surface coil. In vitro proton MRS and imaging was performed on a horizontal 2-T chemical shift imaging (CSI) unit (Gen eral Electric, Milwaukee, WI). A 6-inch birdcage imaging coil was used for imaging ofthe whole excised eye, whereas a custom-made 8-mm diameter solenoid coil was used for spectroscopy of freshly excised tumors (in the range of 300 mg). In the case of the uveal melanoma, no water suppression was necessary, whereas in the case of the ret inoblastomas water suppression was accomplished by the binomial 1-3-3-1 sequence, with a protocol similar to that reported for in vivo studies. 12- 14 The CT studies were performed with a General Electric 9800 CT scanner. To determine the presence of calcifi cation, 1.5-mm contiguous plain axial (transverse) scans were obtained through the orbits. Because ofthe superior contrast resolution ofMRI scans and to avoid unnecessary radiation as well as to contain costs, we did not perform contrast CT studies of the orbit and head in many of the patients.
RESULTS All retinoblastomas had calcification, which was de tected by CT scans. Calcifications as small as 2 mm were 970
•
VOLUME 96
•
NUMBER 7
detected by CT scanning. Calcifications on MRI scans may be seen as varying degrees of hypointensity in all pulse sequences (Fig 1). In contrast to CT scanning, which is highly specific for calcification, MRI may be nonspecific (Fig 2). Retinoblastomas on T 1-weighted and proton weighted MRI scans appeared as areas of moderately high signal (hyperintense) compared with the signal intensity of ~treous (Figs 1B and C, 2A). On T z-weighted MRis, retmoblastomas appeared as areas of moderately (Fig 1D) to markedly (Fig 2B) low signal (hypointense). The non calcified portion of the retinoblastomas was less hypoin tense than partially calcified retinoblastomas on T z weighted MRis. Tumors elevated less than 4 mm could not be definitely identified on MRI scans. One year after enucleation (with no evidence ofoptic nerve involvement on MRI and histologic examination), marked recurrence of tumor along the optic nerve and with extension into the suprasellar region developed in one patient (Fig 3). The intracranial portion of the tumor was isointense to brain in all pulse sequences. None of the patients with Coats' disease had calcifi cation evident on CT scans. The CT and MRI in one of these patients showed total retinal detachment (Fig 4). The subretinal fluid in this case appeared as a hyperintense image in all pulse sequences, similar to other exudative retinal detachments. The eye was removed because ofad vanced disease complicated by glaucoma, and the diag nosis of Coats' disease was confirmed on histologic ex amination. The CT findings of the toxocara granuloma could not be differentiated from noncalcified retinoblas toma. Magnetic resonance imaging, however, showed that the lesion was hyperintense in T 1- and proton-weighted MRis (Fig 5, top) but unlike retinoblastoma, the lesions appeared hyperintense in Tz-weighted MRis (Fig 5, bot tom). Computed tomographic scanning in patients with PHPV (possibly caused by Norrie's disease), ROP, and Warburg syndrome, showed dense vitreous chambers. Magnetic resonance imaging showed marked hyperin tensity of the affected vitreous chamber on T 1- T z-, and proton-weighted MRis (Fig 6). In one patient with War burg syndrome with bilateral suspected PHPV, the MRI scan showed bilateral retinal detachment with hyperin tensity ofsub retinal fluid (Fig 7A) and layered fluid (acute hemorrhage) on follow-up MRis (Fig 7B). In vitro proton MRI using the CSI unit was performed on two excised unfixed eyes with retinoblastoma. The le sions appeared hyperintense in T 1-weighted (Fig 8, 9A) and hypointense in T z-weighted MRis. (In one of these eyes only, T,-weighted MRI scans were obtained.) Reti noblastomas appeared more hyperintense in T 1-weighted MRis ofin vitro (Fig 9A and for comparison Fig 1B; same patient) than the in vivo T 1- and proton-weighted images. The in vitro magnetic resonance spectroscopy was per formed on one of these eyes. This disclosed several res onances that were tentatively assigned to inositol, lactate, N-acetyl aspartate, creatine, and choline (Fig 9B). In vitro MRI of an excised eye with uveal melanoma was performed on the CSI unit. The lesion appeared as hyperintense in T 1-weighted (Fig lOA) and hypointense
MAFEE et al
•
MAGNETIC RESONANCE IMAGING
10
B
a
6
4
2
0
FPM
Fig 9. Retinoblastoma. A, T 1-weighted MRI scan of excised eye (same patient as in Figure I) obtained with the CSI magnetic resonance unit demonstrates a hyperintense mass (arrows). 8, Water-suppressed in vivo proton magnetic resonance spectrum of 300 mg of tumor tissue of the same retinoblastoma tumor obtained with the CSI unit at 2 T. The spectrum was acquired with an 8-mm diameter solenoid coil which nevertheless was not completely filled with the small tumor sample, resulting thus in some loss of homogeneity and of course spectral resolution. The 1-3-3-1 binomial sequence was used for the water suppression, with an interpulse delay of 2.34 msec, for a total of 256 scans. Tentatively, the composite peak centered around 4.1 parts per million may he attributed to inositol, lactate as well as sugars, the peaks at 3.3 and 3.5 parts per million to inositol while the peak approximately 3 parts per million to creatine and phosphocreatine and choline. The peaks at approximately 2.3 to 2.0 parts per million are unassigned, although the peak near 2.0 parts per million may be due to N-acetyl aspartate.
