Graefes Arch Clin Exp Ophthalmol (2013) 251:1841–1848 DOI 10.1007/s00417-013-2337-0
NEUROPHTHALMOLOGY
Effectiveness of averaging strategies to reduce variance in retinal nerve fibre layer thickness measurements using spectral-domain optical coherence tomography Berthold Pemp & Randy H. Kardon & Karl Kircher & Elisabeth Pernicka & Ursula Schmidt-Erfurth & Andreas Reitner
Received: 10 August 2012 / Revised: 11 March 2013 / Accepted: 21 March 2013 / Published online: 16 April 2013 # Springer-Verlag Berlin Heidelberg 2013
Abstract Background Automated detection of subtle changes in peripapillary retinal nerve fibre layer thickness (RNFLT) over time using optical coherence tomography (OCT) is limited by inherent image quality before layer segmentation, stabilization of the scan on the peripapillary retina and its precise placement on repeated scans. The present study evaluates image quality and reproducibility of spectral domain (SD)OCT comparing different rates of automatic real-time tracking (ART). Methods Peripapillary RNFLT was measured in 40 healthy eyes on six different days using SD-OCT with an eyetracking system. Image brightness of OCT with unaveraged single frame B-scans was compared to images using ART of 16 B-scans and 100 averaged frames. Short-term and dayto-day reproducibility was evaluated by calculation of intraindividual coefficients of variation (CV) and intraclass correlation coefficients (ICC) for single measurements as well as for seven repeated measurements per study day.
The authors have full control of all primary data, and agree to allow Graefe’s Archive for Clinical and Experimental Ophthalmology to review their data upon request. B. Pemp (*) : K. Kircher : U. Schmidt-Erfurth : A. Reitner Department of Ophthalmology, Medical University of Vienna, Währinger Gürtel 18-20, 1090 Vienna, Austria e-mail:
[email protected] R. H. Kardon Department of Ophthalmology & Visual Sciences, University of Iowa and Veterans Administration, Iowa City, IA, USA E. Pernicka Institute for Medical Statistics, Medical University of Vienna, Vienna, Austria
Results Image brightness, short-term reproducibility, and day-to-day reproducibility were significantly improved using ART of 100 frames compared to one and 16 frames. Short-term CV was reduced from 0.94±0.31 % and 0.91± 0.54 % in scans of one and 16 frames to 0.56±0.42 % in scans of 100 averaged frames (P≤0.003 each). Day-to-day CV was reduced from 0.98±0.86 % and 0.78±0.56 % to 0.53±0.43 % (P≤0.022 each). The range of ICC was 0.94 to 0.99. Sample size calculations for detecting changes of RNFLT over time in the range of 2 to 5 μm were performed based on intraindividual variability. Conclusion Image quality and reproducibility of mean peripapillary RNFLT measurements using SD-OCT is improved by averaging OCT images with eye-tracking compared to unaveraged single frame images. Further improvement is achieved by increasing the amount of frames per measurement, and by averaging values of repeated measurements per session. These strategies may allow a more accurate evaluation of RNFLT reduction in clinical trials observing optic nerve degeneration. Keywords Reproducibility of results . Optical coherence tomography . Optic nerve
Introduction A variety of neurologic and ophthalmic diseases is associated with changes in retinal nerve fibre layer (RNFL) thickness. Hence, there is considerable interest in studying this parameter in humans. On one hand, acute inflammatory and ischemic diseases of the optic nerve as well as elevated intracranial pressure induce peripapillary RNFL swelling. On the other hand, there are several diseases that cause atrophy of this tissue. A severe reduction of RNFL thickness
1842
is frequently seen in glaucoma and other optic neuropathies. In recent years, several studies have collected evidence for mean RNFL reduction in patients with multiple sclerosis (MS) [1]. This was shown most prominently after elapsed acute optic neuritis indicating localised axonal damage in the optic nerve after acute demyelinating inflammation. In addition, there is also evidence for decreased RNFL thickness in asymptomatic eyes of MS patients, but to a much lower degree suggesting possible subclinical episodes of optic neuritis or retrograde trans-synaptic degeneration [2]. Other recent reports have demonstrated reduced average RNFL thickness in neurodegenerative diseases including Alzheimer’s