To measure the magnitude and direction of anterior scleral canal opening (ASCO) offset relative to the Bruch membrane opening (BMO) (ASCO/BMO offset) to characterize neural canal obliqueness and minimum cross-sectional area (NCMCA) in 69 highly myopic and 138 healthy, age-matched, control eyes.
Using optical coherence tomography (OCT) scans of the optic nerve head (ONH), BMO and ASCO were manually segmented and their centroids and size and shape were calculated. ASCO/BMO offset magnitude and direction were measured after projecting the ASCO/BMO centroid vector onto the BMO plane. Neural canal axis obliqueness was defined as the angle between the ASCO/BMO centroid vector and the vector perpendicular to the BMO plane. NCMCA was defined by projecting BMO and ASCO points onto a plane perpendicular to the neural canal axis and measuring their overlapping area.
ASCO/BMO offset magnitude was greater (highly myopic eyes 264.3 ± 131.1 μm; healthy control subjects 89.0 ± 55.8 μm, P < .001, t test) and ASCO centroid was most frequently nasal relative to BMO centroid (94.2% of eyes) in the highly myopic eyes. BMO and ASCO areas were significantly larger ( P < .001, t test), NCMCA was significantly smaller ( P < .001), and all 3 were significantly more elliptical ( P ≤ .001) in myopic eyes. Neural canal obliqueness was greater in myopic (65.17° ± 14.03°) compared with control eyes (40.91° ± 16.22°; P < .001, t test).
Our data suggest that increased temporal displacement of BMO relative to the ASCO, increased BMO and ASCO area, decreased NCMCA, and increased neural canal obliqueness are characteristic components of ONH morphology in highly myopic eyes.
In patients with axial myopia, elongation of the eye is accompanied by structural changes to the choroid, sclera, retina, and optic nerve head (ONH) tissues that contribute to the clinical appearance of tilt, torsion, and peripapillary atrophy of the myopic optic disc. Recent optical coherence tomography (OCT) studies have described myopic alterations to the macular and peripapillary retina, macular and peripapillary choroid, lamina cribrosa, , and the Bruch membrane opening (BMO), including longitudinal temporal BMO displacement.
While refractive error is used to define myopia and axial length is commonly used to assess its progression, at present there is no OCT parameterization strategy to quantify and stage the morphologic character of myopic alteration to the ONH neural and connective tissues. By “morphologic character” of myopic alteration we mean its magnitude, tissue composition, and sectoral extent. As such, there are also no OCT strategies to account for the presence of nonglaucomatous, myopic structural alteration to the ONH tissues when attempting to detect glaucomatous ONH alterations in highly myopic eyes.
In a recent study of 362 healthy human eyes, we defined the term “neural canal” to be the connective tissue pathway of the retinal ganglion cell (RGC) axons through the ONH as they exit the eye to achieve the orbital optic nerve. We further defined it to extend from BMO through the anterior and posterior scleral canal openings and to consist of “prescleral” and “scleral canal” regions. We proposed that the size, shape and offset of the anterior scleral canal opening (ASCO) relative to BMO (ie, the ASCO/BMO offset) contributes to the direction, obliqueness, and minimum cross-sectional area (NCMCA) of the prescleral neural canal. We further predicted that incorporation of 3-dimensional neural canal connective tissue anatomy into OCT-based ONH phenotyping algorithms would eventually allow the magnitude of myopic ONH neural and connective tissue alteration in a given eye to be quantified separate from traditional measures of axial length or refractive error.
Our working hypothesis is that progressive temporal displacement of BMO relative to the ASCO (ie, progressive nasal ASCO/BMO offset), BMO and ASCO enlargement and posterior bowing of the peripapillary sclera , are core components of ONH morphology that can be used to quantify and stage the morphologic character (ie, phenotype) of myopic alteration within a given ONH in future clinical and genetic studies. As a first step toward characterizing ONH neural canal connective tissue architecture in high myopia, the purpose of the present study was to quantify the size, shape, and offset of the ASCO relative to BMO in highly myopic and non–highly myopic healthy eyes to determine ONH neural canal direction, obliqueness, and minimum cross-sectional area. Second, we wanted to determine the influence of ocular and demographic factors on these parameters.