in T z-weighted MRis (Fig I OB), characteristic of melanotic tumors. 15 The in vitro proton magnetic resonance spec trum of this uveal melanoma is shown in Figure I OC, which shows a resonance at approximately 6.9 parts per million that corresponds to melanin and melanin pre.. cursor. 14
DISCUSSION High-resolution, thin-section CT scanning can detect calcification within retinoblastoma with a high degree of accuracy. 1·4 Calcification may be small and single, large and single, multiple and punctate, or several fine-speckled foci. JA The noncalcified mass of retinoblastoma on CT scans may be very difficult to differentiate from simulating lesions. Computed tomography is of value in accurately assessing the extraocular and intracranial spread of reti noblastomas. Since the first report of the entity known as trilateral retinoblastomas, 16• 17 CT has played an important role in establishing the diagnosis. These ectopic tumors are typically located in the pineal or parasellar region.
Intraocular calcification in children, especially younger than 3 years of age, is highly suggestive of retinoblas toma.18-20 We have seen, however, calcifications in a child younger than 1 year of age and who had bilateral mi crophthalmos with colobomatous cysts. We share our experience with that of others: 1·19 namely, none of the simulating lesions, including PHPV, Coats' disease, toxo cariasis, ROP, chronic retinal detachment, retinal astro cytoma, choroidal osteoma, or optic nerve head drusen tends to contain calcification in the children (:::;;3 years of age) in whom retinoblastoma is usually diagnosed. In children older than 3 years of age, some of the simulating lesions, including retinal astrocytoma, ROP, toxocariasis, and optic nerve head drusen, can produce calcification. Calcifications that are easily detected by CT were difficult to recognize on MRI scans. Dense calcification was seen as an area of hypointensity in T 1-, Tz-, and proton weighted MRis (Fig 1C, D). In some cases, calcifications were not detected on MRI scans. In the diagnosis of ret inoblastoma, MRI is not as specific as CT scanning be cause of its lack of sensitivity in detecting calcification. However, the MRI appearance of retinoblastoma may be 971
OPHTHALMOLOGY
•
JULY 1989
•
VOLUME 96
•
NUMBER 7
Fig 10. Uveal melanoma in an 18-year-old black woman. A, T 1-weighted MRI scan of excised eye obtained with CSI magnetic resonance unit demonstrates a hyperintense mass (arrow). B, the mass (arrow) appears hypointense in Trweighted MRI scan. C, proton spectrum ofthe excised eye showing a large peak at approximately 6.9 parts per million, which may be assigned to aromatic amino acids and melanin. 14•25 The resonance at 4.77 parts per million is from water.
