HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) All Versions of this Article: 2351031301v1 235/1/197    most recent Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by de Graaf, P. Articles by Castelijns, J. A. Search for Related Content PubMed PubMed Citation Articles by de Graaf, P. Articles by Castelijns, J. A.

Retinoblastoma: MR Imaging Parameters in Detection of Tumor Extent¹

¹ From the Departments of Radiology (P.d.G., F.B., J.A.C.), Ophthalmology (A.C.M., S.M.I.), Epidemiology and Statistics (D.L.K.), and Pathology (P.v.d.V.), VU University Medical Center, PO Box 7057, 1007 MB Amsterdam, the Netherlands. Received August 15, 2003; revision requested October 31; final revision received April 20, 2004; accepted June 17. Address correspondence to P.d.G. (e-mail: p.degraaf@vumc.nl).

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
PURPOSE: To assess diagnostic accuracy of preoperatively performed magnetic resonance (MR) imaging for detection of tumor extent in a large patient population with histopathologically proved retinoblastoma.

MATERIALS AND METHODS: Local ethics committee approval and informed consent were not required for retrospective review of patients’ images and records. Fifty-eight eyes in 28 girls (mean age, 21 months; range, 2–59 months) and 28 boys (mean age, 24 months; range, 2–76 months) with retinoblastoma were retrospectively reviewed by one radiologist on unenhanced T1-weighted, dual-echo T2-weighted, and gadolinium-enhanced T1-weighted MR images. MR imaging parameters such as growth pattern, anterior chamber hyperintensity, and involvement of choroid, ciliary body, optic nerve, sclera, orbital fat, and pineal gland were determined. Tumor volume was measured and correlated to metastatic risk factors. Imaging and pathologic findings were compared. Statistical analysis was performed by using logistic regression with log likelihood ratio ² test or Fisher exact test.

RESULTS: Choroidal invasion was suspected with MR imaging in 21 eyes; findings were false-positive in 13 eyes and false-negative in three (73% sensitivity, 72% specificity, 72% accuracy). Anterior chamber hyperintensity on T1-weighted MR images obtained after contrast agent administration correlated well with clinical presence of reactive neovascular processes. MR imaging findings were true-positive in 21 of 32 eyes with proved prelaminar optic nerve invasion (66% sensitivity) and false-positive in one (96% specificity, 79% accuracy). Postlaminar optic nerve invasion was correctly detected in two eyes; in two other eyes, this metastatic risk factor was missed (50% sensitivity, 100% specificity, 97% accuracy). Scleral and extrascleral tumor invasion were correctly excluded in all eyes. Tumor volume was statistically associated with prelaminar optic nerve invasion (P = .001) and choroidal invasion (P = .031).

CONCLUSION: MR imaging is accurate for tumor staging and detection of metastatic risk factors; detection of intraocular tumor infiltration remains difficult. Tumor volume, measured with MR imaging, was associated with prelaminar optic nerve and choroidal involvement.

© RSNA, 2005

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Retinoblastoma, the most common pediatric intraocular neoplasm, is a highly malignant tumor of the primitive neural retina, and it occurs in one in 17 000 live births (1). Because of early diagnosis and improved treatment methods, cure rates greater than 90% have been achieved in western countries. Extraocular extension of retinoblastoma develops in less than 10% of patients (2) and is associated with a considerably higher mortality rate (3–5). Retinoblastoma spreads by direct extension into the orbit along scleral emissary vessels through hematogenous dissemination, invasion of the optic nerve, and dispersion through the cerebrospinal fluid after tumor cells in the optic nerve invade the leptomeninges (6).

The characteristic appearance of retinoblastoma is a unifocal or multifocal elevated white pedunculated mass that is finely vascularized, sometimes with a large feeder vessel and calcified foci caused by tumor necrosis.

The growth pattern of retinoblastoma can be subdivided into three types: endophytic, exophytic, and combined endophytic and exophytic (7). Endophytic tumors arise from inner layers of the retina and grow into the vitreous. Frequently, small clusters of tumor cells detach from the endophytic mass, producing multiple floating tumor islands; this process is known as vitreous seeding. Exophytic tumors start in the outer layers and growth is in the subretinal space, which causes nonrhegmatogenous retinal detachment with subretinal exudate and possible subretinal tumor seeding.

Beside enucleation, conservative treatment methods, such as thermochemotherapy, radioactive plaque therapy, cryotherapy, laser photocoagulation, external-beam radiation therapy, and tumor reduction chemotherapy, can be applied. The treatment of retinoblastoma depends on several parameters: tumor volume and localization, intraocular tumor extension, extraocular stage of disease, and laterality (side) of tumor. In the staging of the extent of disease, magnetic resonance (MR) imaging should include evaluation of intraocular tumor extension (choroid, sclera, prelaminar optic nerve) and extraocular (postlaminar or orbital invasion) or intracranial (pineoblastoma or metastases) tumor spread.

The reported predictors for metastatic retinoblastoma are invasion of the optic nerve (3,8–11), invasion of the choroid (8,9,12), and orbital involvement (3,9). Khelfaoui et al (9) also found that anterior chamber involvement is associated with an increased risk for metastases. Optic nerve invasion is reported to be present more often in patients with exophytic retinoblastoma, tumor thickness of 15 mm or larger, secondary glaucoma, and vitreous hemorrhage (11). A finding of rubeosis iridis should also alert the clinician to suspect optic nerve invasion (13). Factors predictive for choroidal invasion are rubeosis iridis (12,13) and optic nerve invasion (12). There is some controversy about whether exophytic retinoblastoma is a risk factor for choroidal invasion and optic nerve invasion (12,14).

Diagnosis of retinoblastoma is made on the basis of clinical findings at extensive indirect ophthalmoscopy, with the patient receiving a general anesthetic, and at ultrasonography (US). Routinely performed complementary investigations for differential diagnosis and tumor staging are physical examination, examination of blood and cerebrospinal fluid, and MR imaging. Computed tomography is not routinely performed, because cumulative effects of repeated exposure to ionizing radiation could have substantial consequences for patients with hereditary retinoblastoma, such as an increased risk for the development of second primary tumors.