and Parkinson’s disease, which was interpreted as a sign of systemic neurodegeneration [3, 4]. However, the amount of change in these conditions is very small and may be observed only after a long period, which makes it difficult to measure a general reduction of RNFL thickness over time in individuals as well as in a study cohort. Nevertheless, the ability to detect these changes is critical for adequate disease management. Currently, the most accurate method for measuring RNFL thickness noninvasively is optical coherence tomography (OCT) [5, 6]. OCT is based on the different light reflection characteristics of individual ocular tissue layers, to determine their thickness by applying automated segmentation algorithms on the recorded images. There are several devices available, which have shown good reproducibility and repeatability in measurement of peripapillary RNFL thickness [6–12]. However, to reliably detect changes of only a few microns, further improvement of reproducibility would be beneficial. A reduction in variability of measurements would also result in a reduction of sample sizes necessary to detect small changes of mean RNFL thickness in patient groups with suspected neurodegenerative affection of the optic nerve. The aim of the present study was to evaluate short-term variability and day-to-day reproducibility of average peripapillary RNFL thickness measurements using the latest generation of spectral-domain OCT, which provides a number of new features designed to reduce measurement variability. We investigated the beneficial effects of these new scanacquisition features in healthy subjects, acquiring repeat scans on the same day and on different days. In addition, sample size calculations were performed for the detection of mean RNFL thickness changes over time of 2 to 5 μm, based on intraindividual variability data.
Material and methods The study was performed at the Department of Ophthalmology of the Medical University of Vienna after approval of the local ethics board, and adhered to the tenets of the Declaration of Helsinki. Twenty healthy subjects were included in this
Graefes Arch Clin Exp Ophthalmol (2013) 251:1841–1848
prospective cohort study after informed consent was obtained. Assessment of best-corrected visual acuity, slit-lamp biomicroscopy, funduscopy, and measurement of intraocular pressure (IOP) were performed in all subjects. Individuals with an ametropia ≥6 D, a history of retinal disease, abnormalities of the optic disc, or unclear ocular media were not included in the study. Spectral domain OCT OCT imaging was performed in both eyes of each subject using the Spectralis OCT plus (Heidelberg Engineering, Heidelberg, Germany). This fourth generation OCT system has a optical depth resolution of 7 μm and lateral resolution of 14 μm in tissue [13]. Proprietary noise reduction of the device reduces the axial A-scan resolution to a digital resolution of 3.5 μm/pixel. The transverse digital resolution depends on the amount of registered A-scans and can be adjusted. In high-resolution mode, which was used in the present study, a 360° peripapillary RNFL scan contains 1536 A-scans which provides a transverse digital resolution of about 5 μm in a standard eye. The device obtains up to 40.000 A-scans per second and simultaneously images the fundus with an infrared confocal scanning laser ophthalmoscope (SLO). The built-in real-time eye movement tracking system recognises features in the SLO image including blood vessels and the optic disc. This enables stabilisation of the scan coordinates in relation to the retina coordinates and minimises motion artefacts during image acquisition. If eye-tracking is activated, the system is able to average multiple repeated OCT scans, whereupon single B-scan frames are only added to the averaged OCT image if the scanning position matches the original position in the SLO image. After definition of a certain baseline scan as reference, the tracking system further allows image acquisition at the same retinal location throughout each follow-up examination. Latest improvements in scan acquisition with the device are: 1) automatic real-time tracking (ART) with averaging of up to 100 B-scan frames, which was compared to a previous default of 16 ART frames and to single frame images in the present study, and 2) follow-up scan comparison using a predefined reference scan as baseline, upon which subsequent scans are aligned during frame acquisition. This approach has recently been further refined in peripapillary scans to correct for rotation of the fundus image that can occur due to changes in head tilt during positioning. Peripapillary circle scans with a diameter of 12° were obtained in the high-resolution mode of the device. The subjects were asked to look at the nasally projected internal fixation light. After optimisation of fundus image quality and OCT scan height, the scanning circle was placed concentrically around the optic disc. This was done manually at baseline. In follow-up examinations: the built-in alignment technology
