Purchase this article with an account.
Tin Aung, Monisha E. Nongpiur, Benjamin A. Halland, Li-Lian Foo, Mingguang He, Tien Y. Wong, David S. Friedman; Classification Algorithms Based on Anterior Segment OCT Measurements for the Detection of Gonioscopic Angle Closure. Invest. Ophthalmol. Vis. Sci. 2012;53(14):738.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Anterior segment optical coherence tomography (ASOCT) offers the potential to screen for angle closure. We recently found that a combination of 6 ASOCT parameters (anterior chamber area, volume and width [ACA, ACV, ACW], lens vault [LV], iris thickness at 750μm from the scleral spur [IT750] and iris cross-sectional area [IArea]) explained more than 80% of angle width variability. The aim of this study was to evaluate several classification algorithms based on ASOCT measurements for the prediction of gonioscopic angle closure.
A total of 2047 subjects aged ≥50 years, were recruited from a community polyclinic in Singapore. All participants underwent gonioscopy, ASOCT (Carl Zeiss Meditec, Dublin, CA) and IOLMaster (Carl Zeiss Meditec, Dublin, CA). Customized software (Zhongshan Angle Assessment Program, Guangzhou, China) was used to measure ASOCT parameters on horizontal ASOCT scans. IOLMaster was used to measure anterior chamber depth (ACD) and axial length. Six classification algorithms were considered (stepwise logistic regression with Akaike information criterion, random forest, multivariate adaptive regression splines, support vector machine, naïve Bayes classification, and recursive partitioning). Point and interval estimates of the area under the receiver operating characteristic (AUC) curve were generated for these algorithms using both 10-fold cross-validation as well as 50:50 training and validation.
Data on 1368 subjects, including 295 (21.6%) subjects with gonioscopic angle closure were available for analysis. The mean age of subjects was 62.7 ± 7.7 and 54.1% were females. Persons with angle closure were older, had smaller ACD, ACW, ACA, and ACV, greater LV and thicker irides (p<0.05 for all). For both the 10-fold cross-validation and the 50:50 training and validation methods, stepwise logistic regression, incorporating 6 ASOCT parameters, was nominally the best algorithm for detecting eyes with gonioscopic angle closure, with AUC of 0.954 (95% confidence interval (CI); 0.942-0.966) and 0.962 (95% CI; 0.948-0.975) respectively, while recursive partitioning had relatively poorest performance with AUC of 0.860 (95% CI; 0.790-0.930) and 0.905 (95%CI; 0.876-0.933) respectively. Interval estimates of the AUC were also formed using the six algorithms. Stepwise logistic regression provided the most stable algorithm with the least variability.
A classification algorithm, using ASOCT data from a single scan, based on stepwise logistic regression modelling predicts gonioscopic angle closure with greater than 95% accuracy. This has potential to be incorporated into ASOCT angle analysis software for angle closure screening.
This PDF is available to Subscribers Only