MR imaging is considered to be the best imaging technique for the differential diagnosis between retinoblastoma and pseudoneoplastic lesions, particularly in patients with opaque ocular refractive media that reduce the yield of indirect ophthalmoscopy (15–19). Heavily T2-weighted images, unenhanced T1-weighted MR images, and gadolinium-enhanced fat-suppressed T1-weighted MR images proved to be complementary in the characterization and staging of retinoblastoma (15,17,19–23). Many authors reported the role of MR imaging in retinoblastoma, but still very little is known about the diagnostic accuracy of MR imaging in the staging of the extent of disease in retinoblastoma. Until now, to our knowledge, researchers in two studies evaluated the clinical value of MR imaging findings in comparison with histopathologic results after enucleation (20,23). The numbers of patients included in those studies, however, were relatively small.

The purpose of our study was to assess the diagnostic accuracy of preoperatively performed MR imaging for the detection of tumor extent in a large patient population with histopathologically proved retinoblastoma.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Patient Population
Patients were selected from a series of 120 patients with retinoblastoma who were followed up at our institution between December 1992 and October 2002. Patients eligible for this retrospective study were those with histopathologically proved retinoblastoma, who were primarily treated with enucleation (81 patients) and who underwent pretreatment 1.0- or 1.5-T MR imaging, which enabled adequate image interpretation. Inclusion criteria were met by 64 patients. Clinical data were collected from inpatient and outpatient medical records at our institution and included sex, age, affected eye, follow-up period, adjuvant treatment, and results of ophthalmoscopy and US.

Two patients did not receive intravenous injection of contrast material, and in six patients, quality of histopathologic material hampered proper pathologic analysis. These patients were excluded from the study. Thus, the final study population consisted of 56 patients, with a mean age at enucleation of 22 months (median, 17 months; range, 2–76 months). There were 28 girls (age range, 2–59 months; mean age, 21 months; median age, 16 months) and 28 boys (age range, 2–76 months; mean age, 24 months; median age, 19 months). In two boys, bilateral enucleation had been performed. Thus, data obtained about 58 eyes could be evaluated. Mean time between MR imaging and enucleation was 6 days (median, 5 days; range, 1–17 days). All patients underwent ophthalmoscopy and US after they received a general anesthetic. In all patients, tumor calcification was documented by using US. Our local ethics committee did not require its approval or informed consent for retrospective review of a patient’s records and images.

MR Imaging
MR imaging studies were performed at 1.0 T (Magnetom Impact Expert; Siemens, Erlangen, Germany) and 1.5 T (Magnetom Vision, Magnetom Sonata; Siemens), with use of a standard head coil. Before gadolinium-based contrast agent injection, transverse and sagittal T1-weighted MR images were obtained with repetition time msec/echo time msec of 350–500/14–15 and four acquisitions. Transverse T2-weighted MR images were obtained with the following parameters: 2000–2700/19–60 (intermediate-weighted MR images) and 2000–2700/80–120 (T2-weighted MR images) and one acquisition. After intravenous injection of 0.1 mmol/kg gadopentetate dimeglumine (Magnevist; Schering, Berlin, Germany), sagittal, transverse, and coronal T1-weighted MR images (400–575/13–15, three acquisitions) were obtained with (43 patients) or without (13 patients) fat suppression. Fat suppression was accomplished with a frequency-selective presaturation pulse. Section thickness in all images was 3 mm, with an intersection gap of 0.3 mm. A spin-echo technique was used for all imaging sequences, with a matrix of 256 x 256 and a resolution of 0.8 x 0.8 mm. After injection of a gadolinium-based contrast agent, brain imaging was performed with 5-mm-thick T1-weighted MR images obtained with 500–650/14–15 and two acquisitions. During MR imaging, children were sedated with a mixture of 1 mg/kg of promethazine hydrochloride (Centrafarm Services, Etten-Leur, the Netherlands) injection, 1 mg/kg of pethidine (Pharmachemie, Haarlem, the Netherlands), and 0.25 mg/kg of droperidol (Janssen-Cilag, Tilburg, the Netherlands), after premedication with 0.4 mg/kg of diazepam (Stesolid; Alpharma, Baarn, the Netherlands).

Image Interpretation
The MR images were evaluated by an observer (J.A.C.) who had 11 years of experience with orbital and brain MR imaging and who was blinded to histopathologic findings. After injection of a gadolinium-based contrast agent at T1- and T2-weighted MR imaging, evaluation was conducted of the following features: growth pattern (endophytic, exophytic, or a combination of the two); anterior chamber hyperintensity; and involvement of the choroid (tumor extension or inflammation), ciliary body, optic nerve (prelaminar, postlaminar, or no invasion), sclera, orbital fat (tumor extension or inflammation), and pineal gland (trilateral retinoblastoma). T2-weighted MR images were also used to assess vitreous hemorrhage and retinal detachment. These images, in combination with T1-weighted images obtained before gadolinium-based contrast agent injection, were used to evaluate subretinal fluid. Subretinal fluid was classified as follows: category 1, fluid with a sedimentation effect (gravitational fluid-fluid level) classified as subretinal hemorrhage; category 2, hyperintense fluid on precontrast T1- and T2-weighted MR images classified as subretinal hemorrhage or proteinaceous subretinal effusion; and category 3, isointense fluid on precontrast T1-weighted images and hyperintense fluid on T2-weighted images classified as subretinal effusion with low protein content.

Signal intensity (SI) of the ipsilateral and contralateral vitreous was measured in 26 patients on pre- and postcontrast T1-, intermediate-, and T2-weighted MR images. Excluded from SI measurement were patients with bilateral disease, vitreous hemorrhage (clinical, MR imaging, or histopathologic findings), subtotal or total retinal detachment, or a tumor of large size. For each patient, one representative level was selected to measure SI of the ipsilateral and contralateral vitreous with all sequences by using the region-of-interest function of the MR imaging system and custom-designed dedicated software. Regions of interest (approximately 15 mm²) were placed in the section in which ipsilateral and contralateral vitreous had the largest anteroposterior diameter. The proportional difference in SI between ipsilateral and contralateral vitreous was calculated. The transverse relaxation times were estimated from the ratio of the SI on two echoes (echo time 1 [TE1] and echo time 2 [TE2], 20–80 msec and 60–120 msec, respectively) from the equation: T2 (TE2 – TE1)/ln[SI(TE1)] – ln[SI(TE2)] (20). Tumor location was coded as anterior or posterior to the equator. The latter was subcategorized into macular, juxtapapillar, and posterior to the equator but not macular or juxtapapillar. Volume of tumor was measured on transverse postcontrast T1-weighted MR images in 56 enucleated eyes. In two patients with bilateral enucleation, only data from the largest tumor volume were included for consideration. In eyes with multifocality, only the tumor with the largest volume was included. In every section in which tumor was present, this structure was manually surrounded by a region of interest, and the surface of this region was calculated. All these surfaces of tumor in different sections were added and multiplied by section thickness (3 mm) and intersection gap (0.3 mm). On the basis of median tumor volume, tumor volumes were categorized into four subgroups: very small, less than 0.65 cm³; small, 0.65–1.30 cm³; medium, greater than 1.30–1.95 cm³; and large, greater than 1.95 cm³.

