January 2013
Volume 54, Issue 1
Free
Retina  |   January 2013
A Novel Segmentation Algorithm for Volumetric Analysis of Macular Hole Boundaries Identified with Optical Coherence Tomography
Author Notes
  • From the Ophthalmic Imaging Center, Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio. 
  • Corresponding author: Justis P. Ehlers, Cole Eye Institute, Cleveland Clinic, 9500 Euclid Avenue, i32, Cleveland, OH 44195; ehlersj@ccf.org
Investigative Ophthalmology & Visual Science January 2013, Vol.54, 163-169. doi:10.1167/iovs.12-10246
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      David Xu, Alex Yuan, Peter K. Kaiser, Sunil K. Srivastava, Rishi P. Singh, Jonathan E. Sears, Daniel F. Martin, Justis P. Ehlers; A Novel Segmentation Algorithm for Volumetric Analysis of Macular Hole Boundaries Identified with Optical Coherence Tomography. Invest. Ophthalmol. Vis. Sci. 2013;54(1):163-169. doi: 10.1167/iovs.12-10246.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose.: To demonstrate a novel algorithm for macular hole (MH) segmentation and volumetric analysis.

Methods.: A computer algorithm was developed for automated MH segmentation in spectral-domain optical coherence tomography (SD-OCT). Algorithm validation was performed by trained graders with performance characterized by absolute accuracy and intraclass correlation coefficient. A retrospective case series of 56 eyes of 55 patients with idiopathic MHs analyzed using the custom algorithm to measure MH volume, base area/diameter, top area/diameter, minimum diameter, and height-to-base diameter ratio. Five eyes were excluded due to poor signal quality (1), motion artifact (1), and failure of surgical closure (3) for a final cohort of 51 eyes. Preoperative MH measurements were correlated with clinical MH stage, baseline, and 6-month postoperative best-corrected Snellen visual acuity (BCVA).

Results.: The algorithm achieved 96% absolute accuracy and an intraclass correlation of 0.994 compared to trained graders. In univariate analysis, MH volume, base area, base diameter, top area, top diameter, minimum diameter, and MH height were significantly correlated to baseline BCVA (P value from 0.0003–0.011). Volume, base area, base diameter, and height–to-base diameter ratio were significantly correlated to 6-month postoperative BCVA (P value from <0.0001–0.029). In multivariate analysis, only base area (P < 0.0001) and volume (P = 0.0028) were significant predictors of 6-month postoperative BCVA.

Conclusions.: The computerized segmentation algorithm enables rapid volumetric analysis of MH geometry and correlates with baseline and postoperative visual function. Further research is needed to better understand the algorithm's role in prognostication and clinical management.

Introduction
The pathogenesis of idiopathic macular holes (MHs) is believed to involve static and dynamic antero-posterior forces applied by vitreofoveal traction as well as potential tangential traction from involutional changes in the inner retina. 14 As such, the progression of MHs can be described by vitreoretinal forces that lead to a series of evolving anatomical defects. Advances in vitreoretinal surgery have contributed to successful postoperative restoration of foveal microstructure and restoration of visual function in most patients. 57 Spectral-domain optical coherence tomography (SD-OCT) is a noncontact, noninvasive, laser interferometry technique that generates high-resolution, in vivo, cross-sectional images and is used in diagnosing numerous macular diseases. The application of SD-OCT imaging to MHs augments clinical staging by enabling visualization of the foveal and vitreous microstructure and tractional relationships and by calculating size measurements of hole architecture. SD-OCT allows for manual quantification of the width and height of the MH and identifies perifoveal cystoid edema, vitreomacular traction, and perifoveal inner segment and outer segment (IS/OS) integrity, and it allows clear discrimination between lamellar and pseudo-MHs. 8  
Although SD-OCT is being utilized for increasing numbers of ophthalmic diseases, automated analysis of volumetric SD-OCT data has not moved significantly beyond analysis of retinal thicknesses. While modern SD-OCT devices capture a large, three-dimensional (3D) field of view, visualization of the entire dataset on a frame-by-frame basis is time-consuming. In the case of MHs, the horizontal cross-section through the foveal center or across the largest MH width is typically used for clinical evaluation. This may lead to misjudgment of MH size because the true largest width may lie in an oblique axis. As clinical outcomes in MH surgery have been reported to be partly dependent on anatomic configuration, accurate 3D analysis of preoperative MH may enhance prognostication and inform clinical decision-making. 
MHs are well suited for automated segmentation. The hyporeflective cavity of the MH is bounded by the high-contrast boundaries of the retina, RPE, and internal limiting membrane (ILM), making unequivocal boundary delineation by a computer highly accurate. Furthermore, MHs are relatively constrained in their geometric shape and are always found in the foveal location, simplifying the task of fully automated measurement. Previous reports have used preoperative MH OCT findings to predict surgical outcomes. MH minimum diameter less than 400 μm was predictive of higher anatomical closure rate. 9 The ratio of the MH height-to-base diameter ratio (deemed the MH index, MHI) correlated with postsurgical visual function. 10 The integrity of the perifoveal IS/OS has been associated with better postoperative visual outcome. 11  
In this report, a novel computer segmentation algorithm for MHs is developed and tested that provides rapid 3D and two-dimensional (2D) geometric measurements of the MH. Volumetric and area measurements are correlated with clinical staging and with baseline and postoperative visual acuity. 
Methods
Algorithm Development
A novel computerized segmentation algorithm was developed. To reduce bias created by shadowing artifacts and low signal, SD-OCT scans were used only if the signal strength was greater than or equal to 6. To decrease processing time, volumetric data was extracted in 300 by 200 pixel frames for volumetric analysis. In all scans, the foveal center was manually selected by a human grader. For algorithm validation, the foveal SD-OCT cross-section was exported with a resolution of 924 by 616 pixels. 
The segmentation goal was to optimally delineate the boundary between the MH cavity and the adjacent retina and RPE. The algorithm processes volumetric data on a frame-by-frame basis and is comprised of four general steps consisting of preprocessing, segmentation using graph search, reweighting of adjacent edge weights, and calculation of MH parameters (Fig. 1). 
Figure 1. 
 
