Posters

Manual vs automated gonioscopy: a comparison pilot study

Poster Details

First Author: F.Teixeira PORTUGAL

Co Author(s):    P. Jose   D. Sousa   I. Leal   C. Marques-Neves   L. Pinto        

Abstract Details

Purpose:

To evaluate the new version of a prototype capable of iridocorneal angle automated imaging in comparison with manual gonioscopy.

Setting:

Glaucoma clinic at Hospital de Santa Maria in Lisbon, Portugal.

Methods:

Prospective observational single-center study. Manual and automated gonioscopy were performed to consecutive patients in a glaucoma clinic. The same glaucoma specialist performed all manual gonioscopies and graded ICA opening accordingly to Schaffer’s classification. Automated gonioscopy was performed with the new version of the NGS-1 automated gonioscope prototype (Nidek Technologies, Padova, Italy). Two independent observers graded the ICA from the automated gonio-photos. Agreement between automated and manual gonioscopy and also among raters was ascertained by Fleiss’ κappa (Fleiss’ κ) statistic and comparison of area under curves (AUC).

Results:

Fourty-eight eyes of 26 subjects were analyzed. Mean age was 68 ± 12 years, and 50% were female. Angle closure was detected in 16,7% of the eyes with dynamic gonioscopy in comparison with 12,5% using automated images grading. There was moderate agreement for angle closure diagnosis between dynamic and automated gonioscopy (Fleiss’ κ = 0.56±0.12; p value<0,001). The AUC for detecting eyes with angle closure showed good accuracy between automated and manual methods (AUC= 0.81 ± 0.10; p value=0,006). Inter-rater agreement for angle opening assessment of automated images was moderate (Fleiss’ k = 0.57±0.16].

Conclusions:

Manual and automated gonioscopy showed moderate agreement for the assessment of ICA opening status. Moderate agreement was also found between both raters for angle opening. Further improvements of the NGS-1 automated gonioscopy and technique are desired for widespread use in a real-life setting.

Financial Disclosure:

None