Graefes Arch Clin Exp Ophthalmol (2013) 251:1841–1848
automatically tracks and aligns circle scans including rotation orientation, which enables subsequent OCT scans of the same location as the prior scan with a high degree of accuracy. OCT scans were performed seven times on each study day. Scans composed of 16 ART frames were recorded on the first two study days, scans with 100 ART frames were recorded on days 3 and 4. To enable comparison of image properties with unaveraged images, additional OCT recordings with one image frame were performed on study days 5 and 6. The Spectralis software (version 5.3) which was used does not allow recording of single frame images without ART. The first scans of days 1, 3, and 5 were defined as reference scans after review for possible segmentation errors. All other images were acquired using the follow-up function of the device. The automatic segmentation algorithm of the Spectralis OCT software was used to obtain RNFL thickness values. The software calculates global average RNFL thickness and thickness of different sectors. In this study, global values were used for the assessment of variability and reproducibility. ImageJ software (Release 1.44p, NIH, Bethesda, MD, USA) was used to assess image quality. Raw image data were exported from the Spectralis software to avoid unwanted image processing such as normalization and smoothing. Brightness values of the RNFL and the ganglion cell layer (GCL) were measured in the unprocessed OCT images by manually selecting areas of both layers at the layer boundary of the temporal scan segment. This was done in corresponding areas of all images for each subject. The individual brightness histograms were analysed to detect the peak values of brightness distribution in the RNFL and the GCL at the layer edge. The brightness values were computed as arbitrary units (a.u.) on a 256-step scale from complete black pixels (0) to complete white pixels (1). To compare contrast of the different imaging modes at the layer edge of RNFL and GCL, individual luminance differences were calculated and compared. Statistical analysis Statistical analyses were performed using Statistica software (Release 4.5, StatSoft Inc., Tulsa, OK, USA). Data were expressed as means ± standard deviation (SD). Differences between the measurement algorithms were analysed using paired t-tests. A P-value of 0.05 was considered the level of significance. Short-term and day-to-day reproducibility of all three scanning modes were assessed by calculating intraindividual SD and coefficients of variation (CV), applying the formula CV = SD/mean × 100 %. Short-term variability of the three scanning modes was calculated from all seven readings of study days 1, 3, and 5, whereas day-to-day variability was evaluated from the first measurements of all study days for each scanning mode separately. In addition, intraclass correlation
1843
coefficients (ICC) were calculated. ICC calculation is based on a repeated measure ANOVA model using the variance among subjects, the variance among measurements, and the residual error variance [14]. An ICC of 1 represents perfect reproducibility. ICC was calculated from the first two measurements of the first, third, and fifth study day to evaluate short-term reproducibility of each ART mode. Day-to-day variability was calculated from the first measurements of each study day. Similarly, the reproducibility of averaged values from seven repeated measurements per study day was calculated. Sample size calculations were performed based on the intraindividual SD values for each method. These calculations were adjusted for an 80 % power to detect a difference in means of 2 to 5 μm using a paired t-test with a 0.05 two-sided significance level.