Schematic diagram of segmentation algorithm.
Figure 1. 
 
Schematic diagram of segmentation algorithm.
First, every frame is filtered with a 5 by 5 pixel median filter to reduce image noise (Fig. 2A). Next, the constraint boundaries for the graph search are found (Fig. 2B). Anteriorly, the search region is bounded by the ILM. Posteriorly, the search region is bounded by the RPE. The bordering vitreoretinal contour will be important in determining the top edge of segmentation in each frame. The location of the MH is assumed a priori to be located within 2 mm of the foveal center. This is generally true of idiopathic full-thickness MH, and the authors were not able to identify any exceptions in the cohort. The best approximation of the MH location is then determined by finding the region of lowest retinal thickness within the 4-mm diameter search region. 
Figure 2. 
 
(A) Foveal spectral domain optical coherence tomography cross-section of a 45-year-old patient with stage 4 MH. (B) Delineation of the RPE (blue) and internal limiting membrane (red) constraint boundaries. (C) Foreground seed pixels (red) are established by eroding the segmentation result from the adjacent frame. Background seed pixels (blue) are established by selecting pixels a fixed distance from the segmentation result from the adjacent frame. (D) Segmentation result (red) is the solution of the graph cut.
Figure 2. 
 
(A) Foveal spectral domain optical coherence tomography cross-section of a 45-year-old patient with stage 4 MH. (B) Delineation of the RPE (blue) and internal limiting membrane (red) constraint boundaries. (C) Foreground seed pixels (red) are established by eroding the segmentation result from the adjacent frame. Background seed pixels (blue) are established by selecting pixels a fixed distance from the segmentation result from the adjacent frame. (D) Segmentation result (red) is the solution of the graph cut.
Segmentation is performed independently in each frame using the method of graph cuts. The mathematical treatment of graph cuts for image segmentation is complex and can be found in previous reports.12,13 Briefly, an undirected graph G = < V, E > is established by a set of vertices V and edges E. Consider the set of data elements IP representing image pixels and an 8-connected neighborhood system N. For every element p ∈ IP, there exists an edge with nonnegative weight we that connects qN. In addition, there are two vertices named the source S and sink T, each with edges that connect to every IP. A graph cut C is a cut through the edges such that the induced graph G = < V, E\C > bipartitions the graph and separates the source and sink vertices. To use graph cuts to optimally segment a boundary, image intensity is encoded by edge weight to the sink and source vertices, while gradient is encoded by edge weight to neighboring nodes. A graph cut in the network computes the minimization of the energy function  where R(A) specifies the intensity cost, B(A) specifies the boundary cost, and λ is the intensity to boundary cost ratio. In our implementation, we define    where I is pixel intensity, CF and CB are finite foreground and background costs, σ is a noise attenuation factor, and γ is a nonlinear intensity scaling factor. Foreground seed pixels are established by thresholding image intensity in a truncated foveal region (Fig. 2C). The resulting seed markers are morphologically eroded to 25% of original area. For subsequent frames, the segmentation result from the adjacent frame is eroded to 25% of original area. Background seed pixels are established by selecting pixels a fixed distance from the centroid of foreground pixels.  
We chose to use graph cuts implemented in 2D rather 3D because of significant anisotropy of the SD-OCT dataset. While the Cirrus 512 by 128 raster scan has axial and lateral sampling of less than 20 μm, the sampling between frames is 50 μm. Significant change in the MH morphology between adjacent frames, especially in the superior and inferior boundaries of the 3D MH structure where hole morphology is rapidly changing, may not satisfy the neighborhood connectivity constraints of the graph formulation. To improve robustness and increase the between-frame connectivity of the 2D segmentation, edge weights in adjacent frames are reweighted based on the segmentation result in the current frame. Reweighted edges have an additional edge weight term proportional to the distance of the edge to the prior segmentation such that the new segmentation favors a cut of similar shape to the previous result. 
The first frame to be segmented lies at the center of the approximate MH location. Adjacent frames are then segmented. Edge weights in adjacent frames are reweighted by  Where dist is the Euclidean distance function from p to the adjacent graph cut and α is a reweight scaling factor. The graph cut is the set of edges that optimally segregates the hyporeflective MH cavity from the surrounding retina (Fig. 2D).  
The segmentation from individual frames is assembled to generate a 3D surface representation of the MH (Fig. 3). Volumetric calculation is performed by counting voxels inscribed in the surface. Calculation of base area and top area is performed by projecting the base and top boundary points of the segmentation to a best-fit plane and calculating the area of the resulting polygon. The maximum base and top distance is defined as the greatest distance between boundary points in the base and top polygon. The minimum diameter (i.e., smallest hole diameter) is defined as the minimum horizontal width of the MH in the frame with largest base width, which measures the minimum separation distance of the retina. MH height is defined as the average of the normal distances from the top boundary to the RPE. The MHI is calculated by dividing the height by the base diameter. The algorithm was implemented in MATLAB (Mathworks Inc., Natick, MA) and the graph search was implemented in C++ (GNU Compiler Collection; Free Software Foundation Inc., Boston, MA). The human operator reviews the graph cut in each frame to verify that segmentation is plausible. If refinement is needed, manual boundaries in the segmentation (i.e., the top boundary of the MH with adherent posterior vitreous face) can be inputted and the segmentation rerun. 
Figure 3. 
 
