Multi-Quantitative Assessment of AS-OCTA Complemented AS-OCT for Monitoring Filtering Bleb Function After Trabeculectomy

Abstract

Purpose: The purpose of this study was to explore a quantitative grading system of the filtering bleb combined anterior segment optical coherence tomography angiography (AS-OCTA) vascular features and optical coherence tomography (OCT) morphological features.

Methods: One hundred three eyes of 103 patients diagnosed with primary open-angle glaucoma and undergone trabeculectomy over 6 months were divided into success and failure groups according to postoperative intraocular pressure (IOP) level. Vessel density (VD) and vessel diameter index (VDI) were examined by AS-OCTA. Bleb's morphology, including bleb height (BH), and microcyst-structure (MCS) were detected by AS-OCT. Multi-vascular model score (MVMS) was calculated by comprehensive factor analysis, and the comprehensive grading system (MVMS-MCS-BH) was analyzed by linear regression. The efficiency our method was verified by receiver operating characteristic (ROC) analysis.

Results: The VD and VDI were higher in the failure group and closely related to post-trabeculectomy IOP (all P = 0.000). The MVMS was mostly consisted of VD in all regions, and VDIs of nasal, central, and temporal positions in sequence. MVMS ≥0, BH <1.33, and non-MCS were significantly associated with IOP increasing (coefficient = –3.23, –3.69, and 8.10, all P = 0.000). MVMS-BH-MCS got a higher area under curve (AUC), sensitivity, and specificity (0.92, 100%, and 80.30%) than the slit-lamp method (0.62, 72.20%, and 46.43%, respectively).

Conclusions: The quantitative vascular characteristics detected by AS-OCTA were significant for the bleb monitor. The MVMS-BH-MCS grading system had achieved outstanding accuracy in reflecting the surgical results.

Translational Relevance: The multi-vascular biomarker and comprehensive evaluation combined vascular and morphological parameters yield useful information on surgical outcomes, and help ophthalmologists to monitor patients effectively.

Introduction

Glaucoma causes irreversible damage to the optic nerve and visual field.^1,2 In addition, trabeculectomy remains the standard surgical approach to decrease IOP.³ The evaluation and maintenance of filtering blebs are crucial for effective outcome.^4,5 

The function of filtering blebs is relevant to the condition of conjunctival fibrosis, tissue remodeling, and angiogenesis in the surgical area.^6–8 Vascularization of the filtration bleb is the key factor in scar formation, particularly vascularization located between the scleral flap and conjunctiva, and this process is basically stable until 6 months after surgery.⁹ Objective indicators are helpful to evaluate the function of filtering blebs and predict the surgical results.^10,11 

The slit-lamp classification was the most popular approach for filtering bleb assessment, including the Indiana Bleb Appearance Grading Scale (IBAGS) and Moorfields Bleb Grading System (MBGS).^12–15 Neither of them provide quantitative assessments of bleb vascularization. Optical coherence tomography angiography (OCTA) has been created and realized the noninvasive form of angiography.^16,17 In recent years, anterior segment OCTA (AS-OCTA) has gained acceptance, and its value has been validated in various anterior segment diseases, including glaucoma filtering surgery (GFS) and minimally invasive glaucoma surgery (MIGS).^18–25 Previous studies found that superficial vascular density (SVD) in the bleb area was more associated with postoperative IOP and surgical outcome than IBAGS and MBGS.^22,26,27 However, none of them have provided a joint analysis of the vascular and morphological parameters. 

This study aimed to explore a complemented grading system of the filtering blebs combined AS-OCTA vascular features and OCT morphological features. The system could yield useful information on surgical outcomes and help ophthalmologists monitor patients accurately. 

Methods

Patient Selection

This study was designed as a cross-sectional study. Patient recruitment was from December 2020 to October 2021 at the Zhongshan Ophthalmic Center of Sun Yat-Sen University. The study was approved by the Institutional Review Board of Zhongshan Ophthalmic Center of Sun Yat-Sen University (Ethics ID. 2020KYPJ119). The study strictly adhered to the Declaration of Helsinki while being clinically registered (NCT 04515017). In addition, every patient gave written informed consent before the study began. 

For all patients enrolled, data were obtained on the visual acuity, age, sex, IOP detected by Goldmann applanation tonometer (Haag-Streit, Bern, Switzerland), postoperative medication types, postoperative duration, and the use of postoperative prostaglandin (PG; Table 1). The inclusion criteria were as follows: (1) age ≥18 years and the ability to complete all examinations; (2) patients with primary open-angle glaucoma (POAG); and (3) completed trabeculectomy assisted by mitomycin-C (MMC) for more than 6 months. The exclusion criteria were as follows: (1) ocular surgery other than trabeculectomy; (2) a history of ocular trauma, ocular surface inflammation, or other diseases that could influence the vascular status of the conjunctiva, Tenon's capsule, and sclera layers; (3) the patient was unable to complete the image process; and (4) systemic diseases that meant the examination could not be completed. 