Results Four male and 16 female subjects with a mean age of 28± 7 years were included in this study. Measured eyes had a mean BCVA of 1.4±0.3, a spherical error of 1.1±1.8 D, and IOP of 16±2 mmHg. Mean RNFL thickness for the population of normal eyes tested was 97.91±8.67 μm using single measurements of single frames, 97.98 μm±8.59 μm using an average of seven measurements of single frames, 99.50±8.58 μm using single measurements with ART of 16 frames, 99.23± 8.33 μm using an average of seven measurements with 16 frames, 99.40±7.94 μm using single measurements with ART of 100 frames, and 99.15±7.80 μm using an average of seven measurements with 100 frames. RNFL thickness was measured significantly lower in recordings using single frames compared to 16 and 100 frames with all P-values<0.001, whereas there was no significant difference between measurements using 16 and 100 frames (P-value range 0.33–0.87). Sample measurements with OCT images of a single frame, with ART of 16 averaged frames and with ART of 100 frames at the same location of the same eye are presented in Fig. 1. The close-up images of the raw and unprocessed OCT images show enhanced brightness of RNFL and GCL and improved image quality by a considerable reduction of speckle noise using 100 averaged frames. The mean brightness values of RNFL and GCL at the layer boundary increased significantly using ART of 100 (RNFL: 0.175±0.094 a.u.; GCL: 0.017± 0.008 a.u.) compared to 16 frames (RNFL: 0.125±0.062 a.u., P=0.003; GCL: 0.012±0.007 a.u., P=0.001) and also compared to single image frames (RNFL: 0.061±0.033 a.u.; GCL: 0.007±0.004 a.u., P<0.001 each). The brightness values of images with 16 averaged frames were also significantly higher than in single-frame images (P<0.001 each). Similarly, the contrast at the layer edge of RNFL and GCL was significantly higher with ART of 100 frames (0.158±0.089 a.u.) compared
1844
to 16 (0.113±0.056, P=0.005) and single frames (0.055± 0.030, P<0.001). Intraindividual SD (μm) and CV (%) as well as ICC using the six different measurement methods are presented in Table 1. Short-term reproducibility of single measurements was improved using ART of 100 frames compared to single frames, as well as compared to 16 frames. This is reflected by significantly lower values for SD (P=0.001 and P=0.004 respectively, paired t-test) and CV (P<0.001 and P=0.003 respectively). The use of ART of 16 image frames did not result in a lower CV compared to single frames (P=0.605). Day-to-day reproducibility was further improved by averaging of seven repeated measurements per study day using the 100 frames mode, compared to single measurements with 16 frames or single frames, which is reflected by significant reduction of SD (P=0.002 and P=0.005 respectively) and CV (P=0.022 and P=0.006 respectively). ICC values ranged from 0.94 to 0.99. The best values were achieved using ART of 100 frames. Based on the variability data of all approaches, sample size calculations to detect a change over time of 2, 3, 4, and 5 μm are presented in Table 2. Sample sizes are given based on a longitudinal design using pairwise analysis, a doublesided α-error of 0.05, and a β-error of 0.20.
Discussion The results of the present study indicate improved image quality and reproducibility of RNFL thickness measurements using the latest software release of Spectralis OCT, with increased ART of up to 100 frames compared to the previous default maximum of ART of 16 frames as well as compared to single frame scans. This was mainly done to evaluate this device for future studies in which detection of small changes in peripapillary RNFL thickness over time is required. The results also indicate that both the increase of averaged frame number per scan, as well as averaging of repeated measurements, are successful methods for achieving better reproducibility. This is reflected by the fact that sample sizes for the detection of changes in mean RNFL thickness over time are markedly reduced using 100 averaged image frames. The improved reproducibility using more frames per measurement may be due to the additional reduction of speckle noise, the observed enhanced image brightness of the individual retinal layers, and the improved contrast between the individual retinal layers. Enhanced image contrast may facilitate border detection of the true individual retinal layers. Improved image properties as seen with multiple averaged frames may possibly also enhance layer segmentation in distinct parts of the scan that often show
