Diagnostic a-scan of choroidal melanoma: Automated quantification of parameters

Abstract

Aim: To develop an automated algorithm to quantify ultrasonographic A-scan parameters of choroidal melanoma. Methods: The study included 100 consecutive patients with a clinical diagnosis of choroidal melanoma. Ultrasonographic A-scans (8 MHz, 1,550 m/s, tissue sensitivity = 67 dB) were performed by standard techniques. We created and then utilized a MATLAB® script to generate four quantifiable A-scan parameters: (1) tumor height (mm), (2) the number of internal reflectivity peaks (numerical value), (3) median internal reflectivity (%), and (4) angle κ (°). Results: There were small (≤2.5 mm, n = 32), medium (2.6-10.0 mm, n = 53), and large (> 10.0 mm, n = 14) tumors. The mean number of internal reflectivity peaks counted between the two tumor boundary spikes (surface and base) was 10.0 (σ = 8.7, range 1-37). The median value of the internal reflectivity peaks for all cases varied from 19.8 to 99.5 (mean = 68.3, σ = 20.5). A statistically significant correlation was observed between the tumor height categories and each of the three A-scan parameters: the number of internal reflectivity peaks (ρ = 0.90, p < 0.01), median internal reflectivity (ρ = -0.63, p < 0.01), and a positive angle κ (ρ = -0.32, p = 0.03). Conclusions: An automated algorithm can provide quantifiable A-scan parameters for choroidal melanoma.