Detecting and quantifying the clinical phenomenon of “temporal BMO displacement” in myopia is an important goal of this study. It is therefore also important to clarify that the parameter we developed for this purpose (ASCO/BMO offset, Figures 1 and 2 ) uses BMO as the reference opening and measures the offset of ASCO relative to BMO. This means that the clinical phenomenon of “temporal BMO offset” is detected and quantified as “nasal ASCO/BMO offset” within the conventions for that parameter. We choose this convention for defining ASCO/BMO offset because the RGC axons pass through BMO before reaching the ASCO within the neural canal and we felt our concepts of neural canal direction, neural canal obliqueness, and NCMCA were more clinically intuitive when the position of ASCO was characterized relative to BMO in this manner.
Study Subjects and Eyes
Our study adhered to the Declaration of Helsinki for research involving human participants and was approved by the institutional review board of each participating institution. All participants provided written informed consent. Candidate eyes included 74 highly myopic eyes (inclusion criteria below) with and without glaucomatous visual field loss (GLVFL) , from 74 subjects and 362 eyes from 362 healthy subjects from a mixed ethnicity normative database. , , ,
Subjects with highly myopic eyes with GLVFL were recruited prospectively from the glaucoma clinic at the Eye Care Centre, Queen Elizabeth II Health Sciences Centre, Halifax, Nova Scotia, Canada. Highly myopic eyes without GLVFL were recruited consecutively from attendees of a local optometry practice. , Highly myopic eye inclusion criteria included best-corrected visual acuity ≥20/40, spherical equivalent ≤ −6 diopters (D), or axial length ≥26.5 mm; astigmatism <4 D; absence of degenerative myopic changes in the macula; and absence of other retinal or optic nerve disease other than glaucoma.
Criteria for the diagnosis of “myopia with GLVFL” and “myopia without GLVFL” were described previously. In brief, the diagnosis was assigned by consensus among 3 glaucoma subspecialists who evaluated the visual fields and optic disc photographs from all participants independently and were masked from all other demographic and clinical information. To minimize bias in defining glaucoma, visual field appearance was primarily used for designating the diagnostic group of the participants. Eyes were included in the myopic without GLVFL group if their visual field was graded as normal or with abnormalities consistent with myopia, but not glaucoma, independently by all 3 clinicians, irrespective of the grading given to their optic disc. If all 3 clinicians graded the visual field as having glaucomatous abnormalities, the eye was included in the myopic with GLVFL group. In cases in which the 3 clinicians disagreed on the visual field grading, the clinicians used their optic disc evaluation to achieve a consensus assignment into either the myopia with GLVFL or myopia without GLVFL cohorts.
Initially, 131 myopic eyes of 131 subjects were recruited for the study. All clinicians independently agreed that 42 eyes were glaucomatous and 72 eyes were nonglaucomatous based on visual field assessment (ie, complete agreement in 114 [87%] subjects). Of the remaining 17 (13%) eyes, consensus classification after the optic disc evaluation was reached in 16 eyes, whereas the remaining eye in which consensus was not obtained was excluded from the study. Of these 131 myopic eyes, 74 highly myopic eyes (ie, with a myopic refractive error of > −6 D or an axial length ≥26.5 mm) were included for this study. In the present study, 5 eyes were excluded because of the poor quality of OCT images, leaving a final sample of 69 highly myopic study eyes (38 eyes without GLVFL and 31 eyes with GLVFL).
Because we reported significant age effects on several neural canal parameters in our previous study, 138 age-matched healthy control eyes for the 69 highly myopic study eyes were generated from the candidate group of 362 healthy eyes using a 1:2 case-control matching approach. Each matched set consisted of 1 highly myopic eye (case, n = 1) and 2 healthy control eyes (n = 2). The healthy control eyes were matched for age ± 5 years. This procedure was performed by 1 observer (J.W.J.) who was masked to the test results.