Graefes Arch Clin Exp Ophthalmol (2013) 251:1841–1848 Fig. 1 Sample measurements of retinal nerve fibre layer thickness atb identical retinal location of the same eye using one single frame (top), 16 averaged frames (middle) and 100 averaged frames (bottom). Images were recorded using the follow-up function of Spectralis OCT. A considerable reduction of speckle noise and increased brightness of the retinal nerve fibre layer (RNFL) and ganglion cell layer (GCL) can be noted with icreasing image frames of averaged B-scans as seen in the brightness histograms of areas of RNFL and GCL at the layer boundary and also in the close-up image frame of the temporal sector. In this figure the usually processed OCT images of the Spectralis software are replaced by unprocessed images extracted from the raw image data
segmentation errors due to shadow artefacts, for instance scan sections beneath vessels. Repetition of determined RNFL thickness measurement on the other hand reduces the influence of single inaccurate measurements, and therefore also increases reproducibility. Averaging of multiple scans may therefore be an additional approach for minimising longterm variability, to detect very small changes over time. The automated thickness determination seems to be impaired in single frame images, because measurements without averaging produced significantly lower RNFL thickness values than both other imaging modes using averaged OCT images. This result of our study is of considerable importance, because it shows that measurements from single frames may not be comparable to averaged images. Whether the more reproducible values from averaged OCT images represent the true tissue thickness cannot be determined without histological comparison, which is, however, true for all currently available in-vivo imaging devices. To our knowledge, this is the first study comparing image properties and reproducibility of RNFL thickness results using various averaging levels with the Spectralis device. Earlier studies comparing RNFL thickness between several OCT devices have shown significant differences between the various machines, which may be caused rather by different measurement properties and data processing than by averaging [15]. A direct comparison of the commercially available devices with regard to their image properties including brightness and contrast could be interesting, but may not add much information about measurement reliability to the available reproducibility studies [6–12, 15]. The device used in this study has some advantages in measurement of RNFL thickness compared to other OCT platforms. Previous studies have shown that measurements using time-domain OCT have a higher test–retest variability, with intraindividual CV of about 3 to 5 % in normal and glaucomatous eyes [16, 17] and also directly compared to spectral domain OCT [18]. The application of time-domain technique in atrophic disease with small changes over time is therefore limited. Other studies have tested the reproducibility of different spectral domain OCT machines and showed better results using this newer technique with reported CVs of about 1 to 3 % [8–11, 15, 19]. Recently, the same device as used in the present study was evaluated
Graefes Arch Clin Exp Ophthalmol (2013) 251:1841–1848
1845
1846
Graefes Arch Clin Exp Ophthalmol (2013) 251:1841–1848
Table 1 Intraindividual mean standard deviation, intraindividual coefficients of variation and intraclass correlation coefficients for intrasession and intersession variability Image frames
1 16 100
Number of measurements
Short term variability (Day 1)
Single Mean of seven Single Mean of seven Single Mean of seven
Day to day variability
SDi
CVi
ICC
SDi
0.91±0.29 μm – 0.91±0.58 μm – 0.56±0.42 μm –
0.94±0.31 % – 0.91±0.54 % – 0.56±0.42 % –
0.98 – 0.97 – 0.98 –
0.96±0.80 0.81±0.65 0.78±0.59 0.69±0.71 0.64±0.53 0.53±0.44
a
a
CVi μm μm μm μm μm μm
a
0.98±0.86 0.83±0.66 0.78±0.56 0.69±0.65 0.65±0.52 0.53±0.43
ICC % % % % % %
a
0.94 0.98 0.97 0.98 0.99 0.96
a significant difference compared to single measurements with one frame and 16 frames (P<0.006, paired t-test). SDi intraindividual standard deviation, CVi intraindividual coefficients of variation, ICC intraclass correlation coefficients
with regard to its reproducibility and repeatability [12], indicating that Spectralis OCT currently has the best reproducibility in RNFL measurements, with a CVof 1.4 % and an ICC of Table 2 Sample size calculations for the detection of changes of 2 to 5 μm over time using different measurement modes of Spectralis OCT Detectable changes 2 μm
Image frames 1 16 100
3 μm
1 16 100
4 μm
1 16 100
5 μm
1 16 100
Number of measurements
Sample size
1 7 1 7 1
180 155 180 155 145
7 1 7 1 7 1 7 1 7 1 7 1 7 1 7 1 7 1