3D representation of MH from Figure 2 with highlighted top area and top diameter measured across the largest axis (A), minimum diameter (B), and base area and base diameter (C).
Figure 3. 
 
3D representation of MH from Figure 2 with highlighted top area and top diameter measured across the largest axis (A), minimum diameter (B), and base area and base diameter (C).
Algorithm Validation
For validation purposes, masked grading was performed by two vitreoretinal specialists (JPE and AY) on a subset of eyes for comparison to the automated segmentation algorithm. The foveal SD-OCT frame was used for comparison, and the cross-sectional area of the MH (i.e., the area of pixels judged to be part of the MH cavity) was manually measured using ImageJ (freeware; National Institutes of Health, Bethesda, MD). The mean cross-sectional area judged by the expert graders was considered the gold standard. Absolute accuracy and intraclass correlation were used to assess algorithmic performance against this standard. 
Clinical Testing
After Institutional Review Board of the Cleveland Clinic approval, a retrospective case series was performed of patients seen at the Cole Eye Institute from February 2009 to June 2011. All tenets of the Declaration of Helsinki were adhered to for all aspects of this research. Patients were included if they had the diagnosis of idiopathic MH after a comprehensive ophthalmic examination and SD-OCT imaging (Cirrus HD-OCT; Carl Zeiss Meditec, Dublin, CA) consisting of one macular raster scan (512 A-scans per B-scan, 128 B-scans, 6- by 6-mm field of view) and one horizontal five-line pattern. Patients with other diseases that could affect visual function such as AMD, glaucoma, or macular edema from diabetic retinopathy or retinal vein occlusion were excluded. To maintain a valid population for comparison, eyes that failed to achieve postoperative anatomical closure of the MH were excluded. Staging of the MH was judged by biomicroscopic appearance based on the Gass classification with support from the SD-OCT findings. 3,4 All patients underwent standard three port 23 or 25-gauge pars plana vitrectomy (PPV) with or without peeling of the internal limiting membrane (ILM) assisted by indocyanine green staining, gas tamponade with perfluoropropane or sulfur hexafluoride, and prone face positioning. The clinical records were reviewed for baseline best-corrected Snellen visual acuity (BCVA) and Gass classification stage, as well as BCVA and hole closure verified by SD-OCT at postoperative month 1, 3, and 6 visits. 
Statistical Analysis
The SD-OCT macular scan at baseline was analyzed automatically for MH volume, base area, maximum base diameter, top area, maximum top diameter, minimum diameter, height, and MHI. Each MH measurement was compared between Gass stage using one-way ANOVA. Snellen BCVA was converted to logMAR for statistical analysis. MH measurements were compared to baseline BCVA and 6-month postoperative BCVA using Spearman rank correlation. The predictive potential of the MH measurements adjusted for age and baseline BCVA to 6-month postoperative BCVA was assessed by multiple linear regression. A P value of less than 0.05 was considered statistically significant. 
Results
Algorithm Validation and Performance
An initial test set of 25 eyes was analyzed by automated segmentation and compared to masked, expert graders. The algorithm achieved 96% absolute accuracy compared with the gold standard. The mean error between the two graders was 6.1%. The intraclass correlation coefficient comparing computer segmentation to the mean of human graders was 0.994. The intraclass correlation between both graders was 0.992. The average processing time for one SD-OCT dataset was 38 seconds when running on a computer with an Intel Core 2 Quad 2.4 GHz processor and 4 gigabytes of RAM, and review of the segmentation results by the human grader was completed in an average of 50 seconds. In comparison, manual segmentation takes an average of 5 minutes for one SD-OCT dataset. 
Following validation, 54 (of 56) SD-OCT datasets of idiopathic MHs were analyzed using the algorithm. The frame-by-frame segmentation results were reconstructed into a 3D surface, and the MH measurements of volume, base area, base diameter, top area, top diameter, minimum diameter, and height were calculated. The algorithm was successfully completed in all 51 datasets. Five cases (9.8%) underwent human-aided refinement of the segmentation boundary. It was observed that errors in segmentation occurred predominantly at the edge of the MH boundary where there was greater irregularity of the shape of the segmentation result. Segmentation error did not appear to be related to OCT signal strength. 
