Transscleral visible/near-infrared spectroscopy for quantitative assessment of melanin in a uveal melanoma phantom of ex vivo porcine eyes.

Optical spectroscopy has been used as a supplement to conventional techniques for analyzing and diagnosing cancer in many human organs. Because ocular tumors may be characterized by their different melanin content, we investigated the feasibility of using transscleral visible/near-infrared spectroscopy (Vis/NIRS) to estimate the quantity of melanin in a novel uveal melanoma phantom of ex vivo porcine eyes. The phantoms were made by injecting a freshly prepared suspension of 15% (wt/vol) gelatin, 10 mg/ml titanium dioxide (TiO(2)), and natural melanin, isolated from the ink sac of cuttlefish (Sepia officinalis), into the suprachoroidal space of 30 enucleated porcine eyes. The melanin concentrations used were 1 mg/ml, 2 mg/ml, and 3 mg/ml, with 10 eyes in each group. After gelation, the size and location of the phantoms were documented by B-scan ultrasonography and transillumination. Vis/NIRS recordings, covering the wavelength region from 550 to 1000 nm, were performed with two optical fibers separated by 6 mm to deliver and collect the light through the sclera. During all measurements, the exact pressure exerted by the fiber probe on the scleral surface was monitored by placing the eye on an electronic scale. Transscleral Vis/NIRS was performed across the phantom inclusion, as well as on the opposite (normal) side of each eye. A total of three consecutive measurements were carried out alternately on each side of the globe. The spectral data were analyzed using partial least squares regression. In the melanin concentration groups of 1 mg/ml (n = 10), 2 mg/ml (n = 10), and 3 mg/ml (n = 10), the largest basal phantom diameters (mean +/- SD) were 14.9 +/- 1.6 mm, 14.6 +/- 1.5 mm, and 14.3 +/- 1.0 mm, respectively (p > 0.05). The largest phantom thicknesses (mean +/- SD) were 4.0 +/- 0.5 mm, 4.4 +/- 0.7 mm, and 4.5 +/- 0.5 mm, respectively (p > 0.05). Statistical regression modeling of the Vis/NIRS data revealed that it was possible to correctly classify the phantoms according to their melanin concentrations in 84.4% of cases. The correct classification rate for phantoms with the lowest (1 mg/ml) and highest (3 mg/ml) melanin concentrations was 99.2%. The study demonstrates that transscleral Vis/NIRS is a feasible and accurate method for predicting the content of melanin in choroidal lesions. Copyright 2009 Elsevier Ltd. All rights reserved.