145 80 70 80 70 65 65 50 40 50 40 40 40 30 30 30 30 25
7
25
Calculations were adjusted for an 80 % power to detect a difference in means of 2 to 5 μm using a paired t-test with a 0.05 two-sided significance level. Results were rounded up to the nearest 5
0.98. However, the mentioned studies did not apply eye tracking and reference scan definition for repeat measurements. With CV values ranging from 0.53 % to 0.98 %, our study reports even better values of repeatability for spectraldomain OCT in healthy subjects than previous findings, and in particular for the Spectralis device. This may be partly due to eye tracking and reference scan alignment, which was also used in single frame scans. However, even lower intraindividual variation is found after introduction of the new measurement algorithm using 100 averaged image frames compared to measurements with 16 and single frames. ICC of all measurement strategies was very high, which reflects excellent reproducibility of the used OCT device in all imaging modes, but slightly favours measurements with high frame rate. ICC is a generally accepted measure of reliability and is considered more appropriate for concordant measures than other methods [13]. However, it is also population-specific, and may vary from one population to another. The results of our study clearly show an advantage of tracked and multiple averaged images compared to single scans. This also indicates an advantage of the investigated methods with regard to the detection of small changes in mean peripapillary RNFL thickness, compared to devices that do not feature such processing. This means that the averaging strategies which are used may improve reliability of detection of slowly progressive optic neuropathy in different diseases including glaucoma and other diseases of the optic nerve, in an experimental but also in a clinical setting. This study focuses on evaluation of mean RNFL thickness. Whether the applied procedures also enhance reproducibility of sectorial RNFL thickness values was not tested, but seems very likely. In addition to ocular disease, different systemic neurologic disorders have been investigated in recent years with regard to their effects on optic nerve fibres. For example, it has been shown that RNFL thickness was reduced in MS patients after acute ON [1] but also in patients without an episode of ON [20, 21]. Monitoring of RNFL thickness in
Graefes Arch Clin Exp Ophthalmol (2013) 251:1841–1848
MS may be of prognostic value because this parameter is associated with visual impairment [20, 22–24], brain atrophy [25], and also with disability scores [26]. In addition OCT measurement of the RNFL may be a useful additional surrogate marker for the detection of general axonal loss in other neurodegenerative diseases including Alzheimer’s or Parkinson’s disease [3, 4]. Since the amount of RNFL reduction over time found in these pathologies is very small (circa 1 micron per year on average), only a method with optimal reproducibility and repeatability is useful in detecting progredient neurodegeneration of the optic nerve over a relatively small period of time. In addition, neuroprotective effects of different therapeutic approaches could be tested more reliably in clinical trials. The present study measures RNFL thickness in healthy eyes only. Reproducibility in patients with reduced RNFL thickness may be lower. Earlier results in glaucoma patients indicate that the reproducibility of RNFL measurements is not worse in eyes with reduced RNFL thickness using spectraldomain OCT [9–12, 27]. However, a recent study found lower reproducibility of spectral-domain OCT measurements in MS patients compared to healthy subjects [28]. In this study using Spectralis OCT, intraindividual CV was markedly improved after application of average ART of nine frames, which indicates a beneficial effect of OCT image enhancement in patients with RNFL atrophy similar to our results in healthy subjects. A similar improvement could also be possible in swollen RNFL, which however remains to be shown. In conclusion, this study shows that short-term and day-today reproducibility of mean peripapillary RNFL thickness measurements with spectral-domain OCT using eye tracking and reference baseline scans for subsequent scan alignment can be further improved by increasing the number of averaged ART frames per scan image, as well as by averaging values of repeated measurements per session. These strategies may allow a more accurate evaluation of RNFLT reduction in clinical trials observing neurodegenerative disease, and reduce sample sizes necessary to detect small changes of this parameter over time.