Clinical Algorithm Testing
Fifty-six eyes (30 right, 26 left) of 55 patients (23 men, 32 women) with idiopathic MH were initially identified. Five eyes were excluded for the following reasons: motion artifact (1), poor signal quality (1), failure of anatomic closure (3). Algorithm analysis of the three eyes that did not achieve anatomic closure was performed for measurement comparison to other eyes, but the three eyes were excluded from the general analysis given the confounding postoperative visual acuity due to the persistent open hole. This resulted in 51 eyes for the cohort. The cohort was comprised of 3 stage 1 holes, 10 stage 2 holes, 22 stage 3 holes, and 16 stage 4 holes. The mean patient age was 68 ± 11 years. At baseline, patients had a mean BCVA of 0.72 ± 0.37 logMAR (∼20/104 Snellen equivalent). All patients underwent surgical repair with successful closure of their MH with one surgery. Seven patients were pseudophakic at baseline; 31 of the 44 phakic patients at baseline underwent combined PPV and cataract extraction. At the 6-month postoperative visit, the mean BCVA was 0.39 ± 0.37 logMAR (∼20/49 Snellen equivalent). 
The cohort was subdivided by clinical MH stage and analyzed for MH volume, base area, base diameter, top area, top diameter, minimum diameter, height, and MHI (Table 1). Mean hole volume, top area, top diameter, minimum diameter, and height increased with increasing stage (Fig. 4). Mean base area and base diameter were greater in stage 1 MH than stage 2. Mean MHI was greatest in the stage 2 group, followed by the stage 4 group, the stage 3 group, and smallest in the stage 1 group. Whole-group ANOVA demonstrated a significant difference between the Gass stages for volume, base area, base diameter, top area, minimum diameter, and height; there was no significant difference in top diameter and MHI. 
Figure 4. 
 
SD-OCT B-scan and 3D representation of a stage 1 MH in a 73-year-old patient with 20/70 acuity (A, B). A 74-year-old patient with a stage 2 MH and 20/50 acuity (C and D). A 73-year-old patient with a stage 3 MH and 20/70 acuity (E, F). A 52-year-old patient with a stage 4 MH and 20/100 acuity (G, H).
Figure 4. 
 
SD-OCT B-scan and 3D representation of a stage 1 MH in a 73-year-old patient with 20/70 acuity (A, B). A 74-year-old patient with a stage 2 MH and 20/50 acuity (C and D). A 73-year-old patient with a stage 3 MH and 20/70 acuity (E, F). A 52-year-old patient with a stage 4 MH and 20/100 acuity (G, H).
Table 1. 
 
Mean ± SD Measurements of MH Size
Table 1. 
 
Mean ± SD Measurements of MH Size
Stage 1 Stage 2 Stage 3 Stage 4 P Value
Mean volume, mm3 0.026 ± 0.014 0.037 ± 0.017 0.089 ± 0.071 0.179 ± 0.130 0.0005
Mean base area, mm2 0.147 ± 0.082 0.101 ± 0.088 0.364 ± 0.288 0.837 ± 0.668 0.0004
Mean base diameter, μm 511 ± 145 416 ± 185 743 ± 285 993 ± 476 0.001
Mean top area, mm2 0.189 ± 0.168 0.285 ± 0.194 0.345 ± 0.185 0.511 ± 0.066 0.02
Mean top diameter, μm 738 ± 229 752 ± 254 789 ± 189 934 ± 246 0.13
Mean minimum diameter, μm N/A 221 ± 136 323 ± 134 411 ± 172 0.0026
Mean height, μm 189 ± 62 199 ± 79 330 ± 110 401 ± 105 0.0001
Macular hole index 0.370 ± 0.049 0.516 ± 0.148 0.468 ± 0.127 0.515 ± 0.379 0.75
MH measurements were compared to baseline and 6-month postoperative BCVA by Spearman rank correlation (Table 2). There was a strong correlation of MH volume, base area, base diameter, top area, top diameter, minimum diameter, and height to baseline BCVA (P value from 0.0003–0.011) with a correlation of larger measurement to worse visual acuity. The MHI did not correlate with baseline BCVA (P = 0.163). When comparing the baseline MH measurements to the 6-month postoperative BCVA, hole volume, base area, base diameter, and MHI were highly correlated (P value from <0.0001–0.029). There was no significant correlation of 6-month postoperative BCVA with top area (P = 0.44), top diameter (P = 0.58), minimum diameter (P = 0.061), and height (P = 0.80). The multiple regression model correlated age, baseline BCVA, and MH measurements (except base diameter and top diameter because they are redundant to base area and top area) to 6-month postoperative visual outcome. Only base area (P < 0.0001) and volume (P = 0.0028) were statistically significant predictors of 6-month postoperative BCVA (Table 3). 
Table 2. 
 