c 10
8
6
4
2
0
PPM
specific enough to differentiate retinoblastoma from sim ulating lesions. In all retinoblastomas (>3 mm), we ob served a mass that displayed MRI characteristics oflesions with relatively short T 1- and short Trrelaxation times. All retinoblastomas were seen as mildly to moderately hyperintense lesions on T 1- and proton-weighted MRI scans (Figs 1B, C; 2A; 3A). They became hypointense in T 2-weighted MRI scans (Figs 1D, 2B). This is very similar to MRI characteristics of uveal melanoma (Fig 11 ). In this study, none of the patients with PHPV or Coats' dis972
ease and one patient with toxocariasis demonstrated sim ilar MRI characteristics to those of retinoblastoma. Both CT and MRI have some advantages and disad vantages. Computed tomography is the study of choice for detection of calcification. Calcification as small as 2 mm can be detected by high-resolution thin section CT scanning. Magnetic resonance imaging has superior con trast resolution to CT scanning (Fig 2B). The phenomenon of partial volume averaging is more problematic in MRI than CT scanning, because with CT it is possible to obtain
MAFEE et al
•
MAGNETIC RESONANCE IMAGING
thinner sections (1-1.5 mm) than MRI (3 mm). In the study of eyes with suspected retinoblastoma, we recom mend plain CT ofthe eyes and MRI of the eyes and head. The optic nerve involvement, recurrence (Fig 3B, C), extracranial and intracranial (Fig 3C) involvement can be better evaluated by MRI than CT scanning. Magnetic resonance imaging appears to offer more information in the differentiation of pathologic intraocular conditions responsible for leukocoria (Figs 4- 7) and other lesions such as mesectodermal leiomyoma (Fig I2). Since the MRI characteristics of retinoblastomas in our study were similar to that of malignant uveal melanomas 11 • 15 (Figs I C, D; II), we decided to study with in vitro spectroscopy these two lesions in order to obtain additional information from their spectral analysis. Our preliminary study showed that the spectral appearance of retinoblastoma was completely different (Fig 9B) from that of uveal melanoma (Fig lOC). The major resonances in retinoblastoma were due to phosphocreatine and N acetyl aspartate. On the other hand, the major resonance in uveal melanoma was due to melanin or its precursors aromatic amino acids (tyrosin, phenylalinine). The N acetyl aspartate is a major resonance recorded on the spectra ofthe brain. 12 •14 The presence ofN-acetyl aspartate in our retinoblastoma specimen (Fig 9B) may be due to the fact that retinoblastoma is derived from primitive em bryonal retinal cells (either neuronal retinal cells or pho toreceptor). Although we cannot reach any definite con clusion from our preliminary in vitro spectroscopic ex perience of these two lesions, however, it is possible that further future study may prove that theN-acetyl aspartate and melanin precursors might be considered as a tumor marker for retinoblastoma and uveal melanoma, respec tively. In order to understand the technical jargon of magnetic resonance spectroscopy, a glimpse at the fundamental of magnetic resonance spectroscopy (MRS) can bring us to understand the importance of MRS and the potential use
Fig 11. Uveal melanoma. Proton-weighted MRI (top) and T2-weighted MRI (bottom) scans show a hyperintense mass (arrows) in proton weighted and hypointense mass in T 2-weighted MRis. Notice hyperin tense subretinal exudate (curved arrows).
of in vivo magnetic resonance spectroscopy in the future practice of combined magnetic resonance imaging and in vivo magnetic resonance spectroscopy. HIGH-RESOLUTION IN VITRO PROTON MAGNETIC RESONANCE SPECTROSCOPY OF HUMAN EYES
Since the discovery of nuclear magnetic resonance phenomenon, magnetic resonance spectroscopy has proved to be one of the most powerful tools in analytical chemistry. Magnetic resonance spectroscopy is a form of
Fig 12. Mesectodermal leiomyoma. An 8-year-old boy with a large tumor of the left ciliary body. Proton-weighted MRI scan (A) and Trweighted MRI scan (B) showing a hyperintense mass (arrows) arising from left ciliary body. Notice that its MRI characteristics are quite different from those of retinoblastoma.