Inclusion criteria for the candidate group of 362 healthy eyes included: subject age 18-90 years; no history of glaucoma, intraocular pressure (IOP) ≤21 mm Hg; best-corrected visual acuity ≥20/40, refraction < ±6 D sphere and ±2 D cylinder, and glaucoma hemifield test and mean deviation within normal limits. Exclusion criteria included: unusable stereo photographs or insufficient OCT image quality (scan quality score <20); clinically abnormal optic disc appearance; any intraocular surgery (except uncomplicated cataract surgery); and any vitreous, retinal, choroidal, or neuro-ophthalmologic disease.
OCT Image Acquisition, Magnification Estimation, and Segmentation
For each eye, before OCT image acquisition the following measurements were made: visual acuity, refractive correction, curvature of the central, anterior corneal surface (by keratometry), axial length, and IOP (by Goldman applanation) were measured. OCT imaging was then performed, and the eyes were dilated, if necessary, for fundus photographic acquisition. ONH, peripapillary retinal nerve fiber layer (pRNFL), and macula were imaged with spectral-domain OCT (Spectralis, Heidelberg Engineering GmbH, Heidelberg, Germany, software version Heyex 126.96.36.199). To image each eye, the operator manually identified and marked the fovea in a live B-scan, then centered the imaging field on the ONH, where the 2 BMO points in each of 2 perpendicular ONH radial B-scans were identified. These steps established the eye-specific, fovea-BMO (FoBMO) axis, which was used as the reference for the acquisition of all subsequent OCT B-scans. The complete ONH imaging pattern consisted of 24 radial B-scans (15° apart with each B-scan containing 768 A-scans) centered on BMO. Each radial B-scan was acquired 25 times (in enhanced depth imaging mode), and averaged in real time to enhance its signal to noise ratio.
Magnification correction in the healthy control eyes was achieved by the proprietary Spectralis operating software, which uses keratometry measurements (entered into the acquisition module before imaging) and a refractive error estimate derived from the focus setting of the camera head when the operator has brought the retinal image into focus (Heidelberg Engineering Spectralis user manual). That system software is based on the Gullstrand schematic eye model and assumes a default axial length of 24.385 mm. In the highly myopic eyes, the OCT acquisition protocol of the previous study used a default value of 7.7 mm rather than the eye-specific keratometry value. Therefore, in order to account for potential effects of lateral magnification error, a post hoc adjustment of lateral pixel size was determined for each highly myopic eye using the eye-specific keratometry value and a Gullstrand schematic eye model similar to that incorporated within the Spectralis OCT system software (as confirmed previously by personal communication between Brad Fortune and Gerhard Zinser of Heidelberg Engineering, April 2009).
Our methods of OCT image manual segmentation have been described in detail previously. , , , In brief, raw OCT volumes were exported from the device and imported into a custom 3-dimensional visualization and segmentation software (Devers Eye Institute, ATL 3D Suite, Portland, Oregon, USA). ONH and peripapillary landmarks were manually segmented in each radial B-scan and the ONH was reconstructed 3-dimensionally ( Figure 1 ). Segmented landmarks included: the internal limiting membrane (ILM); the posterior surface of the pRNFL, the posterior surface of the Bruch membrane/retinal pigment epithelium complex, BMO, neural canal wall, anterior scleral surface, and the ASCO (segmented on each side of the canal by visually projecting the plane of the peripapillary anterior scleral surface through the neural canal wall and marking their intersection). , All manual segmentations were performed by 2 observers (P.W., C.H.) within the Optic Nerve Head Research Laboratory of Devers Eye Institute. Quantification of all parameters was performed within custom software (Matlab version 188.8.131.527; The MathWorks, Natick, Massachusetts, USA). All left eye data were converted to right eye configuration for analysis.
Fovea-BMO and Fovea-ASCO Distance
Fovea-BMO Distance was measured within the confocal scanning laser ophthalmoscopy (CSSO) image plane as the distance between the BMO centroid and center of the macula lutea. Fovea-ASCO distance was measured as the distance between the ASCO centroid and the center of the fovea projection in micrometers. Both parameters thus measure a 2-dimensional projection of the distance (ie, a chord) between these 2 landmarks rather than the actual distance along the curved surface of the retina.
ONH Neural Canal Connective Tissue Parameters
Detailed descriptions of the following parameters, along with an illustrative video, are published elsewhere. All data are reported in right eye orientation.