Nonparametric Correlation of BCVA at Baseline and at the 6-Month Postoperative Visit to Measurements of MH Size
Table 2. 
 
Nonparametric Correlation of BCVA at Baseline and at the 6-Month Postoperative Visit to Measurements of MH Size
Spearman Correlation to Baseline BCVA Spearman Correlation to 6-Month Postoperative BCVA
Volume 0.488 (0.0003) 0.306 (0.029)
Base area 0.443 (0.0011) 0.470 (0.0005)
Base diameter 0.353 (0.011) 0.371 (0.0074)
Top area 0.365 (0.0084) 0.110 (0.442)
Top diameter 0.384 (0.0055) 0.079 (0.580)
Minimum diameter 0.393 (0.0043) 0.264 (0.061)
Height 0.369 (0.0077) 0.248 (0.80)
Macular hole index −0.198 (0.163) −0.380 (0.0059)
Table 3. 
 
Multiple Regression Analysis of Age, Baseline BCVA, and MH Measurements to 6-Month Postoperative Visual Acuity
Table 3. 
 
Multiple Regression Analysis of Age, Baseline BCVA, and MH Measurements to 6-Month Postoperative Visual Acuity
P Value
Age, y 0.52
Baseline BCVA, logMAR 0.26
Volume, mm3 0.0028
Base area, mm2 <0.0001
Top area, mm2 0.61
Height, μm 0.97
Minimum diameter 0.82
Macular hole index 0.97
Three eyes (excluded from final visual acuity analysis) failed to achieve MH closure following surgical intervention. Algorithm analysis of these eyes was performed. The MH volumes of two of the three eyes were nearly 2 SD larger than the mean MH volume (0.396 and 0.458 mm3). The third eye was a chronic MH; although smaller, it had no macular edema and appeared to be nearly “flat open” at time of surgical intervention. With only three eyes in this group, statistical analysis was not performed on features associated with closure success. 
Discussion
In the present study, we explore the use of a novel, computerized segmentation algorithm to analyze idiopathic MHs imaged by SD-OCT raster scans. We utilize the segmentation to calculate geometric measurements of the MH cavity and then correlate these measurements to MH staging, baseline visual acuity, and postoperative visual outcomes. Automated segmentation provides many advantages in the analysis of MHs. While previous reports analyzing the geometry of MHs have used measurements from one representative cross-sectional image, our algorithmic approach allows the simultaneous and rapid analysis of the 3D geometric qualities of the hole. 
We first sought to examine the trend of MH measurements with clinical staging. Traditionally, the Gass classification of MHs depended on the presence of a full-thickness retinal defect, width of the hole, and posterior vitreous detachment. 3,4,8 The staging criteria is generally associated with greater size of the MH cavity. MH volume, top area, top diameter, minimum diameter, and height increase with increasing stage. This trend is plausible given that volume, top area, and height are geometrically interrelated and correlate with a larger MH. Minimum diameter, measured as the minimum width in the OCT cross-section with greatest base diameter, can be considered the closest correlate to MH size in the Gass classification and was also seen to increase with increasing stage. Base area and base diameter also increase with increasing stage; however, base area and base diameter are larger in the stage 1 group than the stage 2 group. This may be due to the small sample size in the stage 1 group (n = 3). Interestingly, the MHI was largest in the stage 2 group and stage 4 group. 
To address whether MH measurements were related to preoperative visual function, we correlated MH measurements with baseline BCVA. Our results show that hole volume, base area, base diameter, top area, top diameter, minimum diameter, and height were all statistically associated with baseline visual acuity. This finding is in agreement with previous reports that show a significant correlation of base diameter and MH height to preoperative visual acuity. 1114 In this series, three eyes failed to close following surgical repair. Not surprisingly, a trend towards larger MH volume was seen. With only three eyes, the sample size was too small for statistical analysis. We are currently examining the algorithm's predictive value not only for surgical success (e.g., final closure) but also for rate of hole closure using perioperative OCT. If the high-resolution, 3D data set is able to provide information on hole closure rate, this tool may have significant utility in surgical decision-making (e.g., type of tamponade, duration of positioning). 
