Transscleral Optical Spectroscopy of Uveal Melanoma in Enucleated Human Eyes

Abstract

Purpose.: The aims of this study were to use transscleral optical spectroscopy to analyze normal and tumor-infiltrated areas of enucleated human eyes, and to characterize the spectral properties of uveal melanomas in relation to various morphological features.

Methods.: Nine consecutive eyes enucleated for uveal melanoma were examined by transscleral spectroscopy, using a fiber-optic probe that exerted a fixed pressure on the scleral surface. Spectroscopic measurements, covering the wavelength range of 400−1100 nm, were sequentially performed over the uveal melanoma and on the opposite (normal) side of each eye. The eyes were then processed for histological and immunohistochemical analyses. Comparisons between spectral and morphological parameters were performed by Spearman's rank correlation coefficient and unpaired t-test.

Results.: The average reflection intensity obtained from the normal side of the eyes was higher than that from the tumors. The spectral imprint of hemoglobin was lower and that of water was considerably stronger when compared with the tumor side. The diffuse reflection spectra from the melanomas showed a strong correlation with the degree of tumor pigmentation (Spearman's rho = −0.87, P < 0.0001). A weaker correlation was observed between the amount of hemoglobin-related absorption and the density of intratumoral blood vessels (Spearman's rho = −0.25, P = 0.023). The mean diffuse reflection intensity obtained from the spindle cell melanomas was significantly higher than that from the mixed and epithelioid cell melanomas (P < 0.0001).

Conclusions.: Although future in vivo studies are required, these data suggest that transscleral optical spectroscopy is a feasible method for identification and morphological assessment of choroidal tumors.

Introduction

Uveal melanoma is the most common primary intraocular malignancy in adults. The majority of uveal melanomas can be reliably diagnosed by ophthalmoscopy, ultrasonography, and other imaging modalities; but some patients present atypical tumors that pose diagnostic and therapeutic dilemmas. ^1,2 Opaque media may preclude adequate visualization of the tumor and also lead to delayed or missed diagnosis. ^3,4 In such cases, the decision to perform an intraocular biopsy must balance the complexity and potential risks of the surgical procedure, thus emphasizing the need for alternative methods to improve the diagnostic accuracy before initiating treatment. 

Optical spectroscopy is a noninvasive or minimally invasive method that offers biochemical analysis of tissues correlating to tissue morphology. ⁵ Typically, light in the visible and near-infrared wavelength range is used to illuminate the tissue, where it is partially scattered and absorbed. Light scattering is caused by refractive index differences of extra- and intracellular structures, whereas light absorption depends on the presence of various tissue chromophores such as melanin, oxyhemoglobin, deoxyhemoglobin, carotenes, and water. Based on spectral differences associated with parameters like pigmentation, vascularity, oxygenation, fibrosis, and edema, optical spectroscopy allows detailed in vivo analysis of tissues and may help to diagnose cancer and precancerous lesions in many human organs. ^6–8  

Recently, we reported for the first time the feasibility and results of transscleral optical spectroscopy, which was found to be an accurate method for predicting the content of melanin and hemoglobin in experimental choroidal tumors. ^9,10 The fine, homogeneous structure of the sclera and the location of various choroidal lesions just underneath it suggest that a transscleral approach is the most appropriate way to perform optical spectroscopy of intraocular tumors. Owing to morphological characteristics and different contents of melanin and blood, there may be distinct spectral variations between uveal melanomas and other choroidal lesions such as hemangiomas, vasoproliferative tumors, metastases, and choroidal hemorrhages or hematomas. However, significant morphological differences also exist among uveal melanomas in general, which are subclassified into spindle and epithelioid cell types, different stages based on tumor size, and various degrees of pigmentation ranging from amelanotic to darkly pigmented tumors. The objectives of the present work were to evaluate the feasibility of using transscleral optical spectroscopy to analyze both normal and tumor-infiltrated areas of human eyes, and to study the spectral properties of choroidal melanomas in relation to various morphological features. 