Anterior chamber hyperintensity on postcontrast T1-weighted MR images, vitreous hemorrhage, and vitreous tumor seeding were compared with ophthalmoscopic findings obtained with the patient receiving a sedative. An MR imaging finding of retinal detachment was compared with clinical findings (ophthalmoscopic and US findings), because of the large number of artifacts caused by histopathologic preparation of enucleated eyes. Follow-up findings were chosen to be the standard for pineal gland lesions.

Pathologic Analysis
Pathologic analysis was performed by an observer (P.v.d.V., a pathologist with 9 years of experience in ophthalmopathology, who diagnoses virtually all retinoblastoma cases in the Netherlands, as the VU University Medical Center is the center of treatment for this tumor) who was blinded to MR imaging findings. After surgery, enucleated eyes were fixed in 4% formaldehyde for 24 hours, washed with both 70% and 99% alcohol for 24 hours, and embedded in paraffin. Whole-eye sections were cut parallel to the optic nerve with a thickness of 4–6 µm (depending on the amount of tumor calcification) and stained with hematoxylin-eosin. Growth pattern and tumor extension into the choroid, ciliary body, optic nerve, sclera, and orbital fat were assessed.

Tumor extension into the optic nerve was categorized with reference to the cribriform lamina of the sclera as pre- or postlaminar invasion. Also, presence or absence of tumor cells in the vitreous, choroidal inflammation, and orbital inflammation were assessed. Histopathologic data were used as the standard for corresponding MR imaging features. After first evaluation of MR imaging and histopathologic findings, a second review (J.A.C.) of false-positive and false-negative MR imaging findings was performed to evaluate structural causes of disagreement.

Statistical Analysis
Statistical comparisons of SIs were performed by using a nonparametric Mann-Whitney U test for comparing differences in two distributions. The Mann-Whitney U test was selected because of the nonnormality of the distributions, together with a small data set and large differences in standard deviations of the groups. For comparisons, one-tailed tests were used with statistical software (SPSS, version 9.0; SPSS, Chicago, Ill). Accuracy of MR imaging findings was calculated as the percentage of true-positive plus true-negative findings compared with the total number of examined patients. Sensitivity was calculated as a percentage by dividing true-positive findings by the sum of true-positive and false-negative findings. Specificity was calculated as a percentage by dividing true-negative findings by the sum of true-negative and false-positive findings. Logistic regression analysis was performed by using a log likelihood ratio ² test to investigate the association between tumor volume and the dichotomous dependent variables of choroidal and prelaminar optic nerve invasion. The odds ratio was estimated for each independent variable. The Fisher exact test was computed for the association between tumor volume and postlaminar tumor invasion because of the low prevalence of abnormal findings. For significance of differences in distribution of variables, a difference with a P value of .05 was considered statistically significant.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Patient Characteristics
Eighteen (32%) patients, nine boys and nine girls, had bilateral disease; both eyes were enucleated in two of them (n = 20 eyes). Mean age at enucleation was 15 months (median age, 10 months; range, 2–38 months). In patients with unilateral disease, the right eye was involved in 20 patients, 11 boys and nine girls (mean age, 23 months; median age, 18 months; range, 3–76 months); the left eye was involved in 18 patients, eight boys and 10 girls (mean age, 30 months; median age, 28 months; range, 7–65 months) (n = 38 eyes). Histopathologic confirmation was obtained in all 58 eyes. Postoperative adjuvant chemotherapy was applied in six (11%) patients, because of the higher risk of metastases. Mean follow-up time after MR imaging was 51 months (median, 49 months; range, 3–121 months). One patient died because of metastatic disease, without a pineal gland tumor.

MR Imaging Characteristics of Retinoblastoma and Vitreous Tumor Seeding
In all retinoblastomas, MR imaging depicted lesions with characteristic SI: hyperintense with respect to the vitreous on T1-weighted MR images and hypointense on T2-weighted MR images. After intravenous administration of gadopentetate dimeglumine, all tumors showed enhancement.

Mean proportion of ipsilateral to contralateral T2 relaxation times was estimated to be 0.93 ± 0.09 (standard deviation) for vitreous without ipsilateral tumor seeding and 0.90 ± 0.15 for vitreous with ipsilateral tumor seeding (P = .32, Mann-Whitney U test). In eyes with vitreous tumor seeding compared with eyes without tumor seeding, no statistically significant differences were present in proportions of SI between ipsilateral and contralateral vitreous on pre- and post-contrast T1-, T2-, and intermediate-weighted MR images (Table 1).

Vitreous hemorrhage was clinically present in two (4%) of 55 eyes and could not be judged in three. At MR imaging, vitreous hemorrhage was present in three (6%) of 50, and no residual vitreous was left in eight eyes because of the large size of the tumor (four eyes) or total retinal detachment (four eyes). Both clinical and MR imaging evaluations were not possible in one eye. In the remaining 48 eyes, findings in 47 were true-negative and those in one were false-positive (Table 2).

Choroid and Ciliary Body
Choroidal invasion was detected histopathologically in 11 (19%) eyes; eight of those were also rated as abnormal at MR imaging (Table 2). Inhomogeneous contrast enhancement and local thickening of the choroid adjacent to the tumor on T1-weighted gadolinium-enhanced MR images was judged as a possible infiltration (Fig 1). A large number of false-positive findings (13 eyes) were reported, leading to a sensitivity of 73% and a specificity of 72% (Table 2). In all false-positive findings, little normal retinal tissue was left between tumor and choroid, which decreased contrast between two enhancing structures at MR imaging (Fig 2). Only in three eyes was choroidal invasion missed: two eyes showed microscopic invasion and one showed invasion to a larger extent (Fig 3). In three eyes, the choroid was seen as a diffuse thickened hyperintense ring-shaped structure at MR imaging with all sequences. Histopathologic sections showed marked invasion of the choroid by inflammatory cells (Fig 4). In one eye, choroidal inflammation (with a ring-shaped enhancement pattern) penetrated the optic nerve more than 1 mm on postcontrast T1-weighted MR images, which could easily be interpreted as tumor invasion (Fig 5a). At histologic investigation, a granuloma in the optic nerve head was found (Fig 5b).