973
OPHTHALMOLOGY
•
JULY 1989
spectroscopy in which, as a result of the magnetic prop erties of certain nuclei arising from their inherent axial spin, radiofrequency radiation is absorbed by the nuclei in the presence of a strong static magnetic field. The mag netic resonance spectrum, in fact, is a plot of signal in tensity versus resonance frequency. The appearance of magnetic resonance spectrum depends on the density of the extranuclear electron and molecular environment surrounding a nucleus. The presence of neighboring atoms and the diamagnetic "shielding currents" that are asso ciated with the distribution of circulating electrons around adjacent atoms slightly modifies the strength of local magnetic fields at a particular nucleus and hence modifies the resonance frequency,2 1-24 so that the details of mo lecular structure change the magnetic resonance frequency ofa nucleus in that molecule. For example, the resonance frequency of a proton within a water molecule differs slightly from that of a proton within a fat molecule, and the resonance frequency of a 31 P nucleus bound in an ATP molecule varies slightly from that of a 31 P nucleus within a phosphocreatine molecule. These small but spe cific displacements or shifts in frequency produce separate absorption peaks in a spectrum (chemical shift), which are "fingerprints" of molecular conformations and there fore directly aid in the determination of chemical struc ture.21·22 The chemical shift is measured and listed in parts per million with respect to a reference signal. Typical ref erence compounds, each of which has only a single res onance line, are water, chloroform, cyclohexane, benzene and tetramethyl silane for proton spectra, perfluorocyclo butane and trifluoroacetic acid for 19F, and phosphoric acid for 31 P. The chemical shift is of paramount importance in an alytical chemistry, because it allows the assignment ofdif ferent signal (lines) in a magnetic resonance spectrum to corresponding substances or molecular groups. Proton magnetic resonance spectra of live human tissues have been performed by several authors 12- 14·24 in which the major resonances were due to water and fat protons. In a recent report, Luyten and den Hollande~4 recorded res onances from N-acetyl aspartate and phosphocreatine in the human brain. Later on, Barany et al 12·13 recorded the spectra of human brain, which showed several resonances that were assigned to metabolites N-acetyl aspartate, glu tamine, phosphocreatine and creatine, choline derivatives, and taurine. They demonstrated the presence of urea, various amino acids, and nucleotides in liver cancer. They suggested that the presence of a large nucleotide pool is a characteristic of tumor tissue involved in rapid nucleic acid (RNA, DNA) synthesisY Figure 10 shows the in vitro MRI and the magnetic resonance spectroscopy of an eye of an 18-year-old black woman with uveal mela noma. The lesion is hyperintense in T 1-weighted and hy pointense in T rweighted MRis, characteristic of mela notic neoplasms reported by an in vivo MRI study. 15 The magnetic resonance spectrum (Fig lOC) shows a major resonance centered around 6.9 parts per million, which may be assigned to melanin or its precursors Tyrosin and phenylalanine. 14·25 The in vitro spectnm of retinoblas 974
•
VOLUME 96
•
NUMBER 7
toma (Fig 9B) was different from that of uveal melanoma (Figs 9B, 1OC), whereas both lesions appear to have similar characteristics on MRI. Our in vitro proton magnetic res onance spectroscopy data have to be considered as pre liminary results. Caution must be exercised in interpre tation ofspectra. Because of our interest in in vivo proton magnetic resonance spectroscopy, we also tried to obtain in vivo water-suppressed spectra from two patients with retinoblastoma and two with Coats' disease. The spectra, however, were less than optimal because of technical problems and therefore were excluded from this com munication. We believe, however, that the potential of in vivo proton spectroscopy to differentiate leukocoria should be investigated. Our previous investigation of in vivo proton spectroscopy has shown that the levels of some metabolites (e.g., melanin) may serve as a marker in the diagnosis ofmelanoma.14 The technique ofin vivo section selective water-suppressed proton spectroscopy reported by several authors 12·13 has some limitations. Recently, a method ofspatially resolved spectroscopy was introduced by Luyten and den Hollande~ 4 to obtain signal from a particular volume only (voxel magnetic resonance spec troscopy). This technique avoids contributions of signal from superficial tissues. The stimulated echo acquisition mode sequence introduced by Frahm and collaborators25 is a new technique that has some advantages over spatially resolved spectroscopy for recording metabolites by proton magnetic resonance spectroscopy in humans. Our pre liminary results with in vitro proton spectroscopy to ob tain biochemical information (Figs 9B, 1OC) is encour aging. Early experience with in vivo proton magnetic res onance spectroscopy has shown its potential for obtaining important biochemical information, 12- 14·24 thus enhancing the diagnostic sensitivity of magnetic resonance studies. Future work on combined MRI and in vivo magnetic resonance voxel spectroscopy is needed, with the goals of characterizing abnormal conditions according to their spectral patterns and of identifying tumor and metabolic markers.
ACKNOWLEDGMENT The authors thank Joel Schulman, MD, for providing three cases.
