Research Article: Factors Affecting the Ability of the Spectral Domain Optical Coherence Tomograph to Detect Photographic Retinal Nerve Fiber Layer Defects

Date Published: December 23, 2014

Publisher: Public Library of Science

Author(s): Harsha L. Rao, Uday K. Addepalli, Ravi K. Yadav, Nikhil S. Choudhari, Sirisha Senthil, Chandra S. Garudadri, Ted S. Acott.

http://doi.org/10.1371/journal.pone.0116115

Abstract

To evaluate the ability of normative database classification (color-coded maps) of spectral domain optical coherence tomograph (SDOCT) in detecting wedge shaped retinal nerve fiber layer (RNFL) defects identified on photographs and the factors affecting the ability of SDOCT in detecting these RNFL defects.

In a cross-sectional study, 238 eyes (476 RNFL quadrants) of 172 normal subjects and 85 eyes (103 RNFL quadrants with wedge shaped RNFL defects) of 66 glaucoma patients underwent RNFL imaging with SDOCT. Logistic regression models were used to evaluate the factors associated with false positive and false negative RNFL classifications of the color-coded maps of SDOCT.

False positive classification at a p value of <5% was seen in 108 of 476 quadrants (22.8%). False negative classification at a p value of <5% was seen in 16 of 103 quadrants (15.5%). Of the 103 quadrants with RNFL defects, 64 showed a corresponding VF defect in the opposite hemisphere and 39 were preperimetric. Higher signal strength index (SSI) of the scan was less likely to have a false positive classification (odds ratio: 0.97, p = 0.01). Presence of an associated visual field defect (odds ratio: 0.17, p = 0.01) and inferior quadrant RNFL defects as compared to superior (odds ratio: 0.24, p = 0.04) were less likely to show false negative classifications. Scans with lower signal strengths were more likely to show false positive RNFL classifications, and preperimetric and superior quadrant RNFL defects were more likely to show false negative classifications on color-coded maps of SDOCT.

Partial Text

Glaucoma is a progressive optic neuropathy characterized by typical optic disc and retinal nerve fiber layer (RNFL) changes, with or without visual field (VF) defects. RNFL loss is a very early clinical sign of glaucoma, which is reported to be present in majority of glaucoma patients much before any detectable VF defects. [1], [2] Red free fundus photography is currently considered the gold standard for RNFL examination and loss of RNFL is seen as a wedge shaped defect in arcuate pattern, the width of which is larger than the major blood vessel, and reaching the edge of the disc margin. [3] Other techniques available for RNFL evaluation are scanning laser polarimetry and optical coherence tomography (OCT). Spectral domain OCT (SDOCT) is a recent generation of OCT technology which has shown good ability to detect RNFL defects in glaucoma. [4]–[8] However, the ability of SDOCT to detect RNFL defects is influenced by multiple factors. These factors can in general be disease-related, subject- or eye-related and test- or technology-related. Disease-related factor known to influence the ability of SDOCT to detect RNFL defects is the size of the defect. Ability of SDOCT is shown to be better in detecting wider compared to narrower RNFL defects. [4]–[8] Eye-related factors reported to influence the ability of SDOCT to detect RNFL defects are the optic disc size and axial length of the eye. [8], [9] One of the technology-related factors evaluated, but found not to influence the ability of SDOCT to detect RNFL defects is the signal strength of the scan [8].

This was an observational, cross-sectional study of consecutive subjects referred by general ophthalmologists to a tertiary eye care facility between September 2010 and November 2012 for a glaucoma evaluation. All subjects were of Indian origin. Written informed consent was obtained from all subjects to participate in the study and the Institutional Review Board of L V Prasad Eye Institute approved the methodology. All methods adhered to the tenets of the Declaration of Helsinki for research involving human subjects.

Descriptive statistics included mean and standard deviation for normally distributed variables and median and inter-quartile range (IQR) for non-normally distributed variables.

Six hundred and seventy eight eyes of 382 consecutive subjects referred for glaucoma evaluation to our center were analyzed. Forty two eyes with unreliable VFs and 7 eyes with poor quality disc photographs were excluded. Further, 12 eyes with segmentation algorithm failure and 10 eyes with SSI<30 on ONH scans were excluded. Of the remaining eyes, 71 eyes classified as either optic disc or RNFL suspect by either or both the experts, and 51 eyes with optic disc classification as normal and VF classification as glaucoma were also excluded, leaving 485 eyes for the current analysis. Of these, 238 eyes of 172 subjects with optic disc, RNFL and VF classification as “non-glaucoma” formed the control group. Of the remaining 247 eyes with optic disc classification as glaucoma, localized RNFL defects were identified by experts in 85 eyes of 66 subjects, which formed the case group. The initial agreement between glaucoma experts for the overall optic disc classification was 92.7% (kappa = 0.63). The initial agreement between experts for RNFL defect identification was 89.2% (kappa = 0.42). Remaining RNFL defect identification was by consensus. Table 1 shows the demographic, spherical equivalent refraction, VF and SDOCT parameters of the two groups. All VF parameters were significantly different in the glaucoma compared to the control group. Glaucoma patients had significantly smaller optic discs than the control subjects. SSI values were comparable between the two groups. Two hundred and thirty eight control eyes contributed 476 RNFL quadrants (238 superior and 238 inferior) and 85 glaucoma eyes contributed 103 RNFL quadrants with defects (56 superior and 47 inferior RNFL defects) for the analysis. Table 2 shows the diagnostic categorization of the corresponding RNFL quadrants on SDOCT. False positive classification at a p value of <1% was seen in 31 of 476 quadrants (6.5%) and at a p value of <5% was seen in 108 of 476 quadrants (22.8%). False negative classification at a p value of <1% was seen in 31 of 103 quadrants (30.1%) and at a p value of <5% was seen in 16 of 103 quadrants (15.5%). Figs 1 and 2 show examples of false positive and negative RNFL defect classifications respectively on the color-coded maps of SDOCT. Table 3 shows the diagnostic categorization of the corresponding RNFL quadrants on SDOCT separately in glaucoma eyes with and without corresponding VF defects. SDOCT identified the RNFL defect better when the RNFL defect was associated with a corresponding VF defect (p = 0.004, Chi square test). Of the 103 quadrants with RNFL defects, 64 showed a corresponding VF defect in the opposite hemisphere and 39 RNFL defects were preperimetric. In this study to evaluate the misclassification rates of the internal normative database classification of RTVue in detecting photographic wedge shaped RNFL defects in an Indian population, we found that the false positive rate was 22.8% and the false negative rate was 15.5% at a p value of <5%.   Source: http://doi.org/10.1371/journal.pone.0116115