BMO and ASCO size and shape
A plane was fitted to the 48 segmented BMO and ASCO points, respectively ( Figure 2 ), satisfying a least mean square error restraint in each case. Then, the BMO points were projected to the best-fit BMO plane. Using the projected points, a best-fit ellipse was determined and the BMO centroid, area, and ovality index (ellipse long axis length/ellipse short axis length) were calculated. An ASCO centroid, area, and shape index were similarly calculated within the ASCO plane.
ASCO/BMO offset magnitude and direction
ASCO/BMO offset magnitude and direction were defined by projecting the ASCO/BMO centroid vector (connecting the BMO and ASCO centroids) to the BMO plane ( Figure 2 ). ASCO/BMO offset magnitude was defined within the BMO plane as the length of the ASCO/BMO vector component within the BMO plane. ASCO/BMO offset direction was defined within the BMO reference plane by the angle between the projected ASCO/BMO centroid vector and the FoBMO axis (0°) measured clockwise relative to the FoBMO axis (superior 90°, nasal 180°, and inferior 270°).
Neural canal axis, direction and obliqueness
The neural canal axis was defined by the ASCO/BMO centroid vector as described above ( Figure 2 ). Neural canal direction and ASCO/BMO offset direction are therefore identical, were measured identically, and are most commonly referred to as ASCO/BMO offset direction within this manuscript. Neural canal obliqueness was defined by the angle between the neural canal axis vector and a vector perpendicular to the BMO plane, originating at the BMO centroid ( Figure 2 ).
NCMCA ( Figure 3 ) estimates the smallest opening through which the RGC axons pass as they leave the eye. It is calculated within a plane that is perpendicular to the neural canal axis (the neural canal perpendicular plane), by projecting the BMO and ASCO points onto it and quantifying the area that is common to both projections ( Figure 3 ). NCMCA ovality index was calculated as outlined for BMO and ASCO (NCMCA ovality index = ellipse long axis length/ellipse short axis length).
Manual Segmentation Reproducibility
The reproducibility of our study parameters within the 362 healthy, non–highly myopic human eyes from which the healthy control eyes were chosen, has been previously reported to be excellent. Because of the known difficulty in segmenting BMO in highly myopic eyes, we assessed interdelineator reproducibility within the highly myopic eyes of this report by having 2 delineators independently segment 6 of the highly myopic eyes that were chosen to span the range of axial length present within the 68 highly myopic eyes of this study.
Descriptive statistics included the mean and standard deviation for continuous variables and the proportions for categorical variables. Interobserver reproducibility was assessed with the intraclass correlation coefficient of each variable. Baseline characteristics and all OCT ONH parameters were compared between the 2 groups with the t test for continuous variables and χ 2 test for categorical variables.
For multiple comparisons, the Holm-Bonferroni method was used to adjust for type I error. Factors associated with all neural canal connective tissue parameters were initially evaluated with univariable linear regression analysis. Factors associated with each dependent variable with a P ≤ .10 were evaluated in multivariable regression models. Before the multivariable analysis, collinearity between the independent variables was evaluated with correlation analysis. All statistical analyses were performed with IBM SPSS Statistics (v 24.0, IBM Corp, Armonk, New York USA) and GraphPad Prism (v 8.1.2, GraphPad Software, Inc, San Diego, California, USA). P < .05 was considered statistically significant.
The demographic and ocular characteristics of the 69 highly myopic study subjects and 138 age-matched control subjects are summarized in Table 1 . By design, there was no significant difference in the mean age between highly myopic (57.3 ± 9.2) and healthy control subjects (57.2 ± 9.4). As expected, the refraction, axial length, fovea-BMO centroid distance, global pRNFL thickness, and global minimum rim width (MRW) were significantly different between highly myopic and healthy control eyes ( P < .001, t test).