REFERENCES 1. Char DH, Hedges TR Ill, Norman D. Retinoblastoma: CT diagnosis. Ophthalmology 1984; 91 :1347-50. 2. Goldberg MF, Mafee MF. Computed tomography for diagnosis of persistent hyperplastic primary vitreous (PHPV). Ophthalmology 1983; 90:442-51 . 3. Mafee MF, Goldberg MF, Valvassori GE, Capek V. Computed to· mography in the evaluation of patients with persistent hyperplastic primary vitreous (PHPV). Radiology 1982; 145:713-7. 4. Mafee MF, Goldberg MF, Greenwald MF, et al. Retinoblastoma and
MAFEE et al
•
MAGNETIC RESONANCE IMAGING
simulating lesions: role of CT and MR imaging. Radial Clin North Am 1987; 25:667-82. Abramson DH. Retinoblastoma: diagnosis and management. CA 1982; 32:130-40. Shields JA. Diagnosis and Management of Intraocular Tumors. St Louis: CV Mosby 1983; 437. Devesa SS. The incidence of retinoblastoma. Am J Ophthalmol1975; 80:263-5. Char DH. Current concepts in retinoblastoma. Ann Ophthalmol1980; 12:792-804.
and simulating lesions: MR imaging evaluation. Radiology 1986; 160: 773-80. 16. Bader JL, Meadows AT, Zimmerman LE, et al. Bilateral retinoblastoma with ectopic intracranial retinoblastoma: trilateral retinoblastoma. Can cer Genet Cytogenet 1982; 5:203-13. 17. Zimmerman LE, Burns RP, Wankum G, et al. Trilateral retinoblastoma: ectopic intracranial retinoblastoma associated with bilateral retino blastoma. J Pediatr Ophthalmol Strabismus 1982; 19:320-5. 18. Arrigg PG, Hedges TR Ill, Char DH. Computed tomography in the diagnosis of retinoblastoma. Br J Ophthalmol1983; 67:588-91.
9. Haik BG, Saint Louis L, Smith ME, et al. Magnetic resonance imaging in the evaluation of leukocoria Ophthalmology 1985; 92:1143-52.
19. Katz NNK, Margo CE, Dorwart RH. Computed tomography with his topathologic correlation in children with leukocoria. J Pediatr Ophthal mol Strabismus 1984; 21:50-6. 20. Zimmerman RA, Bilaniuk LT. Computed tomography in the evaluation of patients with bilateral retinoblastomas. J Comput Tomogr 1979; 3: 251-7. 21. Mafee MF, Rasouli F, Spigos DG, et al. Magnetic resonance imaging in the diagnosis of nonsquamous tumors of the head and neck. Oto laryngol Clin North Am 1986; 19:523-36. 22. Gore JC, Emery EW, Orr JS, Doyle FH. Medical nuclear magnetic resonance imaging.!. Physical principles. Invest Radiol1981; 16:269 74. 23. Rosen BR, Brady TJ. Principles of nuclear magnetic resonance for medical application. Semin Nucl Med 1983; 13:308-18. 24. Luyten PR, den Hollander JA. Observation of metabolites in the human brain by MR spectroscopy. Radiology 1986; 161:795-8. 25. Frahm J, Merboldt K, Hanicke W. Localized proton spectroscopy using stimulated echoes. J Magn Resonance. 1987; 72:502-8.
5. 6. 7. 8.