Table 1.

 

Demographic and Baseline Characteristics of Study Subjects

 

The limbus-based trabeculectomy assisted MMC was performed by a glaucoma specialist (authors Y.H.Z.), as previously described.²⁸ Briefly, the operator made a rectangular scleral flap at the size of 3 × 4 mm, and the cotton containing MMC was put for 2 minutes (Hanhui Pharmaceuticals Co., Ltd, China; 10 mg/vial). A piece of trabecular meshwork was removed, and a peripheral iridectomy was performed. According to the World Glaucoma Association (WGA) criteria, success group was defined according to the rule that 6 ≤ IOP ≤ 21 mm Hg and a reduction in IOP ≥ 20% (with or without anti-glaucoma medication). 

Outcome Measures

Bleb Vascularity Evaluation Using OCTA

AS-OCTA was performed by a single trained operator using OCTA scanning (6 × 6 mm HD Angio Disc protocol) with XR Avanti and AngioVue software (Optovue, Inc., Fremont, CA, USA), as our previous described.²⁸ The Avanti OCT used the separated spectrum amplitude decorrelation algorithm (SSADA).²⁹ It used an 840-nm centered light source capable of 70,000 A-scans per second, and had a 5 µm axial resolution. For the anterior segment, a long corneal adapter (CAM-L) was used to obtain AS-OCTA images. For each bleb, three 6 × 6 mm acquisitions were performed at the center, temporal, and nasal places of the filtering bleb (Figs. 1A, 1B). The size of the 6 × 6 mm scan pattern in typical mode corresponds to 9 × 9 mm in AS-OCTA images (Fig. 1C). After adjusting the F and Z motor settings (with F = –15D, Z = +9.38) and canceling the automatic tracking mode, the patient was required to look at the external fixed eye position indicator throughout the whole scan. 

Figure 1.

 

Acquisition of AS-OCTA. (A, B) The 6 × 6 mm HD Angio Disc scanning performed in the center of the scleral flap. (C) The 6 × 6 mm AS-OCTA images with division value of 1 mm ruler. (D) The B-scan image showing the superficial (from the conjunctival epithelium to a depth of 150 µm), Tenon's (from a depth of 150 µm to 250 µm), and deep layer (from a depth of 150 µm to 1000 µm) location. AS-OCTA = anterior segment optical coherence tomography angiography.

En face images were delimited by two parallel curved lines that were manually separated. These two lines could be moved from the surface (conjunctival epithelium) to deeper layers (sclera) to explore the different layers of the filtering bleb. Superficial, Tenon's, and deep layer vascular images were acquired by setting the position from the conjunctival epithelium to a depth of 150 µm, from a depth of 150 µm to 250 µm, and from a depth of 150 µm to 1000 µm manually (Fig. 1D). 

The vessel data from AS-OCTA images were generated using the AS-OCTA instrument and quantified using ImageJ software (version 1.48; https://imagej.net/ij/index.html). A signal strength index >60 was considered for the analysis.²³ Low-quality scans were defined as saccade or blinking artifacts disturbing vascularization analysis, and the eyelid-captured part was excluded from the analysis. Three image files were required for the measurements: superficial, Tenon's, and deep vascular layer images. The vessel density (VD) was calculated as a unitless ratio of the total image area occupied by the vessels after binarization of the images. Vessel diameter index (VDI) was calculated as 10 times the ratio of the total image area occupied by the skeletonized vessel area.^28,30 

Bleb Morphology Evaluation Using AS-OCT

Bleb morphology was determined by corneal linear schemes of AS-OCT (AngioVue software; Optovue, Inc., Fremont, CA, USA). The scanning line was set to connect the midpoints of the two sides of the scleral flap. Bleb height (BH) was defined as the longest distance between the conjunctival epithelium and surface of the scleral flap. Microcyst-structure (MCS) was identified by a hyporeflective area available in the bleb wall with a diameter larger than 10 µm (Fig. 2).^31,32 

Figure 2.

 

Bleb morphology evaluation method. (A, B) The scanning position of filtering blebs detected by AS-OCT. (C) AS-OCT image showing the measuring method of bleb height (yellow line) and microcysts (white triangles). AS-OCT = anterior segment optical coherence tomography.

The attending doctor (author H.X.) performed the scanning of AS-OCT and AS-OCTA. In addition, the resident doctor (author M.L.) acquired the images by depth and analyzed the VD, VDI, BH, and MCS. All measurements were obtained from 2:30 PM to 5:00 PM to avoid the effects of IOP fluctuation. The same position was scanned three times for every patient, and the mean values were entered into the analysis. 

