Thank you for your submission!
A Novel Approach to Estimating Choroidal Lesion Thickness Using 2D Ultra-Widefield Optomap Images
Michael D Yu
Michael Yu, MD; Michael Heiferman, MD; Edward Korot, MD; Prithvi Mruthyunjaya, MD, MHS
Byers Eye Institute, Stanford University, Palo Alto, CA
Tumor thickness is a well-established risk factor for transformation of choroidal nevus (CN) into choroidal melanoma (CM) and thus plays an important role in risk stratification of melanocytic choroidal lesions (MCL). Currently, B-scan ultrasonography is the most reliable method for measuring tumor thickness, but its utility in screening may be limited. Herein, we describe a novel technique for rapid extraction of tumor thickness data from 2D ultra-widefield (UWF) dual-wavelength scanning laser ophthalmoscope Optomap images (Optos PLC, Dunfermline, Fife, Scotland, UK).
A consecutive series of patients seen in the Ocular Oncology Service at the Byers Eye Institute (Stanford University, Palo Alto, CA) with clinically diagnosed MCL underwent complete clinical examination, UWF imaging, and standardized B-scan ultrasonography (Eye Cubed, Ellex Medical, Adelaide, Australia). The UWF images were post-processed to isolate the green-wavelength-only image. Using Image J (National Institutes of Health, Bethesda, MD, USA), average pixel intensities within the lesion and of the adjacent retina were obtained, and the difference between both values calculated (“pixel intensity difference”; average lesion intensity minus average adjacent retina intensity). The pixel intensity difference was then plotted against the reference standard for tumor thickness as measured by standardized B-scan ultrasonography. The significance of the relationship between both variables was assessed by linear regression analysis.
A total of 141 MCL (16 CM and 125 CN) of 141 patients were evaluated. Mean ultrasonographic thickness was 1.2 mm (median: 0.8, range: 0.5-7.3). Mean pixel intensity difference was 6.7 (median: 3.8, range: -20.0 – 55.0). The linear correlation coefficient for tumor thickness to intensity difference was 0.85 (p<0.001), indicating a strong positive correlation between tumor thickness and tumor brightness on green-wavelength imaging (Figure 1). Coefficient of determination (R2) was 0.74. A pixel intensity difference threshold of >10 conferred a 100.0% sensitivity and 97.4% specificity for detection of tumors with thickness >2 mm.
Choroidal tumor thickness can be rapidly and reliably estimated using 2D UWF images. With additional validation, this method could augment future high-throughput screening and risk stratification of MCL with UWF images alone.