To evaluate the predictive potential of SD-OCT volumetric MH measurements to visual outcomes, we compared MH measurements to visual outcomes using univariate analysis and multiple regression analysis of nonredundant measurements. Hole volume, base area, base diameter, and MHI were strongly statistically associated with 6-month postoperative visual acuity in univariate comparison. In addition, minimum diameter showed a trend toward association with postoperative visual acuity. This finding echoes previous reports that show a significant association of base diameter and minimum diameter to postoperative visual outcome. 15 Multiple linear regression of the MH measurements adjusted for age and baseline BCVA demonstrated that only MH volume and base area were significant predictors of postoperative visual acuity. Previous studies have shown an association of the MHI to postoperative visual outcome. 10,11 While we found a significant association in the univariate correlation, MHI was not a significant predictor of vision in the multiple regression model. We hypothesize that the lack of association is because the regression model included other parameters more strongly associated with postoperative visual acuity. 
The strength of the volumetric measurements relative to other linear measurements in prognosticating visual outcomes requires testing in a larger scale study. The preliminary results of this study are promising given that in the multivariate models, the volumetric measurements were among the strongest associated variables. MH volume takes into account the aggregate contribution of top diameter of the MH, base diameter, minimum diameter, and height. Furthermore, volumetric calculation is resistant to error from asymmetrically shaped MHs, which have a true size that is not well represented by a horizontal OCT cross-section. The MH base area is also a better representation of the area of outer retinal disruption than a single linear cross-section. Our data demonstrate that base area and volume are significant predictors of visual outcome even while adjusting for the MHI, baseline vision, and the age of the patient. Finally, our data suggest that volumetric MH geometry may be a stronger predictor of visual outcome than preoperative visual acuity. This suggests that preoperative assessment of MH geometry may provide improved information to patients regarding expectations for visual outcomes. Other studies report the use of electroretinograms and microperimetry by scanning laser ophthalmoscope to predict postoperative visual outcome. 1618 In addition, SD-OCT studies examining the predictive potential of photoreceptor IS/OS measurements to visual outcome have been done. 11,19,20 Our findings demonstrate that evaluating and including the noncentral OCT cross-sections of a MH and 3D MH morphology holds prognostic value, but more research is needed to demonstrate the comparative strength of its prognostication versus standard linear measurements. 
The clinical findings in this study are the result of an automated segmentation technique that reduces subjectivity, improves consistency, and has accuracy suitable for clinical evaluation. However, there are several limitations of the study. First, pre- and postoperative phakic status was not controlled in this retrospective study. However, multiple regression analysis of the subgroups based on phakic status demonstrated identical significant predictors of 6-month postoperative acuity of base area and volume. An additional limitation is that the segmentation validation protocol in this study required human graders to evaluate a single SD-OCT frame rather than a volumetric data set, which raises some concerns regarding extrapolation from the area to volume calculation. However, because there is no accepted method of calculating the base area, top area, or volume of the MH structure from a series of multiple frames for human graders, this was determined to be the best surrogate for validation of the volumetric measurement. Additionally, subtle motion artifact in the SD-OCT image may translate into increased area, diameter, or volume calculation, which may introduce greater variation into the correlation with pre- and postoperative visual acuity. We reviewed OCT B-scans and 3D renderings for evidence of motion artifact and excluded one eye for serious motion artifact within the extent of the MH. Finally, the 3D MH morphology was characterized by volume and area measurements. 
Further study is warranted in exploring other clinically relevant morphologic descriptors, including IS/OS loss, and their impact on this multivariate model. Additional research is also needed to better define the prognostic value and relative strength of this model compared to linear measurements in a larger sample size. Finally, assessment of the utility of the volumetric algorithm in predicting rate of hole closure is needed to determine whether this algorithm may be used for enhancing surgical decision-making (e.g., need for face-down positioning). Currently, we are evaluating the use of this algorithm in MH surgery to identify subtle changes in MH geometry following surgical intervention and the implications for rate of hole closure as determined with intraoperative and perioperative OCT. 
