RNFL thickness (RNFLT) measurements from scanning laser polarimetry (SLP) and visual field (VF) sensitivity from standard automated perimetry were made available for 535 eyes from three centers «Zhu H, Crabb DP, Schlottmann PG ym. Predicting vis...»1. In a training dataset, structure-function relationships were characterized by using linear regression and a type of neural network: radial basis function customized under a Bayesian framework (BRBF). These two models were used in a test dataset to (1) predict sensitivity at individual VF locations from RNFLT measurements and (2) predict the spatial relationship between VF locations and positions at a peripapillary RNFLT measurement annulus. Compared with linear regression, BRBF yielded a nearly twofold improvement (P < 0.001; paired t-test) in performance of predicting VF sensitivity in the test dataset (mean absolute prediction error of 2.9 dB [SD 3.7] versus 4.9 dB [SD 4.0]). The predicted spatial structure-function relationship showed better agreement (P < 0.001; paired t-test) with anatomic prior knowledge when the BRBF was compared with the linear regression (median absolute angular difference of 15° vs. 62°).
When Capriolin considered only visual field damage alone or structural damage alone «Caprioli J. Discrimination between normal and glau...»2, the sprcificity was poor (6–35 %). Combining structure and function increase the specificity to 76 % and sensitivity to 90 %.
A randomised sample of 3001 Caucasian participants aged 45-49 years of the Northern Finland Birth Cohort Eye Study was examined using 24-2 SAP, optic nerve head (ONH) and retinal nerve fibre layer (RNFL) photography as well as SD-OCT of the peripapillary RNFL «Karvonen E, Stoor K, Luodonpää M ym. Combined stru...»3. The S-F report was generated by Forum Glaucoma Workplace software. OCT, SAP and the S-F analysis were evaluated against clinical glaucoma diagnosis, that is, the positive '2 out of 3' rule based on the clinician's evaluation of ONH and RNFL photographs and visual fields (VFs).
At a specificity of 98%, the sensitivity for glaucomatous damage was 26% for abnormal OCT, 35% for SAP and 44% for S-F analysis. Estimated areas under the curve were 0.74, 0.85 and 0.76, and the corresponding positive predictive values were 8 %, 10% and 12%, respectively. By applying a classification tree approach combining OCT, SAP and defect localisation data, a sensitivity of 77% was achieved at 90% specificity. In a localisation analysis of glaucomatous structural and functional defects, the correlation with glaucoma increased significantly if the abnormal VF test points were located on borderline or abnormal OCT zones.
SAP performs slightly better than OCT in glaucoma screening of middle-aged population. However, the diagnostic capability can be improved by S-F analysis.