Methods

Patients and Specimens

Nine consecutive patients enucleated for posterior uveal melanoma at the Department of Ophthalmology of Haukeland University Hospital, between January 2010 and June 2011, were included in the study. Eight eyes underwent primary enucleation, and one eye (patient no. 9, the Table) was enucleated 15 months after iodine-125 brachytherapy due to lack of local tumor control. Prior to surgery, ophthalmoscopy, ultrasonography, fluorescein angiography, and magnetic resonance imaging revealed the location, size, and shape of the tumors, as well as the degree of exudative retinal detachment. The tumors were grouped according to the TNM classification, ¹¹ and the degree of retinal detachment was scored semiquantitatively from 1 to 3. Immediately after enucleation, the eyes were brought to the laboratory for spectroscopic analysis. The study was registered and approved by the Regional Committee for Medical and Health Research Ethics, Western Norway, and followed the official ethical regulations for clinical research and the Declaration of Helsinki. All patients gave their written informed consent prior to participation in the study. 

Table 1. 

 

Patient and Melanoma Characteristics

Patient no.	Age (y)	Sex F/M	Location Relative to Equator	Shape	Retinal Detachment	Diameter (mm)	Height (mm)	TNM Class	Cell Type	Pigment (%)	Blood Vessel Density	Mitotic Figures
1	50	F	Ant. + post.	Mushroom	++	15.1	11.4	T3a	Mixed	85.5	45	8
2	27	F	Peripapillar	Dome	+++	11.5	4.0	T2a	Mixed	4.5	345	5
3	63	M	Posterior	Mushroom	+	13.5	13.9	T3a	Spindle	4.4	30	1
4	87	M	Ant. + post.	Dome	++	17.0	10.0	T3a	Epithelioid	4.3	52	9
5	55	M	Anterior	Mushroom	++	15.6	13.3	T4b	Epithelioid	45.7	49	3
6	87	M	Ant. + post.	Dome	+++	19.5	13.5	T4b	Spindle	<1	148	7
7	64	F	Peripapillar	Dome	++	11.4	4.9	T2a	Spindle	2.3	182	3
8	63	F	Peripapillar	Dome	++	9.8	2.8	T1a	Spindle	3.6	29	<1
9	68	M	Posterior	Dome	++	12.8	2.6	T2a	Epithelioid	3.0	24	<1

Spectroscopy Instrumentation

A schematic illustration of the experimental setup and the principle of transscleral optical spectroscopy are shown in Figure 1. The spectroscopy instrumentation has previously been described in detail. ¹² Briefly, a high-powered projector quartz tungsten halogen lamp (Oriel Simplicity 66765; Newport Corporation, Irvine, CA) produced light with a smooth spectrum across the visible and near-infrared region. A 600-μm diameter optical fiber (BFL48-600; Thorlabs, Newton, NJ) was used to deliver the light to the sclera, and a second similar fiber collected the light. The optimal distance between the source and detector fiber was estimated using Monte-Carlo light-propagation simulations based on the optical properties and geometries of ocular tissues. ¹³ A center-to-center fiber separation of 6 mm was found to give high fractions of backscattered photons both when the measurements were performed on areas with an underlying melanoma and on normal regions of the eye. Light detection was performed by a fiber-coupled spectrometer (QE65000; Ocean Optics, Dunedin, FL). To control the placement of the two optical fibers and their pressure on the scleral surface, the fibers were mounted on micrometer-translator stages (PT/1M; Thorlabs), which enabled them to be moved both vertically and horizontally. By placing the eye on an electronic scale (Kern 572; Kern & Sohn GmbH, Balingen, Germany), the pressure exerted on the eye surface was monitored and kept constant for all measurements. A polytetrafluoroethylene disc (LBS-SRS-99-020; Spectralon, Labsphere Inc., North Sutton, NH) was used as the optical reference standard. 

Figure 1. 

 

Schematic illustration of the experimental setup and the principle of transscleral optical spectroscopy: (1) High-powered projector quartz tungsten halogen lamp. (2) Human uveal melanoma-containing eye placed on an electronic scale. (3) Cross-section of the eye showing the optical fibers for incident and detected light centered on the scleral surface over the choroidal melanoma. (4) Spectrometer. (5) Computer for spectral analysis.