In conclusion, we used a computerized segmentation algorithm to analyze idiopathic MHs on SD-OCT and evaluated various measurements of MH architecture. Measurements of MH size were associated with baseline and postoperative visual function. Preoperative MH base area and volume may be clinically useful predictors of postoperative visual outcome for MH surgery. This tool may have significant applications to other ophthalmic pathologic conditions as well as utility in rapidly analyzing morphological changes induced by surgical manipulation with intraoperative OCT. 
References
Bainbridge J Herbert E Gregor Z. Macular holes: vitreoretinal relationships and surgical approaches. Eye (Lond) . 2008; 22: 1301–1309. [CrossRef] [PubMed]
Johnson RN Gass JD. Idiopathic macular holes. Observations, stages of formation, and implications for surgical intervention. Ophthalmology . 1988; 95: 917–924. [CrossRef] [PubMed]
Gass JD. Idiopathic senile macular hole. Its early stages and pathogenesis. Arch Ophthalmol . 1988; 106: 629–639. [CrossRef] [PubMed]
Gass JD. Reappraisal of biomicroscopic classification of stages of development of a macular hole. Am J Ophthalmol . 1995; 119: 752–759. [CrossRef] [PubMed]
Freeman WR Azen SP Kim JW Vitrectomy for the treatment of full-thickness stage 3 or 4 macular holes. Results of a multicentered randomized clinical trial. The Vitrectomy for Treatment of Macular Hole Study Group. Arch Ophthalmol . 1997; 115: 11–21. [CrossRef] [PubMed]
Park DW Sipperley JO Sneed SR Dugel PU Jacobsen J. Macular hole surgery with internal-limiting membrane peeling and intravitreous air. Ophthalmology . 1999; 106: 1392–1397. [CrossRef] [PubMed]
Haritoglou C Gass CA Schaumberger M Macular changes after peeling of the internal limiting membrane in macular hole surgery. Am J Ophthalmol . 2001; 132: 363–368. [CrossRef] [PubMed]
Altaweel M Ip M. Macular hole: improved understanding of pathogenesis, staging, and management based on optical coherence tomography. Semin Ophthalmol . 2003; 18: 58–66. [CrossRef] [PubMed]
Ip MS Baker BJ Duker JS Anatomical outcomes of surgery for idiopathic macular hole as determined by optical coherence tomography. Arch Ophthalmol . 2002; 120: 29–35. [CrossRef] [PubMed]
Kusuhara S Teraoka Escaño MF Fujii S Prediction of postoperative visual outcome based on hole configuration by optical coherence tomography in eyes with idiopathic macular holes. Am J Ophthalmol . 2004; 138: 709–716. [CrossRef] [PubMed]
Oh J Smiddy WE Flynn HW Gregori G Lujan B. Photoreceptor inner/outer segment defect imaging by spectral domain OCT and visual prognosis after macular hole surgery. Invest Ophthalmol Vis Sci . 2010; 51: 1651–1658. [CrossRef] [PubMed]
Boykov Y Jolly M. Interactive graph cuts for optimal boundary & region segmentation of objects in N-D images. In: Proceedings Eighth IEEE International Conference on Computer Vision 2001 . Los Alamitos, CA: IEEE Computer Society; 2001: 741.
Boykov Y Veksler O Zabih R. Fast approximate energy minimization via graph cuts. IEEE Trans Pattern Anal Mach Intell . 2001; 23: 1222–1239. [CrossRef]
Mikajiri K Okada AA Ohji M Analysis of vitrectomy for idiopathic macular hole by optical coherence tomography. Am J Ophthalmol . 1999; 128: 655–657. [CrossRef] [PubMed]
Ullrich S Haritoglou C Gass C Macular hole size as a prognostic factor in macular hole surgery. Br J Ophthalmol . 2002; 86: 390–393. [CrossRef] [PubMed]
Terasaki H Miyake Y Tanikawa A Focal macular electroretinograms before and after successful macular hole surgery. Am J Ophthalmol . 1998; 125: 204–213. [CrossRef] [PubMed]
Amari F Ohta K Kojima H Yoshimura N. Predicting visual outcome after macular hole surgery using scanning laser ophthalmoscope microperimetry. Br J Ophthalmol . 2001; 85: 96–98. [CrossRef] [PubMed]
Apostolopoulos MN Koutsandrea CN Moschos MN Evaluation of successful macular hole surgery by optical coherence tomography and multifocal electroretinography. Am J Ophthalmol . 2002; 134: 667–674. [CrossRef] [PubMed]
Ooka E Mitamura Y Baba T Foveal microstructure on spectral-domain optical coherence tomographic images and visual function after macular hole surgery. Am J Ophthalmol . 2011; 152: 283–290. [CrossRef] [PubMed]
Grigoropoulos VG Theodossiadis GP Theodossiadis PG. Association of the preoperative photoreceptor layer defect as assessed by optical coherence tomography with the functional outcome after macular hole closure: a long follow-up study. Ophthalmologica . 2011; 225: 47–54. [CrossRef] [PubMed]
Footnotes
 Supported by grant from Research to Prevent Blindness (DX, PKK).
Footnotes
 Disclosure: D. Xu, P; A. Yuan, None; P.K. Kaiser, P; S.K. Srivastava, P; R.P. Singh, None; J.E. Sears, None; D.F. Martin, None; J.P. Ehlers, P
Figure 1. 
 