View larger version (152K): [in this window] [in a new window] [Download PPT slide]   Figure 1. Bilateral retinoblastoma in 2-year-old boy. Transverse T1-weighted gadolinium-enhanced MR image (370/14) of left eye shows inhomogeneous enhancement pattern and local thickening of the choroid, adjacent to the tumor, and these findings were suspicious for extensive tumor invasion (arrows). Histopathologic examination results confirmed these findings.

View larger version (130K): [in this window] [in a new window] [Download PPT slide]   Figure 2a. Retinoblastoma in 18-month-old boy. (a) Transverse gadolinium-enhanced T1-weighted MR image (500/15) of right eye shows retinoblastoma and choroid with similar SI. Choroidal invasion could not be excluded with MR imaging, because tumor and choroid showed similar SI. Notice asymmetric enhancement of the iris (arrow), which was clinically confirmed by the presence of rubeosis iridis. (b) Histopathologic examination did not show invasion of the choroid (C) by retinoblastoma (R). Choroid is not affected, because retinal pigment epithelium layer (arrowheads) is intact. (Hematoxylin-eosin stain; original magnification, x20 objective.)

View larger version (145K): [in this window] [in a new window] [Download PPT slide]   Figure 2b. Retinoblastoma in 18-month-old boy. (a) Transverse gadolinium-enhanced T1-weighted MR image (500/15) of right eye shows retinoblastoma and choroid with similar SI. Choroidal invasion could not be excluded with MR imaging, because tumor and choroid showed similar SI. Notice asymmetric enhancement of the iris (arrow), which was clinically confirmed by the presence of rubeosis iridis. (b) Histopathologic examination did not show invasion of the choroid (C) by retinoblastoma (R). Choroid is not affected, because retinal pigment epithelium layer (arrowheads) is intact. (Hematoxylin-eosin stain; original magnification, x20 objective.)

View larger version (133K): [in this window] [in a new window] [Download PPT slide]   Figure 3a. Retinoblastoma in 3-year-old boy. (a) Transverse gadolinium-enhanced T1-weighted fat-suppressed MR image (575/15) of right eye shows tumor mass (arrowheads) surrounded by hyperintense subacute subretinal bleeding. Notice local thickening of choroid (black arrow) lateral to the optic nerve caused by massive tumor invasion. Increased SI (white arrow) posterior to the eyeball is present and was suspicious for postlaminar tumor invasion. (b) Transverse T2-weighted MR image (2200/120) shows characteristic hypointense retinoblastoma surrounded by hyperintense subretinal fluid. (c) Histopathologic specimen shows massive invasion of retinoblastoma (R) into the choroid (C). Sclera (S) is not affected. (Hematoxylin-eosin; original magnification, x10 objective.) (d) Histopathologic specimen shows same findings as in c and postlaminar optic nerve (O) invasion (arrow). (Hematoxylin-eosin; original magnification, x20 objective.)

View larger version (120K): [in this window] [in a new window] [Download PPT slide]   Figure 3b. Retinoblastoma in 3-year-old boy. (a) Transverse gadolinium-enhanced T1-weighted fat-suppressed MR image (575/15) of right eye shows tumor mass (arrowheads) surrounded by hyperintense subacute subretinal bleeding. Notice local thickening of choroid (black arrow) lateral to the optic nerve caused by massive tumor invasion. Increased SI (white arrow) posterior to the eyeball is present and was suspicious for postlaminar tumor invasion. (b) Transverse T2-weighted MR image (2200/120) shows characteristic hypointense retinoblastoma surrounded by hyperintense subretinal fluid. (c) Histopathologic specimen shows massive invasion of retinoblastoma (R) into the choroid (C). Sclera (S) is not affected. (Hematoxylin-eosin; original magnification, x10 objective.) (d) Histopathologic specimen shows same findings as in c and postlaminar optic nerve (O) invasion (arrow). (Hematoxylin-eosin; original magnification, x20 objective.)

View larger version (137K): [in this window] [in a new window] [Download PPT slide]   Figure 3c. Retinoblastoma in 3-year-old boy. (a) Transverse gadolinium-enhanced T1-weighted fat-suppressed MR image (575/15) of right eye shows tumor mass (arrowheads) surrounded by hyperintense subacute subretinal bleeding. Notice local thickening of choroid (black arrow) lateral to the optic nerve caused by massive tumor invasion. Increased SI (white arrow) posterior to the eyeball is present and was suspicious for postlaminar tumor invasion. (b) Transverse T2-weighted MR image (2200/120) shows characteristic hypointense retinoblastoma surrounded by hyperintense subretinal fluid. (c) Histopathologic specimen shows massive invasion of retinoblastoma (R) into the choroid (C). Sclera (S) is not affected. (Hematoxylin-eosin; original magnification, x10 objective.) (d) Histopathologic specimen shows same findings as in c and postlaminar optic nerve (O) invasion (arrow). (Hematoxylin-eosin; original magnification, x20 objective.)

View larger version (137K): [in this window] [in a new window] [Download PPT slide]   Figure 3d. Retinoblastoma in 3-year-old boy. (a) Transverse gadolinium-enhanced T1-weighted fat-suppressed MR image (575/15) of right eye shows tumor mass (arrowheads) surrounded by hyperintense subacute subretinal bleeding. Notice local thickening of choroid (black arrow) lateral to the optic nerve caused by massive tumor invasion. Increased SI (white arrow) posterior to the eyeball is present and was suspicious for postlaminar tumor invasion. (b) Transverse T2-weighted MR image (2200/120) shows characteristic hypointense retinoblastoma surrounded by hyperintense subretinal fluid. (c) Histopathologic specimen shows massive invasion of retinoblastoma (R) into the choroid (C). Sclera (S) is not affected. (Hematoxylin-eosin; original magnification, x10 objective.) (d) Histopathologic specimen shows same findings as in c and postlaminar optic nerve (O) invasion (arrow). (Hematoxylin-eosin; original magnification, x20 objective.)