Measurement Procedures

After enucleation, the eye was gently rinsed with saline to remove residual blood from the surface. The exact location and extent of the uveal melanoma were visualized by transocular transillumination, thereby facilitating a correct positioning of the optical fibers during spectroscopy. The eye was then placed on the electronic scale, and the fibers (with a fixed source−detector separation of 6 mm) were centered on the scleral surface over the melanoma and gradually lowered until an additional weight of 1.5 g was reached. Spectroscopy was performed with an exposure time of 10 to 15 seconds, and the spectral data were transferred to a computer and analyzed using data analysis software (MATLAB, Version 7.8 [R2009a]; The MathWorks Inc., Natick, MA). A similar measurement procedure was performed on the opposite, normal side of the eye. Ten consecutive measurements were carried out alternately on each side of the globe. A reference standard spectrum was recorded between each measurement. 

Preparation for Light Microscopy and Immunohistochemistry

Following completion of the spectroscopic recordings, the eye was fixed in 10% phosphate-buffered formalin and processed by standard methods. One paraffin block with the maximum tumor bulk was chosen from each eye for histological analysis. Specimens including the melanoma and surrounding tissues were cut in 5-μm sections and mounted on glass slides. For conventional histopathological evaluation, the sections were stained with hematoxylin-eosin and examined by light microscopy (Leica DM L; Leica Microsystems, Wetzlar, Germany). 

Immunostaining of blood vessels was performed on 5-μm sections dewaxed in xylene and dehydrated through graded ethanol. For microwave antigen retrieval, the sections were boiled in target retrieval solution, pH 6.0 (Dako, Glostrup, Denmark) for 10 minutes at 750 W; followed by 20 minutes at 350 W; and then incubated with the monoclonal CD31 antibody (clone JC/70A [M0823]; Dako) diluted 1:25 for 30 minutes at room temperature. Staining was performed on an automated horizontal slide-processing system (Dako Autostainer; Dako), visualized by a detection system (EnVision Kit; Dako) with diaminobenzidine peroxidase as substrate. Antibody binding was detected by enzyme- and antibody-marked polymer (K4011 for polyclonal rabbit antibodies, K4007 for mouse antibodies; DakoCytomation, Glostrup, Denmark). The sections were finally counterstained with hematoxylin for 1 minute. 

Quantization of Morphological Parameters

For the evaluation of tumor pigmentation, vascularity, and mitotic figures, multiple sections from representative areas of each tumor were analyzed. The degree of pigmentation was determined by counting the number of melanin containing cells relative to the total number of cells in 10 high-power fields (HPF; 400× magnification) across representative areas of the tumor. In tumors with inhomogeneous pigmentation, the level of pigmentation was assessed at low magnification by calculating the sum of areas with predominantly pigmented cells relative to the total tumor area in each section. The pigmented regions were manually outlined on the digitized microscope images, and the touch count and area analyses were performed using a microscope-mounted camera (ColorView IIIu; Soft Imaging Systems GmbH, Münster, Germany) with imaging software (Olympus DP Software; Olympus, Tokyo, Japan). 

Blood vessel density (BVD) was determined by counting CD31-positive structures in at least 10 randomly chosen fields at 200× magnification. Any immunostained endothelial cell cluster or vessel, clearly separated from adjacent vessels, was considered to be a single countable blood vessel. The results were averaged and calculated as the number of blood vessels per square mm. Mitotic frequency (number of cells with mitotic figures in 10 HPF) was also recorded. All histological analyses were done by an experienced pathologist (SJM) who was blinded to the spectroscopic data. 

Statistical Analysis

Correlations between the diffuse reflection spectra and morphological parameters were analyzed with Spearman's rank correlation coefficient. Each variable was tested for the hypothesis of no correlation against the alternative that there was a nonzero correlation. To test the differences between melanomas of different cell types, a two-tailed unpaired t-test was performed. 