References 1. Sergott RC (2007) Historical perspective and future prospective for retinal nerve fiber loss in optic neuritis and multiple sclerosis. Int Ophthalmol Clin 47:15–24 2. Albrecht P, Ringelstein M, Mueller A, Keser N, Dietlein T, Lappas A, Foerster A, Hartung H, Aktas O, Methner A (2012) Degeneration of retinal layers in multiple sclerosis subtypes quantified by optical coherence tomography. Mult Scler 18(10):1422–1429. doi:10.1177/ 1352458512439237 3. Parisi V, Restuccia R, Fattapposta F, Mina C, Bucci MG, Pierelli F (2011) Morphological and functional retinal impairment in Alzheimer’s disease patients. Clin Neurophysiol 112:1860–1867
1847 4. Rohani M, Langroodi AS, Ghourchian S, Falavarjani KG, Soudi R, Shahidi G (2012) Retinal nerve changes in patients with tremor dominant and akinetic rigid Parkinson’s disease. Neurol Sci Jun 3 [Epub ahead of print]. doi:10.1007/s10072-012-1125-7 5. Schuman JS, Pedut-Kloizman T, Hertzmark E, Hee MR, Wilkins JR, Coker JG, Puliafito CA, Fujimoto JG, Swanson EA (1996) Reproducibility of nerve fiber layer thickness measurements using optical coherence tomography. Ophthalmology 103:1889–1898 6. Knight OJ, Chang RT, Feuer WJ, Budenz DL (2009) Comparison of retinal nerve fiber layer measurements using time-domain and spectral-domain optical coherence tomography. Ophthalmology 116:1271–1277 7. Budenz DL, Chang RT, Huang X, Knighton RW, Tielsch JM (2005) Reproducibility of retinal nerve fiber thickness measurements using the stratus OCT in normal and glaucomatous eyes. Invest Ophthalmol Vis Sci 46:2440–2443 8. Menke MN, Knecht P, Sturm V, Dabov S, Funk J (2008) Reproducibility of nerve fiber layer thickness measurements using 3D fourier-domain OCT. Invest Ophthalmol Vis Sci 49:5386–5391 9. González-García AO, Vizzeri G, Bowd C, Medeiros FA, Zangwill LM, Weinreb RN (2009) Reproducibility of RTVue retinal nerve fiber layer thickness and optic disc measurements and agreement with Stratus optical coherence tomography measurements. Am J Ophthalmol 147:1067–1074 10. Vizzeri G, Weinreb RN, Gonzalez-Garcia AO, Bowd C, Medeiros FA, Sample PA, Zangwill LM (2009) Agreement between spectraldomain and time-domain OCT for measuring RNFL thickness. Br J Ophthalmol 93:775–781 11. Kim JS, Ishikawa H, Sung KR, Xu J, Wollstein G, Bilonick RA, Gabriele ML, Kagemann L, Duker JS, Fujimoto JG, Schuman JS (2009) Retinal nerve fibre layer thickness measurement reproducibility improved with spectral domain optical coherence tomography. Br J Ophthalmol 93:1057–1063 12. Wu H, de Boer JF, Chen TC (2011) Reproducibility of retinal nerve fiber layer thickness measurements using spectral domain optical coherence tomography. J Glaucoma 20:470–476 13. Castro Lima V, Rodrigues EB, Nunes RP, Sallum JF, Farah ME, Meyer CH (2011) Simultaneous confocal scanning laser ophthalmoscopy combined with high-resolution spectral-domain optical coherence tomography: a review. J Ophthalmol 2011:743670 14. Kramer MS, Feinstein AR (1981) Clinical biostatistics. LIV. The biostatistics of concordance. Clin Pharmacol Ther 29:111–123 15. Pierro L, Gagliardi M, Iuliano L, Ambrosi A, Bandello F (2012) Retinal nerve fiber layer thickness reproducibility using seven different OCT instruments. Invest Ophthalmol Vis Sci 53:5912–5920 16. Budenz DL, Fredette MJ, Feuer WJ, Anderson DR (2008) Reproducibility of peripapillary retinal nerve fiber thickness