Schematic diagram of segmentation algorithm.
Figure 1. 
 
Schematic diagram of segmentation algorithm.
Figure 2. 
 
(A) Foveal spectral domain optical coherence tomography cross-section of a 45-year-old patient with stage 4 MH. (B) Delineation of the RPE (blue) and internal limiting membrane (red) constraint boundaries. (C) Foreground seed pixels (red) are established by eroding the segmentation result from the adjacent frame. Background seed pixels (blue) are established by selecting pixels a fixed distance from the segmentation result from the adjacent frame. (D) Segmentation result (red) is the solution of the graph cut.
Figure 2. 
 
(A) Foveal spectral domain optical coherence tomography cross-section of a 45-year-old patient with stage 4 MH. (B) Delineation of the RPE (blue) and internal limiting membrane (red) constraint boundaries. (C) Foreground seed pixels (red) are established by eroding the segmentation result from the adjacent frame. Background seed pixels (blue) are established by selecting pixels a fixed distance from the segmentation result from the adjacent frame. (D) Segmentation result (red) is the solution of the graph cut.
Figure 3. 
 
3D representation of MH from Figure 2 with highlighted top area and top diameter measured across the largest axis (A), minimum diameter (B), and base area and base diameter (C).
Figure 3. 
 
3D representation of MH from Figure 2 with highlighted top area and top diameter measured across the largest axis (A), minimum diameter (B), and base area and base diameter (C).
Figure 4. 
 
SD-OCT B-scan and 3D representation of a stage 1 MH in a 73-year-old patient with 20/70 acuity (A, B). A 74-year-old patient with a stage 2 MH and 20/50 acuity (C and D). A 73-year-old patient with a stage 3 MH and 20/70 acuity (E, F). A 52-year-old patient with a stage 4 MH and 20/100 acuity (G, H).
Figure 4. 
 
SD-OCT B-scan and 3D representation of a stage 1 MH in a 73-year-old patient with 20/70 acuity (A, B). A 74-year-old patient with a stage 2 MH and 20/50 acuity (C and D). A 73-year-old patient with a stage 3 MH and 20/70 acuity (E, F). A 52-year-old patient with a stage 4 MH and 20/100 acuity (G, H).
Table 1. 
 
Mean ± SD Measurements of MH Size
Table 1. 
 
Mean ± SD Measurements of MH Size
Stage 1 Stage 2 Stage 3 Stage 4 P Value
Mean volume, mm3 0.026 ± 0.014 0.037 ± 0.017 0.089 ± 0.071 0.179 ± 0.130 0.0005
Mean base area, mm2 0.147 ± 0.082 0.101 ± 0.088 0.364 ± 0.288 0.837 ± 0.668 0.0004
Mean base diameter, μm 511 ± 145 416 ± 185 743 ± 285 993 ± 476 0.001
Mean top area, mm2 0.189 ± 0.168 0.285 ± 0.194 0.345 ± 0.185 0.511 ± 0.066 0.02
Mean top diameter, μm 738 ± 229 752 ± 254 789 ± 189 934 ± 246 0.13
Mean minimum diameter, μm N/A 221 ± 136 323 ± 134 411 ± 172 0.0026
Mean height, μm 189 ± 62 199 ± 79 330 ± 110 401 ± 105 0.0001
Macular hole index 0.370 ± 0.049 0.516 ± 0.148 0.468 ± 0.127 0.515 ± 0.379 0.75
Table 2. 
 
Nonparametric Correlation of BCVA at Baseline and at the 6-Month Postoperative Visit to Measurements of MH Size
Table 2. 
 
Nonparametric Correlation of BCVA at Baseline and at the 6-Month Postoperative Visit to Measurements of MH Size
Spearman Correlation to Baseline BCVA Spearman Correlation to 6-Month Postoperative BCVA
Volume 0.488 (0.0003) 0.306 (0.029)
Base area 0.443 (0.0011) 0.470 (0.0005)
Base diameter 0.353 (0.011) 0.371 (0.0074)
Top area 0.365 (0.0084) 0.110 (0.442)
Top diameter 0.384 (0.0055) 0.079 (0.580)
Minimum diameter 0.393 (0.0043) 0.264 (0.061)
Height 0.369 (0.0077) 0.248 (0.80)
Macular hole index −0.198 (0.163) −0.380 (0.0059)
Table 3. 
 
Multiple Regression Analysis of Age, Baseline BCVA, and MH Measurements to 6-Month Postoperative Visual Acuity
Table 3. 
 
Multiple Regression Analysis of Age, Baseline BCVA, and MH Measurements to 6-Month Postoperative Visual Acuity
P Value
Age, y 0.52
Baseline BCVA, logMAR 0.26
Volume, mm3 0.0028
Base area, mm2 <0.0001
Top area, mm2 0.61
Height, μm 0.97
Minimum diameter 0.82
Macular hole index 0.97
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×