I read with great interest the article by O’Connell et al. on reliability testing of retinal oximetry with the Oxymap oximeter, which appeared in the June 2014 issue of Optometry and Vision Science. The Oxymap oximeter and accompanying software, which I took part in developing and testing, automatically detects retinal vessels and calculates oxygen saturation at every pixel along the detected vessel segments. The authors describe a very interesting approach to the analysis of these oximetry images. In short, they take every pixel detected by the software and use statistical methods to obtain saturation values for categories of pixels corresponding to different diameters. Their method, for example, reports one saturation value for all pixels, which belong to cross sections of venules of diameters 60 to 70 micrometers, one value for venules of 70 to 80 micrometers, and so on (up to 170 to 180 micrometers and the same for arterioles). The authors test the reliability of this approach and compare the results with my group’s previous results, which were obtained with different analysis methods. I think this approach is highly interesting and could potentially be developed further. However, there are several issues that need to be resolved first and I disagree with several statements in the article. Below are my concerns.

Purpose. To determine intrasession and intersession repeatability of retinal vessel oxygen saturation from the Oxymap Retinal Oximeter using a whole image-based analysis technique and so determine optimal analysis parameters to reduce variability.Methods. Ten fundus oximetry images were acquired through dilated pupils from 18 healthy participants (aged 22 to 38) using the Oxymap Retinal Oximeter T1. A further 10 images were obtained 1 to 2 weeks later from each individual. Analysis was undertaken for subsets of images to determine the number of images needed to return a stable coefficient of variation (CoV). Intrasession and intersession variability were quantified by evaluating the CoV and establishing the 95% limits of agreement using Bland and Altman analysis. Retinal oxygenation was derived from the distribution of oxygenation values from all vessels of a given width in an image or set of images, as described by Paul et al. in 2013.Results. Grouped in 10-mu m-wide bins, oxygen saturation varied significantly for both arteries and veins (p < 0.01). Between 110 and 150 mu m, arteries had the least variability between individuals, with average CoVs less than 5% whose confidence intervals did not overlap with the greater than 10% average CoVs for veins across the same range. Bland and Altman analysis showed that there was no bias within or between recording sessions and that the 95% limits of agreement were generally lower in arteries.Conclusions. Retinal vessel oxygen saturation measurements show variability within and between clinical sessions when the whole image is used, which we believe more accurately reflects the true variability in Oxymap images than previous studies on select image segments. Averaging data from vessels 100 to 150 Km in width may help to minimize such variability.

Ocular disease can both alter the retina's oxygen requirements, and decrease its ability to cope with changes in metabolic demand. We examined the influence of a moderate intraocular pressure (IOP) elevation on three outcome measures: arterial and venous oxygen saturation, blood flow, and the pattern electroretinogram (PERG). We increased IOP to ˜30 mmHg in 23 healthy participants (22–39 years) using a mechanical probe applied to the eyelid, thereby lowering ocular perfusion pressure (OPP) by 30

Purpose. To use a novel image analysis approach to consider how oxygen saturation changes as a function of vessel width and distance from the nerve and between superior and inferior retinal hemifields.Methods. Ten images were acquired from one eye of 17 participants (mean [standard deviation] age, 28 [4] years; range, 22-38 years) using the Oxymap T1 retinal oximeter. Every pixel identified by the detection algorithm was extracted, and frequency histograms of retinal vessel oxygen saturation were plotted for each vessel diameter (70-170 mu m). Histograms were fitted with two Gaussian models to identify peak arteriole and venule oxygen saturation. Mean (+/- standard error of the mean) arteriole and venule oxygen saturation at each vessel width were calculated. Data were also analyzed in (1) annuli of 100 mu m centered on the optic nerve or (2) upper and lower hemifields demarcated by the center of the optic nerve.Results. Venous oxygen saturation was higher in smaller vessels than in larger vessels. Arterial oxygen saturation remained relatively constant with vessel width. Oxygen saturation was lower in veins nearer the optic nerve. The upper retinal hemisphere showed higher venous oxygen saturation compared with the lower hemifield.Conclusions. The current objective analysis approach provides a more complete picture of retinal oxygen saturation at the posterior pole as a function of vessel width and retinal location.

Background: To evaluate the effect of phacoemulsification on trabeculectomy function.Design: Retrospective case-control study.Participants: Forty-eight patients who underwent trabeculectomy surgery and had at least 2 years of follow up.Methods: Patients were classified into two groups: patients who had phacoemulsification subsequent to trabeculectomy (Trab_phaco, n = 18) and patients who were pseudophakic for greater than 6 months preceding their trabeculectomy (Phaco_trab, n = 30). Groups were matched for length of follow up of 2 years from time of trabeculectomy.Main Outcome Measures: The primary outcome measures were target intraocular pressure of criteria A, <= 12 mmHg; B, <= 15 mmHg; C, <= 18 mmHg with or without additional topical treatment. A separate measure for bleb function failure was also used; with failure defined as the need for additional topical antiglaucoma therapy or surgical intervention to achieve control of intraocular pressure.Results: There was no significant difference in achieving each intraocular pressure criterion between groups (12 months, P = 1.0; 24 months, P = 0.330). In the first 12 months, significantly more trabeculectomies in the Trab_phaco group failed, requiring additional intervention to control the IOP (39%) compared with the Phaco_trab (10%) group (P = 0.028). Although this trend continued at 24 months, there were no significant differences in failure rates (P = 0.522).Conclusions: Phacoemulsification performed after trabeculectomy significantly increased rates of bleb failure in the following 12 months but not at 24 months.

Purpose To use a novel image analysis approach to consider how oxygen saturation changes as a function of vessel width and distance from the nerve and between superior and inferior retinal hemifields. Methods Ten images were acquired from one eye of 17 participants (mean [standard deviation] age, 28 [4] years; range, 22–38 years) using the Oxymap T1 retinal oximeter. Every pixel identified by the detection algorithm was extracted, and frequency histograms of retinal vessel oxygen saturation were plotted for each vessel diameter (70–170 μm). Histograms were fitted with two Gaussian models to identify peak arteriole and venule oxygen saturation. Mean (±standard error of the mean) arteriole and venule oxygen saturation at each vessel width were calculated. Data were also analyzed in (1) annuli of 100 μm centered on the optic nerve or (2) upper and lower hemifields demarcated by the center of the optic nerve. Results Venous oxygen saturation was higher in smaller vessels than in larger vessels. Arterial oxygen saturation remained relatively constant with vessel width. Oxygen saturation was lower in veins nearer the optic nerve. The upper retinal hemisphere showed higher venous oxygen saturation compared with the lower hemifield. Conclusions The current objective analysis approach provides a more complete picture of retinal oxygen saturation at the posterior pole as a function of vessel width and retinal location.