Automated in vivo detection and localization of posterior staphyloma in ultrasound b-modes

In high and pathologic myopia eyes, staphyloma, an outpouching of the eyewall, is associated with risks of permanent vision loss. Automated detection of staphyloma in ultrasound B-modes could assist clinicians in quickly identifying patients at risk. In this study, an algorithm to detect the presence of staphyloma and locate the staphyloma apex was developed. A 10 MHz ultrasound probe was used to acquire 127 B-modes, from 82 eyes and 52 patients. Sixty-one of the B-modes included staphyloma, as determined by an experienced ophthalmologist who also marked the staphyloma apex in each image where present. In each B-mode, the vitreoretinal boundary was automatically detected by a thresholding method. The local radius of curvature (K) was computed, and the ability of the standard deviation of K to detect staphyloma was evaluated by receiver-operating characteristic analysis. The staphyloma apex was automatically detected as the point on the vitreoretinal boundary furthest from the center of the transducer. The area under the curve using standard deviation of K for staphyloma detection was 0.927. The average error in detected staphyloma apex location was 1.34 ± 1.36 mm. These results demonstrate the potential of automated staphyloma detection and localization from ultrasound, which could assist in quick diagnosis of staphyloma.