View larger version (130K): [in this window] [in a new window] [Download PPT slide]   Figure 4a. Bilateral retinoblastoma in 10-month-old boy. (a) Transverse T1-weighted MR image (480/15) of left eye obtained after gadolinium-based contrast agent administration shows hyperintense enhancing ring, which appears to be a diffuse thickened choroid, probably related to inflammation secondary to an intraocular tumor (arrows). (b) Microscopic specimen shows a detail of optic nerve (O), choroid (C), and sclera (S). Inflammatory cells, present in choroid, caused MR image enhancement. (Hematoxylin-eosin stain; original magnification, x20 objective.)

View larger version (125K): [in this window] [in a new window] [Download PPT slide]   Figure 4b. Bilateral retinoblastoma in 10-month-old boy. (a) Transverse T1-weighted MR image (480/15) of left eye obtained after gadolinium-based contrast agent administration shows hyperintense enhancing ring, which appears to be a diffuse thickened choroid, probably related to inflammation secondary to an intraocular tumor (arrows). (b) Microscopic specimen shows a detail of optic nerve (O), choroid (C), and sclera (S). Inflammatory cells, present in choroid, caused MR image enhancement. (Hematoxylin-eosin stain; original magnification, x20 objective.)

View larger version (132K): [in this window] [in a new window] [Download PPT slide]   Figure 5a. Bilateral retinoblastoma in 2-month-old girl. (a) Transverse T1-weighted fat-suppressed MR image (541/15) of right eye obtained after administration of gadolinium-based contrast agent shows strong enhancement and diffuse thickening of the choroid, probably related to choroidal inflammation. This enhancement extends into the optic nerve (arrow) and was suspicious for postlaminar tumor invasion. (b) Microscopic specimen shows retinoblastoma (R), with massive necrosis, sclera (S), and inflammation of choroid (C). A granuloma (arrow) was detected in the optic nerve head and was responsible for contrast enhancement pattern on MR image. No tumor invasion into optic nerve (O), choroid, or sclera was present. (Hematoxylin-eosin stain; original magnification, x5 objective.)

View larger version (129K): [in this window] [in a new window] [Download PPT slide]   Figure 5b. Bilateral retinoblastoma in 2-month-old girl. (a) Transverse T1-weighted fat-suppressed MR image (541/15) of right eye obtained after administration of gadolinium-based contrast agent shows strong enhancement and diffuse thickening of the choroid, probably related to choroidal inflammation. This enhancement extends into the optic nerve (arrow) and was suspicious for postlaminar tumor invasion. (b) Microscopic specimen shows retinoblastoma (R), with massive necrosis, sclera (S), and inflammation of choroid (C). A granuloma (arrow) was detected in the optic nerve head and was responsible for contrast enhancement pattern on MR image. No tumor invasion into optic nerve (O), choroid, or sclera was present. (Hematoxylin-eosin stain; original magnification, x5 objective.)

View larger version (129K): [in this window] [in a new window] [Download PPT slide]   Figure 6a. Unilateral retinoblastoma in 20-month-old girl. (a) Transverse T1-weighted MR image (575/15) of right eye obtained after gadolinium-based contrast agent administration shows large tumor mass with inhomogeneous enhancement. Tumor invasion of the ciliary body (arrows) was suspected but was a false-positive finding, probably because of insufficient spatial resolution. The iris manifests as a hyperintense band anterior to the lens as a result of rubeosis iridis detected with ophthalmoscopic examination. (b) Histopathologic specimen shows a detail of the ciliary body (CB) (arrows) without tumor infiltration. Tumor (R) is adjacent to the ciliary body, anterior chamber, and iris (I). (Hematoxylin-eosin stain; original magnification, x10 objective.)

View larger version (134K): [in this window] [in a new window] [Download PPT slide]   Figure 6b. Unilateral retinoblastoma in 20-month-old girl. (a) Transverse T1-weighted MR image (575/15) of right eye obtained after gadolinium-based contrast agent administration shows large tumor mass with inhomogeneous enhancement. Tumor invasion of the ciliary body (arrows) was suspected but was a false-positive finding, probably because of insufficient spatial resolution. The iris manifests as a hyperintense band anterior to the lens as a result of rubeosis iridis detected with ophthalmoscopic examination. (b) Histopathologic specimen shows a detail of the ciliary body (CB) (arrows) without tumor infiltration. Tumor (R) is adjacent to the ciliary body, anterior chamber, and iris (I). (Hematoxylin-eosin stain; original magnification, x10 objective.)

View larger version (99K): [in this window] [in a new window] [Download PPT slide]   Figure 7a. Unilateral retinoblastoma in 12-month-old boy. (a) Transverse gadolinium-enhanced T1-weighted fat-suppressed MR image (475/14) shows an enhancing tumor mass (arrowheads) of right eye. Notice extreme enhancement of the iris (arrow) compared with appearance of left eye. Clinical examination revealed rubeosis iridis of the right eye. (b) Histopathologic specimen shows a detail of the anterior chamber (AC), cornea (CO), and lateral part of the iris (I). Marked increase of blood vessels is present in the iris (arrowheads). (Hematoxylin-eosin stain; original magnification, x20 objective.)

View larger version (72K): [in this window] [in a new window] [Download PPT slide]   Figure 7b. Unilateral retinoblastoma in 12-month-old boy. (a) Transverse gadolinium-enhanced T1-weighted fat-suppressed MR image (475/14) shows an enhancing tumor mass (arrowheads) of right eye. Notice extreme enhancement of the iris (arrow) compared with appearance of left eye. Clinical examination revealed rubeosis iridis of the right eye. (b) Histopathologic specimen shows a detail of the anterior chamber (AC), cornea (CO), and lateral part of the iris (I). Marked increase of blood vessels is present in the iris (arrowheads). (Hematoxylin-eosin stain; original magnification, x20 objective.)

View larger version (145K): [in this window] [in a new window] [Download PPT slide]   Figure 8a. Unilateral retinoblastoma in 12-month-old boy. (a) Transverse gadolinium-enhanced T1-weighted MR image (475/14) of right eye shows prelaminar optic nerve invasion (arrow) as an interruption of linear enhancement pattern of the choroidoretinal complex at the optic disc. (This finding was rated as true-positive.) (b) Histopathologic specimen shows retinoblastoma (R) that infiltrated optic nerve (O) up to lamina cribrosa (LC), which indicated prelaminar invasion. (Hematoxylin-eosin stain; original magnification, x5 objective.)