Traditional Classification Method Using Slit-lamp Microscopy

This study utilized the vascularity grading from IBAGS, which was classified: V0 = avascular white; V1 = avascular cystic; V2 = mild vascularity; V3 = moderate vascularity; and V4 = extensive vascularity. The filtering bleb's vessel was also analyzed by MBGS method with: 1 = avascular; 2 = normal; 3 = mild; 4 = moderate; and 5 = severe. 

Statistical Analysis

SPSS version 25.0 was used for analysis (SPSS, Chicago, IL, USA). The Shapiro–Wilk test was applied to test the distribution of variables. Non-normality data were described as median (interquartile range [IQR]). The Mann-Whitney U test, Kruskal-Wallis test, and Chi-square test were used to evaluate the differences between non-normally distributed variables. Spearman's correlation analysis was used for the correlation analysis. For factor analysis, principal component analysis (PCA) was used to generate the factors. The sample was considered adequate to perform factor analysis when the Kaiser-Meyer-Olkin (KMO) test was >0.5. The Bartlett test of sphericity was used to determine the homogeneity of the data. Varimax rotation was used after the initial factor solution. The optimal number of factors was assessed from the screen plot.³³ The multi-vascular model score (MVMS) was calculated by comprehensive factor analysis. The cutoff points of BH and MVMS were 1.33 mm and the value of 0 assessed by “minimum P value method.”^34,35 The factors affecting IOP, including age, sex, MCS, BH, and MVMS, were analyzed using univariate and multivariate linear regression from stepwise regression. Operating characteristic curve (ROC), area under the curve (AUC), and Youden index were calculated. Statistical significance was set at P < 0.05. 

Results

Patient Characteristics

Initially, 118 patients met the inclusion and exclusion criteria. Among these patients, eight patients refused AS-OCTA scanning, and seven patients had poor cooperation at the time of the examination. Ultimately, we enrolled 103 eyes of 103 patients with glaucoma for analysis. 

The failure group had higher IOP, and longer postoperative duration than the success group, as shown in Table 1 (P = 0.000 and 0.006, respectively). Other baseline parameters, including age, sex, postoperative medication type, and the rate of PG administration, were not significantly different between the success and failure groups (see Table 1). 

Quantitative Evaluation of Filtering Blebs

VD and VDI of the superficial, Tenon's, and deep layers were higher in the failure group than in the success group at all positions of the filtering blebs (all P = 0.000; Table 2, Fig. 3). Successful blebs had more MCS than failed blebs (P = 0.005), but BH showed no difference between the two groups (P = 0.357; Table 3). 

Table 2.

 

Vascular Parameters Between Successful and Failed Group

 

Figure 3.

 

Cases showing anterior segment photographs, and AS-OCTA images. (A) Anterior segment photograph of the success group. (B-D) AS-OCTA images in superficial layer, Tenon's layer, and deep layer of the success group. (E) Anterior segment photograph of the failure group. (F-H) AS-OCTA images in superficial layer, Tenon's layer, and deep layer of the failure group. AS-OCTA = anterior segment optical coherence tomography angiography.

Table 3.

 

Morphological Parameters in Bleb Area Between Successful and Failed Group

 

Factor Analysis of Bleb Vascularity Parameters

The factor analysis of the bleb vascularity parameters is shown in Table 4. For these parameters, a four-factor solution was the most tenable based on the screen plot. The first factor comprised superficial vessel density (SVD), Tenon's vessel density (TVD), and deep vessel density (DVD) in the bleb and peri-bleb areas. The second factor was contamination superficial vessel diameter index (SVDI) and Tenon's vessel diameter index (TVDI) in the nasal area. The third factor was SVDI, TVDI, and the deep vessel diameter index (DVDI) in the bleb. The fourth factor was identical to the TVDI and DVDI of the temporal bleb. These factors were named VD (factor 1), nasal VDI (factor 2), central VDI (factor 3), and temporal VDI (factor 4). This comprehensive principal component model was named MVMS, and its formula can be defined as follows:  

\begin{eqnarray*} && {\rm Y = 38.05\% \; factor \; 1 + 22.33\% \; factor \; 2} \nonumber \\ && {\rm \; + 22.19\% \; factor \; 3 + 17.43\% \; factor \;4.} \end{eqnarray*}

 

Table 4.

