The RTVue SDOCT with the corneal lens module attachment demonstrated poor to average detection of the SS, as judged against the ASOCT (26.9% vs. 69.1%, respectively). The SS is the key landmark for quantitative and qualitative angle assessment. Its major advantage in this study over the ASOCT was in its unique ability to detect the SL in 44.1% of subjects. In contrast, Wong et al. report in vivo visualization of the SS in almost 80% and of the SL in approximately 90% of quadrants with the use of another SDOCT device, the Cirrus.
11 The difference between the Cirrus and the RTVue in the ability to detect ACA landmarks in the two studies could be due to a number of factors.
11 First, an RTVue image is the product of 16 processed and averaged images, while the Cirrus (and the ASOCT for that matter) is a single-frame image. Alternatively, differing proprietary methods of reducing signal-to-noise ratios could be to blame. It could be technician related; however, the single technician who performed the imaging in this study has extensive experience with all manner of anterior segment imaging devices. The most important factor to consider is the motion artifact caused by ocular movements that may lead to an indistinct image depending on the scan time. The use of an external fixation target for directing the subject's gaze cannot be standardized, and the images acquired might not have been centered similarly in all instances, leading to edge artifact and unclear images. There are also two good methodological reasons why the two SDOCT studies have different detection rates for the angle structures. Wong et al. have smaller numbers and only look at the more easily visible horizontal meridians. In that study, 17/90 (18.8%) have one angle closed. In our study, 122/324 (37.7%) had one angle closed. It is therefore unsurprising that the detection of SL and the TM may have been more difficult when more subjects with closed angles were included in the case mix.
Wylegala et al. image 54 eyes using the RTVue and the Visante ASOCT and analyze them both qualitatively and quantitatively.
12 They report good correlation between anterior segment parameters, including central corneal thickness, trabecular iris angle, and angle opening distance. They also report that all anterior chamber angle structures (SS, TM, and Schlemm's canal) are visible with the RTVue; however, they do not state if they achieve this in all images or how many images have poor visibility of the ACA structures.
The depth and width of view provided by SDOCT does not allow imaging of the entire angle up to the iris root; hence the entire angle cannot be viewed in a single image as in the ASOCT. Gonioscopy can be used to assess the entire circumference of the angle, whereas these imaging devices only deliver a cross-sectional view of the angle. A subject with a gonioscopically predominantly closed angle in a quadrant with a small open segment would have been erroneously classified as open if that meridian were chosen for imaging. In addition, angle features such as peripheral anterior synechiae, iris strands, and pigment cannot be determined. There is also an inherent limitation in currently available ASOCT devices for imaging vertical quadrants. This is due to poorer rates of detection of the SS in images obtained in the superior and inferior quadrants compared with images of the horizontal quadrants.
5
We found that the agreement between the two devices and gonioscopy was moderate when only one quadrant out of two was used to define closed angles; however, this result should be interpreted with caution since only 35 eyes imaged by the SDOCT and 56 imaged by the ASOCT were analyzed. This was because only the horizontal meridians were used for determining agreement. Although this is a commonly used and time-saving protocol for anterior segment imaging, it must be remembered that this is not a complete four-quadrant scan, which would be the ideal approach to decide the status of the angle in an eye. Furthermore, the CAM-L module is an optional attachment for the RTVue, produced by the same manufacturer. It is possible that a fully integrated device may work better in practice. Our cited value of 0.60 using AC1 statistics for agreement between RTVue and gonioscopy compares favorably with the κ = 0.63 between Cirrus and gonioscopy calculated by Wong et al.11
This study has a few limitations. As stated earlier, the angle imaging was performed by directing the subject's gaze using an external fixation light to center the iridocorneal angle in the instrument's field of view. This might have led to inadequate image centration in the SDOCT in contrast to the ASOCT, which has an external fixation device incorporated. We used a 3-mm scan length as recommended by the manufacturer while imaging the angle with the SDOCT; however, we do not know if the 6-mm scan would provide more useful information about the same area. Since a single observer performed the gonioscopy, there could have been a systematic bias in the angle grading. In addition, differences in testing conditions exist for gonioscopy and for the imaging devices.
The imaging devices use infrared light and do not require contact with the eye. Inadvertent indentation might have occurred during gonioscopy despite efforts to avoid this. Similarly, even though ambient light is kept to a minimum during gonioscopy, the light may have led to a marginal degree of pupil constriction and angle opening. Hence, gonioscopy may have missed eyes with irido-angle contact. The height of irido-angle contact used to define a closed angle is also different with each technique. With imaging devices like the SDOCT and ASOCT, a closed angle is defined as any contact between the iris and angle wall anterior to the SS, but if this contact did not occur up to the level of the TM, the angle would have been graded as open on gonioscopy. Thus, it is also possible that imaging may have overdetected eyes with closed angles leading to more false positives. In images where the SS was not detected, an angle with any contact between the TM (when identified) and the iris was considered closed. This inconsistency of definitions used for closed angles would not only impact the incomparability of alternative SDOCT devices at detecting angle closure compared with the RTVue but also would mean a more lenient definition of closed angles had been used for the SDOCT device. Regarding the SDOCT images, excluding the large number of poor quality scans would have made it more difficult for the device to detect the true closed quadrants from an already low potential of 122/324 quadrants available on gonioscopy. Reproducibility of detection is important, especially for difficult-to-detect structures. However, the reproducibility of the RTVue at detecting structures such as Schlemm's canal was not measured in this study. In a separate study using AC1 statistics, we found the inter-observer agreement for identification of the Schlemm's canal using the iVue (Optovue) varied from 0.522 to 0.970 depending on the quadrant investigated, indicating significant variability (Quek, Tun, Narayanaswamy, et al. unpublished data, 2011). The iVue is an equivalent, more compact version of the RTVue device, from the same manufacturer, which has integrated anterior and posterior segment scanning capabilities.
In summary, the RTVue SDOCT device was found to have limited ability to image angle structures such as the TM, SS, and SL. This made assessment of angle closure status impossible in about 60% of angle quadrants. For quadrants in which closed angles could be assessed, there was moderate agreement with gonioscopy.