10. Haik BG, Saint Louis L, Smith ME, et al. Computed tomography of the nonrhegmatogenous retinal detachment in the pediatric patient. Ophthalmology 1985; 92:1133-42. 11. Mafee MF, Peyman GA, Peace JH, et al. Magnetic resonance imaging in the evaluation and differentiation of uveal melanoma. Ophthalmology 1987; 94:341-8. 12. Barany M, Spigos DG, Mok E, et al. High resolution proton magnetic resonance spectroscopy of human brain and liver. Magn Resonance Imaging 1987; 5:393-8. 13. Barany M, Langer BG, Glick RP, et al. In vivo H-1 spectroscopy in humans at 1.5T. Radiology 1988; 167:839-44. 14. Malee MF, Puklin J, Barany M, Cohen S, et al. MRI and in vivo proton spectroscopy of lesions of the globe. Semin Ultrasound CT MR 1988; 9:59-71. 15. Mafee MF, Peyman GA, Grisolano J, et al. Malignant uveal melanoma
Discussion by
Barrett G. Haik, MD The diagnosis of retinoblastoma is one of the greatest chal lenges in ophthalmology, because this is one of the few cancers treated without histologic confirmation. To assure appropriate therapy, and to permit maximum ocular salvage and minimum tumor associated mortality, differentiation from a host of benign simulating lesions must be rapidly accomplished. Ophthalmoscopic recognition of retinoblastoma is quite re liable in the classic situation where creamy-pink, single or mul tiple, solid, convex, or pedunculated masses are seen arising from the retina and growing into the vitreous cavity. Associated find ings of intratumor calcification or tumor seeding give additional support to an ophthalmoscopic diagnosis of this disease. Un fortunately, retinoblastoma may assume any of a wide array of appearances depending on the growth pattern and tumor size, making ophthalmoscopic diagnosis much more difficult. There fore, standard ancillary diagnostic studies such as ultrasound and computed tomography (CT) are often critical in establishing a secure diagnosis. The study by Mafee et al was undertaken in order to more clearly define the role of magnetic resonance imaging (MRI} and proton spectroscopy in the evaluation ofpatients with leukocoria. It represents the largest series ofleukocoric patients studied with MRI to date. Their findings support the viewpoint that CT is currently superior to MRI for diagnosis of retinoblastoma, but
From Tulane University School of Medicine, New Orleans. Supported in part by St. Giles Foundation, Brooklyn, New York.
that MRI has significant and dramatic advantages in the detec tion of its extraocular spread. We agree with the authors' findings of CT's superiority in establishing the diagnosis ofretinoblastoma, because it can detect intraocular calcification with greater sensitivity than can MRI. Calcium produces a very weak signal on almost all MRI se quences, because this tissue contains few protons, and these are fixed to the calcium and not easily precessed by radiofrequency stimulus. However, because retinoblastoma is a heterogeneous tumor consisting of soft tissue and calcium in varying propor tions, one can at times see calcium at a hypointense foci within the more hyperintense soft tissue mass. Magnetic resonance imaging may be superior to CT in de tecting subtle scleral invasion with tumors. Sclera, like bone, has a low hydrogen density and, thus, is also hypointense on both T 1- and Tz-accentuated images. Therefore, any disruption of scleral contours by a higher intensity abnormality is evidence of tumor invasion. Although optic nerve, chiasma! and intracranial extension of retinoblastoma may be detected on contrast enhanced CT, these abnormalities are more easily detected with MRI and represent MRI's true value in patient evaluation. Detection is facilitated for two reasons: ( 1) MRI is not subject to beam hardening or volume averaging artifacts associated with dense bone in the orbital apex, optic canal, or chiasmal region, and (2} MRI is exquisitely sensitive to changes in both the hydration and fat content of neural tissue. The diagnosis of pinealoblastoma, as sociated with the hereditary form of retinoblastoma, is similarly accomplished with more ease through MRI.
975
OPHTHALMOLOGY
•
JULY 1989
As this study clearly shows, both CT and MRI play a critical role in the evaluation of patients with leukocoria. While the detection of calcification is a key element in diagnosis, and is best established on CT, it is evident that extraocular spread of the tumor is best detected on MRI. It is thus obvious that both of these diagnostic imaging techniques are powerful adjuncts in the evaluation of patients with leukocoria. The second part of the report presents the results of the first leukocoric eyes studied with in vitro proton magnetic resonance spectroscopy. The very preliminary results from this study suggest that both Coats' disease and retinoblastoma are distinguishable based on spectroscopy. However, as the authors point out, these results should be reviewed with a great deal of caution because this technique is quite complex, only a small number of speci
976
•
VOLUME 96
•
NUMBER 7
mens were studied, and last, there is no established database for comparison. The idea of in vivo magnetic resonance spectroscopy has in trigued physicians since its theoretical conception, because the technique holds the potential to provide the ultimate in diag nostic information-a tumor biochemical profile. It is encour aging to know that this technique is no longer restricted to the laboratory, but is now clinically feasible. I congratulate the authors for their pioneering work and for giving us an insight into the value of magnetic resonance spec troscopy. If in vivo magnetic resonance spectroscopy advances one half as rapidly as magnetic resonance imaging has over the past 5 years, then we are truly on the verge of a noninvasive biopsy.