Results

Patients and Specimens

The study included four women and five men with a mean age at the time of enucleation of 63 years (range, 27–87 years). In all patients, the main tumor site was the choroid. The ciliary body was involved in two eyes, while three patients had a peripapillary melanoma. All tumors were surrounded by varying degrees of exudative retinal detachment. The mean largest basal tumor diameter was 14.0 mm (range, 9.8−17.0 mm), and the mean largest tumor thickness was 8.5 mm (range, 2.6−13.9 mm). On gross examination, six tumors appeared as dome-shaped and three as mushroom-shaped choroidal masses, with a varying degree of pigmentation ranging from amelanotic white to dark black. The patient and tumor characteristics are summarized in the Table. 

Histological and Immunohistochemical Findings

Based on the histological findings, three melanomas were classified as epithelioid, four as spindle B, and two as mixed cell type. No signs of necrosis were observed in any of the eyes. As shown in the Table and in Figures 2A and 2B, the level of pigmentation, derived from the percentage of pigmented cells (in seven homogeneously pigmented tumors) or pigmented areas (in two inhomogeneously pigmented tumors), varied widely between the different melanomas. Variation of BVD was also observed, with a mean and median intratumoral blood vessel count per mm² of 100 and 49, respectively (range, 24−345; Fig. 2C). The mean and median numbers of mitotic figures in 10 HPF were 5 (range, <1−9). The histopathological appearance of the melanoma in patient no. 9 differed markedly from the primarily enucleated melanomas (Fig. 2D). This eye had 15 months earlier undergone iodine-125 brachytherapy with a prescription dose of 90 Gy to the tumor apex. The melanoma showed a moderate degree of pigmentation, mainly due to pigment-laden macrophages within the tumor tissue, and irradiation-induced changes like vacuolization of tumor cells, sclerosis and thickening of vessel walls, and interstitial fibrosis. A sheet of fibrovascular granulation tissue was observed on the sclera overlying the tumor, corresponding to the area of previous iodine plaque placement (Fig. 3). 

Figure 2. 

 

Photomicrographs showing: (A) Heavily pigmented melanoma in patient no. 1. (B) Essentially amelanotic spindle cell melanoma in patient no. 6 (hematoxylin-eosin stain). (C) CD31-positive blood vessels in the spindle cell melanoma in patient no. 3 (CD31 immunostain). (D) Residual melanoma after irradiation in patient no. 9. Note the thickened vessel walls and the stromal fibrosis with accumulation of pigment-laden macrophages (hematoxylin-eosin stain). Original magnifications 200×, scale bars = 100 μm.

Figure 3. 

 

Photomicrograph of the outer sclera in patient no. 9. A fine layer of well-vascularized fibrous tissue (asterisk) is seen on the scleral surface (arrow) corresponding to the area previously covered by the iodine-125 plaque during brachytherapy (hematoxylin-eosin stain). Original magnification 200×, scale bar = 100 μm.

Spectroscopic Analysis

The average intensity of the diffuse reflection spectra measured from the normal side of the eyes was slightly higher than that from the melanomas. A relatively high absorption was observed at 500 to 600 nm and a strong absorption peak was seen around 980 nm, corresponding to the light absorption of hemoglobin and water, respectively. At the normal side of the eyes, the spectral imprint of blood was lower and that of water was considerably stronger when compared with the tumor side. The diffuse reflection spectra obtained from all the melanomas and from the normal side of the eyes are illustrated in Figures 4A and 4B, respectively. To further demonstrate the spectral differences between the melanomas, line graphs of the mean diffuse reflection spectra for each patient and measurement site are presented in Figures 5A and 5B. 

Figure 4. 

 

Diffuse reflection spectra. (A) Obtained from uveal melanomas. (B) Obtained from the normal side of the eyes. Every horizontal line in the images represents a single spectrum. Ten consecutively acquired spectra are presented for each patient and measurement site. The colors indicate the intensity of the diffuse reflection as given by the color bar to the right.

Figure 5. 

 

Diffuse reflection spectra. (A) Obtained from uveal melanomas. (B) Obtained from the normal side of the eyes. Each curve represents the mean of 10 consecutively acquired diffuse reflection spectra for each patient and measurement site. The patient number is color-coded as given by the legend to the right.

