Pixel Intensity to Estimate Choroidal Tumor Thickness Using 2-Dimensional Ultra-Widefield Images

Abstract

ImportanceTumor thickness is a well-established risk factor for malignant transformation of choroidal nevus into melanoma. To date, reliable evaluation of tumor thickness relies on B-scan ultrasonography, which is frequently unavailable at nonsubspecialty clinics where most melanocytic choroidal lesions (MCLs) are first diagnosed.ObjectiveTo describe a technique for potential rapid and reliable estimation of MCL thickness using only Optomap ultra-widefield (UWF) images without B-scan ultrasonography.Design, Setting, and ParticipantsThis retrospective, exploratory, cross-sectional analysis of consecutive MCLs diagnosed at Byers Eye Institute, Stanford University, Palo Alto, California, investigates the quantitative correlation between B-scan ultrasonographic thickness of MCLs with the relative pixel intensity of MCLs on green-channel (GC) Optomap UWF images. Pixel intensity overlying the lesion was standardized to the pixel intensity surrounding the lesion (pixel intensity difference [PID]), which was then correlated with ultrasonographic tumor thickness in all study lesions using linear regression analysis. Data were collected from January 1, 2019, to July 1, 2021.Main Outcomes and MeasuresThe correlation between ultrasonographic tumor thickness and PID was measured. Performance (sensitivity and specificity) of the regression analysis trendline in estimating tumor thickness was also determined.ResultsA total of 138 MCLs from 138 patients (mean age, 57.0 years; 51% female) were included, comprising 125 that were nevi and 13 that were melanoma. The mean ultrasonographic tumor thickness was 1.1 mm (median, 0.8 mm; range, 0.2-5.5 mm), with a mean PID of 2.13. Stratifying lesions by ultrasonographic thickness (<1.0 mm vs 1.0-2.0 mm vs >2.0 mm), the mean PID increased (−1.95 vs 3.72 vs 17.62; P < .001) as mean thickness increased (0.5 mm vs 1.4 mm vs 3.3 mm; P < .001). PID was correlated with tumor thickness (R2 = 0.823; 95% CI, 0.770-0.875; P < .001) across lesions of all sizes.Conclusions and RelevanceIn this proof-of-principle study, GC-derived lesion intensity correlated well with ultrasonographic tumor thickness. Leveraging this correlation, this study demonstrates a technique for potentially reliable and rapid estimation of tumor thickness using UWF Optomap images without the use of B-scan ultrasonography.