View larger version (153K): [in this window] [in a new window] [Download PPT slide]   Figure 8b. Unilateral retinoblastoma in 12-month-old boy. (a) Transverse gadolinium-enhanced T1-weighted MR image (475/14) of right eye shows prelaminar optic nerve invasion (arrow) as an interruption of linear enhancement pattern of the choroidoretinal complex at the optic disc. (This finding was rated as true-positive.) (b) Histopathologic specimen shows retinoblastoma (R) that infiltrated optic nerve (O) up to lamina cribrosa (LC), which indicated prelaminar invasion. (Hematoxylin-eosin stain; original magnification, x5 objective.)

View larger version (141K): [in this window] [in a new window] [Download PPT slide]   Figure 9a. Unilateral retinoblastoma in 20-month-old girl. (a) Transverse gadolinium-enhanced T1-weighted MR image (575/15) of right eye shows large tumor mass associated with prelaminar optic nerve invasion (arrow), which was not suspected. Notice absence of bulk of tumor mass near the optic disc. (b) Histopathologic specimen depicts optic disc (O) invaded by retinoblastoma (R). (Hematoxylin-eosin stain; original magnification, x10 objective.)

View larger version (143K): [in this window] [in a new window] [Download PPT slide]   Figure 9b. Unilateral retinoblastoma in 20-month-old girl. (a) Transverse gadolinium-enhanced T1-weighted MR image (575/15) of right eye shows large tumor mass associated with prelaminar optic nerve invasion (arrow), which was not suspected. Notice absence of bulk of tumor mass near the optic disc. (b) Histopathologic specimen depicts optic disc (O) invaded by retinoblastoma (R). (Hematoxylin-eosin stain; original magnification, x10 objective.)

View larger version (128K): [in this window] [in a new window] [Download PPT slide]   Figure 10a. Bilateral retinoblastoma in 15-month-old girl. (a) Transverse T1-weighted MR image (375/14) of left eye obtained after gadolinium-based contrast agent administration shows retinoblastoma at posterior part of the eyeball with normal thickness of underlying optic nerve disc and linear enhancement pattern of choroidoretinal complex (arrow) and choroid. Both prelaminar and postlaminar tumor invasion of the optic nerve were not suspected. (b) Histopathologic specimen of optic nerve (O) revealed postlaminar tumor (R) invasion (arrow). (Hematoxylin-eosin stain; original magnification, x10 objective.)

View larger version (131K): [in this window] [in a new window] [Download PPT slide]   Figure 10b. Bilateral retinoblastoma in 15-month-old girl. (a) Transverse T1-weighted MR image (375/14) of left eye obtained after gadolinium-based contrast agent administration shows retinoblastoma at posterior part of the eyeball with normal thickness of underlying optic nerve disc and linear enhancement pattern of choroidoretinal complex (arrow) and choroid. Both prelaminar and postlaminar tumor invasion of the optic nerve were not suspected. (b) Histopathologic specimen of optic nerve (O) revealed postlaminar tumor (R) invasion (arrow). (Hematoxylin-eosin stain; original magnification, x10 objective.)

View larger version (135K): [in this window] [in a new window] [Download PPT slide]   Figure 11a. Unilateral retinoblastoma in 26-month-old girl. (a) Transverse gadolinium-enhanced T1-weighted MR image (575/15) of left eye. Normal thickness of the optic disc is seen with linear enhancement pattern (arrow), a finding that was not suspicious for invasion of this structure. Enhancement of choroid (arrowheads) adjacent to the tumor does not indicate tumor infiltration. (b) Macroscopic histopathologic specimen shows retinoblastoma (R) and confirms absence of optic nerve (O) and choroidal (c) invasion. Retinal detachment (*) is an artifact caused by histopathologic preparation. (Hematoxylin-eosin stain; original magnification, x3.5 objective.)

View larger version (138K): [in this window] [in a new window] [Download PPT slide]   Figure 11b. Unilateral retinoblastoma in 26-month-old girl. (a) Transverse gadolinium-enhanced T1-weighted MR image (575/15) of left eye. Normal thickness of the optic disc is seen with linear enhancement pattern (arrow), a finding that was not suspicious for invasion of this structure. Enhancement of choroid (arrowheads) adjacent to the tumor does not indicate tumor infiltration. (b) Macroscopic histopathologic specimen shows retinoblastoma (R) and confirms absence of optic nerve (O) and choroidal (c) invasion. Retinal detachment (*) is an artifact caused by histopathologic preparation. (Hematoxylin-eosin stain; original magnification, x3.5 objective.)

View larger version (159K): [in this window] [in a new window] [Download PPT slide]   Figure 12. Unilateral retinoblastoma in 17-month-old girl. Transverse gadolinium-enhanced T1-weighted MR image (575/15) of left eye. Temporal part of the eye shows enhancing retinoblastoma (hyperintense signal), combined with subretinal hemorrhage (temporal and nasal). Choroidoretinal complex shows linear enhancement pattern (arrowhead), which indicates absence of prelaminar invasion. Normal thickness of optic disc is seen. Histopathologic examination results confirmed these findings.

View larger version (94K): [in this window] [in a new window] [Download PPT slide]   Figure 13a. Unilateral retinoblastoma in 14-month-old boy. (a) Transverse gadolinium-enhanced T1-weighted fat-suppressed MR image (575/15) of eyes. Contrast between tumor and vitreous is diminished because of subretinal blood. Increased SI of the postlaminar optic nerve in continuity with tumor was suspicious for postlaminar tumor invasion (long arrow). Notice shallowness of anterior chamber combined with SI increase of the iris (short arrow). Rubeosis iridis was detected with ophthalmoscopic examination. Shallowness is probably related to presence of neovascular glaucoma combined with elevated intraocular pressure, which is an indication for enucleation without delay. (b) Transverse T2-weighted MR image (2200/120) shows total retinal detachment (black arrow) with subretinal hypointense tumor mass (white arrow). Characteristic blood-fluid level (arrowheads) indicates acute subretinal hemorrhage. (c) Histopathologic specimen shows retinoblastoma (R) with postlaminar optic nerve invasion, detachment of retina (arrowheads), and shallow anterior chamber (AC). Choroid (C) and sclera (S) were not invaded by tumor. (Hematoxylin-eosin stain; original magnification, x3.5 objective.) Optic nerve (O) is also involved. Inset: Detail of optic nerve shows postlaminar tumor invasion (arrows). (Original magnification, x10 objective.)

