Purpose: To characterize preperimetric retinal nerve fiber layer (RNFL) defects on speckle noise-reduced spectral-domain optical coherence tomography (SD-OCT), and to determine whether detection of preperimetric RNFL defects can be improved by speckle noise reduction.
Patients and methods: Thirty-two eyes of 32 patients with preperimetric glaucoma and 30 normal eyes of 30 volunteers underwent complete ophthalmic examinations and scanning by speckle noise-reduced SD-OCT (Spectralis), single-scan SD-OCT (RTVue-100), and single-scan time-domain (Stratus) OCT.
Results: All 40 RNFL defects identified by photography had angular widths <30 degrees and no disruption of RNFL reflectivity on Spectralis. Circumpapillary RNFL (cpRNFL) boundaries were accurately determined by Spectralis for 38 (95.0%) of the 40 defects, by RTVue-100 for 25 (62.5%), and by Stratus OCT for 23 (57.5%). Sensitivity for the detection of RNFL defects (cpRNFL thinning to <1% of normal for an age-matched database) was 15% for Stratus, 42.5% for RTVue, and 47.5% for Spectralis on sector maps and significantly higher for SD-OCT on temporal-superior-nasal-inferior-temporal (TSNIT) thickness graphs: RTVue-100 (57.5%; P=0.031) and Spectralis (90.0%; P=0.0001). On the basis of TSNIT thickness graphs, sensitivity for the detection of RNFL defects was significantly higher for Spectralis compared with RTVue-100 (P=0.001) and Stratus (P=0.0005). Spectralis TSNIT graphs were more sensitive (P=0.001) for glaucoma detection without significant reduction (P=0.125) in specificity compared with Spectralis sector maps.
Conclusions: Our results suggest that accurate measurement of cpRNFL thickness by speckle noise-reduced SD-OCT and a comparison of the results with normative database using TSNIT graphs are required to improve the sensitivity for detecting preperimetric RNFL defects.