The spectroscopic identification of melanin was difficult because it has no distinct absorption features. Light absorption of melanin increases almost exponentially from the near-infrared toward shorter wavelengths. For the two melanomas with the strongest pigmentation (patients no. 1 and 5), a clear differentiation in the spectra could be observed. Both presented considerably lower diffuse reflection intensity around 900 nm (Figs. 4A, 5A). For a refined analysis of all the eyes studied, each spectrum was divided into 12 equally wide spectral bands. Each band was examined for correlations with the morphological features of the melanomas. In a wide range of 575−1100 nm, the diffuse reflection spectra obtained from the melanomas showed a strong correlation with the degree of pigmentation (Spearman's rho = −0.75, P < 0.0001). Specifically in the narrower band of 620−664 nm, a slightly increased correlation coefficient was found (Spearman's rho = −0.87, P < 0.0001; Fig. 6A). 

Figure 6. 

 

Scatter plot of the diffuse reflection intensity. (A) Versus tumor pigmentation. (B) Versus intratumoral blood vessel density. (C) Box plot displaying the diffuse reflection intensity for the mixed and epithelioid cell melanomas and for the spindle cell melanomas (middle line: the median; bottom and top box edges: the 25th and the 75th percentiles; whiskers: the range).

For the analysis of BVD, a weak correlation was observed between the amount of hemoglobin-related absorption in the wavelength range of 394 to 420 nm and the number of intratumoral blood vessels (Spearman's rho = −0.25, P = 0.023; Fig. 6B). When analyzing the spectral data from melanomas with different cell types (excluding patient no. 9), significant differences in diffuse reflection across the spectrum could be measured (Fig. 6C). The spindle cell melanomas (n = 4 eyes and 40 measurements) with a mean diffuse reflection intensity of 0.0069 at the short end of the spectrum (530–575 nm), guided the light more efficiently compared with the mixed and epithelioid cell melanomas (n = 4 eyes and 40 measurements) with a mean intensity of 0.0026 (P < 0.0001, two-tailed, two-sample unpaired t-test). The diffuse reflection spectrum obtained from the postirradiation melanoma in patient no. 9 differed markedly from the other melanoma spectra. It showed a higher light absorption at shorter wavelengths due to a very strong imprint of hemoglobin around 500–600 nm (Figs. 4A, 5A). 

No clear spectral correlations were observed between the diffuse reflection spectra and any of the other tumor characteristics. Because of the small number of eyes studied, further statistical analyses of the spectral data were not performed. 

Discussion

Owing to the technical difficulties and possible risks associated with biopsy of intraocular tumors, diagnosis and treatment decisions are often based solely on clinical and imaging findings. Alternative noninvasive methods for tumor analysis are therefore desired to make a more conclusive diagnosis before proceeding with definitive treatments like irradiation or enucleation. 

In this study, we have used transscleral optical spectroscopy to characterize the spectral properties of uveal melanomas in relation to various morphological features in enucleated human eyes. Marked differences were observed between the absorption spectra obtained from the normal side of the eyes, the untreated melanomas, and the irradiated melanoma in patient no. 9. The absorption peaks indicating less blood and higher measured water content in the normal eye wall compared with the uveal melanomas may be explained by the thickness of the probed tissue and the generally high degree of vascularity and pigmentation in malignant melanomas. It is likely that the increased water absorption observed in the spectra from the thinner, normal side of the eyes is caused by light interacting with the high (99%) water content of the vitreous body. ¹⁴ These results are consistent with our earlier findings in porcine eyes, where statistical regression modeling of the spectral data yielded a high discrimination rate between the normal side and the tumor side of the eyes. ¹⁰  