View larger version (95K): [in this window] [in a new window] [Download PPT slide]   Figure 13b. Unilateral retinoblastoma in 14-month-old boy. (a) Transverse gadolinium-enhanced T1-weighted fat-suppressed MR image (575/15) of eyes. Contrast between tumor and vitreous is diminished because of subretinal blood. Increased SI of the postlaminar optic nerve in continuity with tumor was suspicious for postlaminar tumor invasion (long arrow). Notice shallowness of anterior chamber combined with SI increase of the iris (short arrow). Rubeosis iridis was detected with ophthalmoscopic examination. Shallowness is probably related to presence of neovascular glaucoma combined with elevated intraocular pressure, which is an indication for enucleation without delay. (b) Transverse T2-weighted MR image (2200/120) shows total retinal detachment (black arrow) with subretinal hypointense tumor mass (white arrow). Characteristic blood-fluid level (arrowheads) indicates acute subretinal hemorrhage. (c) Histopathologic specimen shows retinoblastoma (R) with postlaminar optic nerve invasion, detachment of retina (arrowheads), and shallow anterior chamber (AC). Choroid (C) and sclera (S) were not invaded by tumor. (Hematoxylin-eosin stain; original magnification, x3.5 objective.) Optic nerve (O) is also involved. Inset: Detail of optic nerve shows postlaminar tumor invasion (arrows). (Original magnification, x10 objective.)

View larger version (149K): [in this window] [in a new window] [Download PPT slide]   Figure 13c. Unilateral retinoblastoma in 14-month-old boy. (a) Transverse gadolinium-enhanced T1-weighted fat-suppressed MR image (575/15) of eyes. Contrast between tumor and vitreous is diminished because of subretinal blood. Increased SI of the postlaminar optic nerve in continuity with tumor was suspicious for postlaminar tumor invasion (long arrow). Notice shallowness of anterior chamber combined with SI increase of the iris (short arrow). Rubeosis iridis was detected with ophthalmoscopic examination. Shallowness is probably related to presence of neovascular glaucoma combined with elevated intraocular pressure, which is an indication for enucleation without delay. (b) Transverse T2-weighted MR image (2200/120) shows total retinal detachment (black arrow) with subretinal hypointense tumor mass (white arrow). Characteristic blood-fluid level (arrowheads) indicates acute subretinal hemorrhage. (c) Histopathologic specimen shows retinoblastoma (R) with postlaminar optic nerve invasion, detachment of retina (arrowheads), and shallow anterior chamber (AC). Choroid (C) and sclera (S) were not invaded by tumor. (Hematoxylin-eosin stain; original magnification, x3.5 objective.) Optic nerve (O) is also involved. Inset: Detail of optic nerve shows postlaminar tumor invasion (arrows). (Original magnification, x10 objective.)

Tumor Volume and Its Correlations
Mean primary tumor volume in 56 eyes was 1.33 cm³ (median, 1.30 cm³; range, 0.04–2.88 cm³). In eight (14%) eyes, mean primary tumor volume was very small; in 20 (36%), it was small; in 19 (34%), it was medium; and in nine (16%), it was large.

The number of eyes with prelaminar invasion increased with increasing tumor volume. Very small tumors showed 25% prelaminar invasion. Small, medium, and large tumors showed an invasion risk of 40%, 68%, and 78%, respectively. A significant association between tumor volume and prelaminar invasion was found (² = 11.1, odds ratio = 5.5, P = .001) (Table 3). Location of tumor in the eye had no significant relationship with prelaminar tumor invasion risk. Risk of tumor extension into the choroid also increased with increasing tumor volume from 0% for very small tumors (<0.65 cm³), to 10% for small tumors, to 26% for medium tumors, and to 33% for tumors with a volume greater than 1.95 cm³. The association between tumor volume and choroidal invasion was significant (² = 4.7, odds ratio = 3.7, P = .031) (Table 3). Furthermore, very small tumors did not invade the choroid, and in the group of small tumors, two eyes demonstrated some degree of invasion. These findings resulted in a 7% risk for choroidal invasion in tumors smaller than 1.30 cm³. Postlaminar invasion occurred in four eyes, and tumor volume was more than 0.65 cm³ in all of these eyes. Two (10%) eyes with postlaminar invasion were in the group of small tumors. In both the medium tumor group and the large tumor group, invasion was present in one eye. No significant association was found (P = .926, Fisher exact test), probably because of the small number of eyes with postlaminar invasion (Table 3).

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

In regard to retinoblastoma, the data about diagnostic accuracy of MR imaging in the staging of the extent of disease are scarce. To our knowledge, researchers in only two previous studies reported their results about the clinical value of MR imaging findings in comparison with histopathologic findings after enucleation (20,23) (Table 4). In both studies, the number of patients included was relatively small. Barkhof et al (20) analyzed 18 patients with a 0.6-T MR imaging unit with thicker sections and no fat suppression. Schueler et al (23) compared preoperative MR images with histologic findings after enucleation in 21 cases. Because of the increase in spatial and contrast resolution, possibilities for the detection of smaller lesions of intraocular retinoblastoma appeared promising. Sensitivity and specificity in regard to optic nerve invasion and sensitivity in the detection of choroidal invasion, however, were disappointing. Furthermore, no correlation between the degree of choroidal infiltration at histologic analysis and the appearance of this finding on MR images was present.

High level of vascular endothelial growth factor, which is produced in hypoxic retinal and retinoblastoma cells (13), may mediate hypoxia-initiated tumor angiogenesis (26), and it increases fluid permeability from blood vessels (27). The elevated SI within the vitreous fluid was thought to be related to protein leakage into the vitreous, which shortens T1 and T2 relaxation times (28).

Accuracy of MR imaging in the recognition of endophytic and exophytic growth patterns was not high in our study. Exophytic tumors are commonly detached, with a characteristic V-shaped configuration, subretinal fluid collections, and a typical hemispherical appearance. Accuracy of MR imaging in the detection of retinal detachment with accumulation of subretinal fluid is high, but the value of these attendant characteristics at MR imaging in the evaluation of tumor growth pattern is restricted, because retinal detachment and subretinal fluid occurred in tumors with both endophytic and exophytic growth patterns.