measurements with stratus OCT in glaucomatous eyes. Ophthalmology 115:661–666 17. Tzamalis A, Kynigopoulos M, Schlote T, Haefliger I (2009) Improved reproducibility of retinal nerve fiber layer thickness measurements with the repeat-scan protocol using the Stratus OCT in normal and glaucomatous eyes. Graefes Arch Clin Exp Ophthalmol 247:245–252 18. Leung CK, Cheung CY, Weinreb RN, Qiu Q, Liu S, Li H, Xu G, Fan N, Huang L, Pang CP, Lam DS (2009) Retinal nerve fiber layer imaging with spectral-domain optical coherence tomography: a variability and diagnostic performance study. Ophthalmology 116:1257–1263 19. Mwanza JC, Chang RT, Budenz DL, Durbin MK, Gendy MG, Shi W, Feuer WJ (2010) Reproducibility of peripapillary retinal nerve fiber layer thickness and optic nerve head parameters measured with cirrus HD-OCT in glaucomatous eyes. Invest Ophthalmol Vis Sci 51:5724–5730 20. Talman LS, Bisker ER, Sackel DJ, Long DA Jr, Galetta KM, Ratchford JN, Lile DJ, Farrell SK, Loguidice MJ, Remington G, Conger A, Frohman TC, Jacobs DA, Markowitz CE, Cutter GR,
1848
21.
22.
23.
24.
Ying GS, Dai Y, Maguire MG, Galetta SL, Frohman EM, Calabresi PA, Balcer LJ (2010) Longitudinal study of vision and retinal nerve fiber layer thickness in multiple sclerosis. Ann Neurol 67:749–760 Garcia-Martin E, Pueyo V, Martin J, Almarcegui C, Ara JR, Dolz I, Honrubia FM, Fernandez FJ (2010) Progressive changes in the retinal nerve fiber layer in patients with multiple sclerosis. Eur J Ophthalmol 20:167–173 Costello F, Hodge W, Pan YI, Eggenberger E, Coupland S, Kardon RH (2008) Tracking retinal nerve fiber layer loss after optic neuritis: a prospective study using optical coherence tomography. Mult Scler 14:893–905 Fisher JB, Jacobs DA, Markowitz CE, Galetta SL, Volpe NJ, NanoSchiavi ML, Baier ML, Frohman EM, Winslow H, Frohman TC, Calabresi PA, Maguire MG, Cutter GR, Balcer LJ (2006) Relation of visual function to retinal nerve fiber layer thickness in multiple sclerosis. Ophthalmology 113:324–332 Henderson AP, Trip SA, Schlottmann PG, Altmann DR, GarwayHeath DF, Plant GT, Miller DH (2008) An investigation of the retinal nerve fibre layer in progressive multiple sclerosis using optical coherence tomography. Brain 131:277–287
Graefes Arch Clin Exp Ophthalmol (2013) 251:1841–1848 25. Gordon-Lipkin E, Chodkowski B, Reich DS, Smith SA, Pulicken M, Balcer LJ, Frohman EM, Cutter G, Calabresi PA (2007) Retinal nerve fiber layer is associated with brain atrophy in multiple sclerosis. Neurology 69:1603–1609 26. Sepulcre J, Murie-Fernandez M, Salinas-Alaman A, GarcíaLayana A, Bejarano B, Villoslada P (2007) Diagnostic accuracy of retinal abnormalities in predicting disease activity in MS. Neurology 68:1488–1494 27. Hood DC, Anderson SC, Wall M, Raza AS, Kardon RH (2009) A test of a linear model of glaucomatous structure-function loss reveals sources of variability in retinal nerve fiber and visual field measurements. Invest Ophthalmol Vis Sci 50:4254–4266 28. Garcia-Martin E, Pueyo V, Pinilla I, Ara JR, Martin J, Fernandez J (2011) Fourier-domain OCT in multiple sclerosis patients: reproducibility and ability to detect retinal nerve fiber layer atrophy. Invest Ophthalmol Vis Sci 52:4124–4131
This study was not sponsored by an organization, therefore no financial relationship exists for any author. The authors have no proprietary or financial interest in any aspect of this study.