When analyzing the diffuse reflection spectra of the uveal melanomas, we found that their melanin content was strongly correlated with the absorption of light in the wavelength region between 575 and 1100 nm. This is also in accordance with our previous study on melanin-containing experimental tumors, and suggests that light absorbance by melanin can be a useful parameter for the discrimination of choroidal tumors. ⁹ Although the degree of pigmentation varies considerably between uveal melanomas, the presence or absence of melanin is important to distinguish between melanomas and relevant nonpigmented choroidal lesions like hemangiomas, vasoproliferative tumors, metastases, and hemorrhages. The relatively weak correlation between BVD and the absorption spectra of hemoglobin may be due to the reduced blood volume in enucleated eyes, providing an uneven blood filling of the intratumoral vessels. In vitro optical spectroscopy of human brain tumors has shown more variable results and less pronounced hemoglobin absorption when compared with in vivo measurements. ^15,16 By means of transscleral optical spectroscopy, we have recently demonstrated that different blood concentrations in experimental choroidal tumors can be distinguished with a very high degree of certainty. ¹⁰ Under in vivo conditions, the method may therefore be of particular diagnostic value for the differential diagnoses mentioned above. A higher intensity of the reflected light was observed for the melanomas of spindle cell type compared with the mixed and epithelioid cell melanomas. It is tempting to speculate that this could be related to the more uniform cells and isotropic morphology of the spindle cell melanomas, influencing the scattering and propagation of light within the tissue. However, the small number and variable pigmentation of the tumors limit further interpretation of these results. 

The diffuse reflection spectrum of the irradiated uveal melanoma in patient no. 9 differed significantly from those of the other tumors by a stronger light absorption in the shorter wavelength region. This is most likely caused by the extensive fibrosis, deposition of collagen, and accumulation of melanin-containing macrophages commonly found in irradiated tumor stroma. ^17,18 Considering the relatively low number of intratumoral blood vessels, the spectrum showed a surprisingly strong imprint of hemoglobin. During the spectroscopy measurements, however, a red discoloration of the sclera was observed, which proved to be due to a fine layer of highly vascularized granulation tissue from the previous plaque placement. As blood contamination and vessels close to the optical fibers have a strong impact on the spectral imprint of hemoglobin, ¹⁶ it will be important for future clinical studies to inspect the scleral surface and remove excess tissue and residual blood prior to the measurements. 

The main limitations of this study include the small sample size and the in vitro technique. A fundamental challenge of biomedical optics is the validation of measurement results obtained from excised organs. In postmortem tissues, absorption spectra and optical properties related to physiological parameters such as blood flow and oxygenation will change significantly. ¹⁹ Recently, we have developed a handheld fiber optic probe for in vivo transscleral spectroscopy, geometrically optimized for light propagation through the layered structure of the eye wall. ¹² With this probe, a slightly smaller separation (5 mm) between the source and detector fibers was found to be optimal in terms of relevant optical and spectroscopic parameters. The study further indicated that our method of transscleral spectroscopy is a safe procedure that does not damage the eyes. In the future, we intend to apply this probe for in vivo characterization of intraocular tumors. It should be noted that the main purpose of this method is not to analyze uveal melanomas per se, but to differentiate them from other types of intraocular tumors. Transscleral spectroscopy is probably most suitable for choroidal tumors located at or anterior to the equator. In order to probe more posteriorly situated lesions, a minimally invasive procedure in the form of a conjunctival incision may be needed to allow free access to the sclera. Likewise, the method can be useful to establish a correct diagnosis during various forms of ocular surgery and for tumor demarcation and follow-up in conjunction with episcleral brachytherapy. 

In summary, markedly different spectral responses were observed from the uveal melanomas compared with the surrounding, normal parts of the eye. When comparing the spectroscopic results with standard morphological criteria, a strong correlation was found for the degree of tumor pigmentation, and a somewhat weaker correlation was obtained for the density of intratumoral blood vessels. Significant variations in the reflection intensity were also seen between melanomas of different cell types. Although further research—especially in vivo studies—is required, our results indicate that transscleral optical spectroscopy has the potential to be implemented in the panel of noninvasive diagnostic techniques currently used for the evaluation of intraocular tumors. 

Acknowledgments

The authors thank Laila Vårdal for technical assistance with the histological preparation of the specimens. 

Footnotes

 Supported by grants from the Western Norway Regional Health Authority and a Linnaeus grant for the Lund Laser Centre.

Footnotes

 Disclosure: J. Krohn, None; P. Svenmarker, None; C.T. Xu, None; S.J. Mørk, None; S. Andersson-Engels, None