|Highly Myopic Eyes, n = 69 Mean (SD)||Healthy Control Eyes, n = 138 Mean (SD)||P Value|
|Age, y (SD)||57.3 (9.2)||57.2 (9.4)||.927|
|Female gender, n (%)||33 (47.8)||81 (58.7)||.138|
|Left eye, n (%)||34 (49.3)||61 (44.2)||.490|
|IOP on imaging day, mm Hg (SD)||15.4 (3.5)||14.6 (2.8)||.070|
|CCT, μm (SD)||550.7 (37.2)||558.1 (29.8)||.126|
|Refraction, diopters (SD)||−7.61 (2.27)||-0.14 (1.82)||<.001 a|
|Axial length, mm (SD)||26.96 (1.07)||23.62 (0.96)||<.001 a|
|Cornea curvature, mm (SD)||7.76 (0.32)||7.71 (0.25)||.2752|
|Fovea-BMO distance, μm (SD)||4458.8 (473.3)||4391.0 (282.8)||<.001 a|
|Fovea-ASCO distance μm (SD)||4672.4 (524.8)||4444.0 (298.6)||<.001 a|
|Global pRNFLT, μm (SD)||74.2 (13.9)||96.4 (10.5)||<.001 a|
|Global MRW, μm (SD)||231.2 (81.1)||328.0 (58.9)||<.001 a|
a Statistically significant differences ( P < .05, t test or χ 2 test) that remain significant using a correction for multiple comparisons (Holm-Bonferroni method).
While comparisons between highly myopic eyes with (n = 38) and without (n = 31) GLVFL were not a primary goal of this study, there were significant differences in age, IOP on examination day, axial length, global pRNFL thickness, global MRW, and visual field mean deviation between the 2 subgroups ( P < .05, t test). However, only axial length, global pRNFL thickness, global MRW, and mean deviation remained significant after applying a Holm-Bonferroni correction for multiple comparisons ( Supplemental Table 1 ).
Study parameter intraclass correlation coefficient values for the highly myopic eyes of this study were excellent, ranging from 0.836-0.998.
Highly myopic versus healthy control eye comparisons
Fovea-BMO and fovea-ASCO distance
Both the fovea-BMO distance (4458.8 μm [473.3] vs 4391.0 μm [282.8]) and the fovea-ASCO distance (4672.4 μm [524.8] vs 4444.0 μm [298.6]; P < .001, t test corrected for multiple comparisons, Table 1 ) were increased in the highly myopic compared to the healthy control eyes.
BMO, ASCO, and NCMCA size and shape
The BMO and ASCO areas of highly myopic eyes (2.323 ± 0.798 mm 2 and 2.263 ± 0.750 mm 2 ) were significantly larger than those of the healthy control eyes (1.795 ± 0.354 mm 2 and 2.166 ± 0.402 mm 2 , respectively; P < .001, t test). NCMCA was significantly smaller in the highly myopic eyes (0.857 ± 0.559 mm 2 ) compared with control eyes (1.280 ± 0.378 mm 2 ; P < .001, t test). BMO, ASCO, and NCMCA ovality indices were significantly higher in highly myopic eyes compared with control eyes ( P s ≤ .001, t test). These differences remained significant after correction for multiple comparisons ( Table 2 , Figures 4 and 5 ).
|Highly Myopic Eyes, n = 69 (Mean ± SD)||Healthy Control Eyes, n = 138 (Mean ± SD)||P Value|
|BMO area (mm 2 )||2.323 ± 0.798||1.795 ± 0.354||<.001 a|
|ASCO area (mm 2 )||2.263 ± 0.750||2.166 ± 0.402||<.001 a|
|NCMCA (mm 2 )||0.857 ± 0.559||1.280 ± 0.378||<.001 a|
|BMO ovality index||1.132 ± 0.086||1.125 ± 0.058||<.001 a|
|ASCO ovality index||1.145 ± 0.087||1.127 ± 0.063||<.001 a|
|NCMCA ovality index||2.780 ± 0.979||1.557 ± 0.567||<.001 a|
|ASCO/BMO offset magnitude (μm)||264.3 ± 131.1||89.0 ± 55.8||<.001 a|
|ASCO/BMO offset direction (degrees)||156.4 ± 37.5||140.6 ± 61.5||<.001 a|
|Neural canal obliqueness (degrees)||65.17 ± 14.03||40.91 ± 16.22||<.001 a|