Abstract
Purpose :
Trabecular-iris angle (TIA) anatomy plays an important role in evaluating and treating glaucoma although the gold standard measuring method remains highly subjective. This study tested the reliability of manual and automated angle measurements using Ultrasound Biomicroscope (UBM) images from different anatomical locations.
Methods :
Sequential 2D UBM images of five eyes from one adult and four pediatric patients were converted to 3D images. Four images including eight TIAs at anatomical locations separated by 45° about the eye center were sliced from each 360° 3D reconstruction of the iridocorneal junction. Eight trained observers measured 40 angles manually according to Palvin’s definition using ImageJ. Same angles were analyzed by a completely automatized algorithm and the results were compared to the means of observers’ measurements. Intra-Class Correlation (ICC) estimates and their 95% confident intervals were calculated using SPSS version 25 based on the alpha, absolute-agreement, 2-way mixed-effects model for single measures. An angle map of the adult patient was created by the automated analysis of the 360 images separated by 0.5° about the eye center.
Results :
Means and standard deviations of first and second measurements ranged 19.3°–48.9°±1.9-10.4 and 19.1°–49.0°±1.6 -11.2 respectively. Variability of TIA with anatomic locations in each eye ranged within 25° in manual measurements. ICC for inter-observer agreement of first and second measurements were 0.727 [0,626-0,821] and 0.680 [0,570-0,786] (p<0,0005). Intra-observer agreement showed ICCs ranged 0,667 - 0,860 (p<0,0005). Means of first and second measurements showed ICC 0,942 [0,786-0,977] (p<0,0005). ICC for automated measurements was significant for 78 years old adult patient with no ocular pathology (0,717 (p=0,014)) and 6 years old patient with congenital glaucoma (0,654 (p=0,039)). Overall ICC for automated results was 0,542 [0,278- 0,729] (p<0,0005).
Conclusions :
Manual measurements in accordance with Palvin’s definition resulted in good intra and inter-observer agreements. Automated angle detection algorithm demonstrated promising results in two patients. With 3D reconstruction and processing, TIA was assessed at multiple locations. Angle maps may explain the variability of TIA and provide crucial information for evaluation and treatment of glaucoma.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.