Abstract
Purpose: :
To compare the diagnostic performance of Glaucoma Probability Score(GPS) with that of Moorfields Regression Analysis(MRA) using HRT 3.
Methods: :
This prospective cross-sectional study included 110 eyes of normal subjects and 70 eyes of subjects with early to moderate POAG.CSLO images were obtained by using HRT version 3.0.The images were analysed with MRA using a contour line drawn by a single operator and operator independent GPS.The receiver operating characteristic(ROC) curves between normal and glaucomatous subjects were plotted for the stereometric parameters of both the software versions.Agreement between MRA and GPS classifications was quantified using agreement coefficient k.
Results: :
The mean age of normal subjects was 44.49 + 8.58 yrs and glaucoma subjects was 56.36 + 10.97 yrs(p=0.04).The average mean deviation for visual fileds for normal subjects was -2.88 + 1.42db and for glaucoma subjects was -5.75 + 3.18db(p<0.005).Among normal subjects,MRA was ONL for 1 eye whereas GPS was ONL for 34 eyes.The sensitivity and specificity were evaluated using most specific(borderline results included as test negatives) and least specific criteria(borderline results included as test positives).With MRA sensitivity and specificity were 24.2% and 99.1%(most specific criteria) and 43.9%and 99.1%(least specific criteria).With GPS sensitivity and specificity were 56.4%and 90.6%(most specific criteria) and 68.8% and 32.1%(least specific criteria).The sensitivity increased with increasing disc size for both MRA and GPS.The specificity remained similar for varying disc sizes for MRA whereas the same decreased for GPS with increasing disc size.The individual stereometric parameters with the best discrimination were cup/disc area ratio(AUROC = 0.792) and vertical cup disc ratio(AUROC=0.779) for MRA and cup size measure(AUROC = 0.662) and cup depth(AUROC= 0.569) for GPS.The agreement coefficient(weighted k) for the overall MRA and GPS classifications was 0.157(CI = 0.09-0.22).
Conclusions: :
There was a poor agreement between the overall MRA and GPS classifications.It is recommended that MRA should be used for optic nerve head topography in routine clinical practice as GPS tends to give higher false positive results.
Keywords: imaging/image analysis: clinical