The choroid, which is the enhancing inner layer of the bulbar wall between hypointense vitreous and sclera, can be infiltrated by tumor, and tumor infiltration causes replacement of normal structure. On postcontrast T1-weighted MR images, tumor infiltration may be seen as a local thickening of the choroid or a discontinuity in normal linear enhancement beneath the retinoblastoma. Involvement of the choroid increases the risk of metastatic disease because of its rich vascular network (8,9,12). Detection of choroidal invasion is sometimes hampered by the amount of retinal destruction. The entire replacement of retinal pigment epithelium by tumor hampers discrimination between enhancing choroid and enhancing tumor tissue. In all false-positive MR imaging observations, no delineation of tumor tissue was possible, probably because of partial-volume effects. Histopathologic findings corresponded well with MR imaging findings.

Inflammatory reaction of the choroid, secondary to retinoblastoma, appears to be responsible for effusion at the choroidal interstitium. Serous choroidal effusion is demonstrated as a ring-shaped area of high SI at MR imaging with all sequences (29). Accumulation of fluid with high protein content in the choroidal interstitium, as a result of inflammatory transudation from the choroidal vasculature in combination with inflammation hyperemia, may be responsible for SI increase on both T1- and T2-weighted MR images. After gadolinium-based contrast agent administration, the ring-shaped configuration becomes clearer. Choroidal inflammation combined with a granuloma in the optic nerve head may lead to an erroneous diagnosis, because the contrast enhancement pattern may mimic optic nerve invasion.

Detection of ciliary body invasion by retinoblastoma caused similar problems, with false-positive findings as a consequence. Discrimination between tumor and ciliary body was not reliable because of partial-volume effects and lack of spatial resolution. Involvement of the ciliary body, as part of the anterior uveal tract, could also be associated with an increased risk for hematogenous dissemination and should be considered as an indication for enucleation without delay.

Neovascularization in combination with protein leakage from vessels possibly causes increased SI of the anterior chamber. The presence of rubeosis iridis (iris neovascularization) is strongly correlated to the presence of choroidal and prelaminar invasion, and the presence of rubeosis iridis makes enucleation the only therapeutic option (12,13). Marked enhancement on postcontrast T1-weighted MR images of the anterior chamber may represent a vascular reaction to intraocular ischemia, although in five patients no clinical explanation was found for local SI increase. A differentiation between reactive hyperemia and rubeosis iridis was not possible at MR imaging. SI elevation in the anterior chamber could provide clinicians complementary information in estimation of the risk of metastatic disease.

Optic nerve invasion, especially to the postlaminar degree, represents a sign of tumor aggressiveness and is associated with a significant increase of metastatic disease and mortality rate (3,8–11). With use of gadolinium-based contrast agents for contrast enhancement, tumor invasion of the optic nerve can reliably be determined by observation of interruption of linear enhancement at the choroidoretinal complex. Accuracy in detection of SI changes at the prelaminar optic disc is rather high, and this criterion seems very specific for tumor invasion. Sensitivity remains low, because MR imaging is not able to depict microscopic invasion. In some cases, even substantial prelaminar invasion was missed because of unimpaired linear enhancement. The volume of the tumor appears to be the best predictor for optic nerve invasion. Thickening of the intact enhancing choroidoretinal complex also proved to be predictive. Enhancement confined to a tumor that involves the optic disc and interrupts linear enhancement of the choroidoretinal complex is very specific for tumor invasion of the prelaminar optic nerve, but in the absence of this sign, microscopic invasion cannot be excluded, whereas other signs may also predict prelaminar optic nerve invasion.

Sensitivity of MR imaging in the detection of postlaminar optic nerve invasion was also low and, again, specificity was high. Enhancement extending from the tumor into the optic nerve is a confident sign of postlaminar invasion, but absence of enhancement does not always exclude the presence of postlaminar optic nerve involvement. The diagnostic accuracy for postlaminar tumor invasion seems high, but data may be inaccurate because of the small number of patients.

Scleral invasion and extraocular dissemination were not present in our patients. Although exclusion of this form of tumor invasion was histopathologically proved in all eyes, the clinical value of these results remains unknown, because the sensitivity of MR imaging is probably not high enough to detect microscopic and small areas of scleral and extrascleral tumor extension.

In recent years, the role for nonoperative treatment modalities has increased and chemoreduction therapy, combined with focal treatment modalities, has been introduced (30). In consequence of this trend, more children are treated without proper histopathologic assessment of risk factors for metastasis. Several studies focused on clinical features that may be used to predict the presence of metastatic disease and extraocular recurrence (3,8,11,12,31).

We found a strong association between increase in tumor volume and risk of prelaminar optic nerve invasion (P = .001). Location of the tumor in the eye, however, was not significantly related to the risk of optic nerve invasion. Some peripheral tumors were accompanied by prelaminar optic nerve invasion, possibly because of the simultaneous presence of vitreous seeding. Frequent appearance of vitreous or subretinal tumor seeding and a continuous fluid flow from the vitreous into the optic nerve is suggested by Wolter (32), thus explaining the selective motion of free-floating retinoblastoma cells toward the optic nerve head. Postlaminar optic nerve invasion did not show a significant relationship with tumor volume, but the value of this finding is limited because of the small number of eyes with postlaminar optic nerve invasion.

Although in a previous report clinical tumor diameter and tumor thickness were not predictive of choroidal invasion (12), our study showed a significant association between tumor volume and choroidal invasion (P = .031). Absence of choroidal invasion under very small tumors could be related to the unique resistance to tumor invasion demonstrated by the retinal pigment epithelial layer and the Bruch membrane (33).

This study had limitations. It was a retrospective study, and one experienced radiologist performed image analysis. Our MR imaging technique, with use of a standard head coil for evaluation, was limited. Use of orbital surface coils, which provide an increase in spatial resolution and in image contrast and a reduction in artifacts, may improve sensitivity and specificity. Some of the MR imaging studies were performed without fat suppression after gadolinium-based contrast agent administration, which limits evaluation of tumor invasion into orbital fat. Furthermore, improvement of study results could also be realized by future use of gadolinium-enhanced three-dimensional gradient-echo sequences instead of spin-echo techniques.

In conclusion, the results of this study suggest that modern MR imaging is more sensitive and more specific in tumor staging and determination of metastatic risk factors compared with the results documented in previous reports. Detection of intraocular tumor infiltration remains difficult, however, because spatial resolution and signal-to-noise ratio are still not high enough to reduce false-positive and false-negative observations. Tumor volume, as measured with MR imaging, was predictive of prelaminar optic nerve infiltration and choroidal involvement.