 

Factor Analysis of Vascular Parameters of Filtering Blebs

 

Correlation Analysis Between Parameters and IOP

There was a statistically significant correlation between IOP and VD and VDI in the superficial, Tenon's, and deep layers at any position of the bleb (Table 5). Comparing the correlation among the three layers, VD in Tenon's layer was most correlated with IOP in the relative area (r = 0.51, 0.55, and 0.65, all P = 0.000). Meanwhile, comparing the quantitative vascular patterns at the center and periphery of the filtering bleb, VD in the nasal position showed a relatively greater association with IOP (r = 0.58, 0.65, and 0.63, all P = 0.000). In addition to the vascularity evaluation, MCS and BH were also correlated with IOP (r = –0.353, –0.277, P = 0.000, 0.005) in Table 6. 

Table 5.

 

Correlation Between the Vascular Parameters and IOP

 

Table 6.

 

Correlation Among the MVMS, Morphological Parameters, and IOP

 

We also analyzed the correlation between MVMS and IOP, and MVMS was the multi-vascular parameter calculated from the factor analysis. MVMS had the highest r value (r = 0.67) compared with the individual VD and VDI parameters and morphological parameter detected by AS-OCT (see Table 6). 

Quantitative Predictive Factors for Postoperative IOP

Based on the “minimum P value method,” the breaking point of MVMS was zero. Age and sex were not associated with IOP in the single linear regression analysis (P = 0.420 and 0.499, respectively). Meanwhile, MCS, BH, and MVMS were performed using multiple linear regression analysis. The appearance of MCS, BH ≥1.33, and MVMS ≥0 were the factors significantly associated with IOP (coefficient [95% confidence interval {CI}] = –3.23 [–5.78 to –0.68], –3.69 [–6.18 to –1.21], and 8.10 [5.57 to 10.62], all P = 0.000; Table 7). Comparing these three parameters, high-level MVMS (≥0) of filtering blebs showed an 8.10-fold increased risk of IOP elevation compared to patients with low-level MVMS (<0). The other two morphological patterns scanned by AS-OCT could also affect the bleb function. In detail, non-MCS and low-level BH (<1.33 mm) could lead to a 3.23-fold and 3.69-fold risk, respectively, for failed IOP control. 

Table 7.

 

Linear Regression Analysis for IOP

 

Evaluation of Clinical Associations of Quantitative Parameters

The ROC analysis with respect to the operation results was shown in Table 8 and Figure 4. MVMS had higher AUC (0.90) than classic slit-lamp methods (0.62 for IBAGS, and 0.61 for MBGS). The MVMS-BH-MCS got the highest AUC, sensitivity, and specificity than other methods which equaled to 0.92, 100%, and 80.30%, respectively. 

Table 8.

 

ROC Analysis for Success and Failure Outcomes

 

Figure 4.

 

The ROC curve for MVMS-BH-MCS, MVMS, BH, MCS, IBAGS, and MBGS with respect to the surgery results. ROC = receiver operating characteristic curve; MVMS = multi-vascular model score; BH = bleb height; MCS = microcyst-structure; IBAGS = Indiana bleb appearance grading scale; MBGS = Moorfields Bleb Grading System; AUC = area under curve; CI = confidence interval.

The Youden index of MVMS-BH-MCS was 17.61. Besides, the corresponding IOP was closed to 15 mm Hg determined by the loess plot which less than the 21 mm Hg (Fig. 5). 

Figure 5.

 

The loess plot of IOP and MVMS-BH-MCS. The cutoff value of MVMS performed at the value of 17.61, the corresponding IOP value is closed to 15 mm Hg. IOP = intraocular pressure; MVMS = multi-vascular model score; BH = bleb height; MCS = microcyst-structure.

Discussion

We explored a quantitative multi-vascular model of filtering blebs, and a comprehensive grading system combining MVMS BH, and MCS detected by AS-OCT and AS-OCTA. Denser and thicker blood vessels were associated with the failure of filtering blebs. The VD and VDI were correlated with post-trabeculectomy IOP, especially for the Tenon's layer. The MVMS was mostly consisted of VD of all regions, and VDIs of the nasal, central, and temporal positions in sequence. MVMS-BH-MCS complemented grading system yielded early-warning information of the transition of IOP, and got more outstanding accuracy in reflecting surgery outcomes than the slit-lamp method. 

The failed filtering blebs always presented more abundant and thicker vessels. In addition, the VD and VDI were positively related to post-surgical IOP, especially for the Tenon's layer and nasal region toward blebs. A highly increased conjunctival and episcleral vascularity was a sign of surgical failure, and the most significant advantages of AS-OCTA was to provide quantitative information on subconjunctival vessels.²² The assessment of central filtering bleb vessels had predictive value to surgical outcomes.^26,36 However, previous AS-OCTA researches lacked systematic analysis of vascularity containing different depths and regions. 

We explored a multi-vascular model that combined all quantitative vessel parameters. The variety of vessels differentiated by depths and locations of filtering bleb yielded useful information.^{11–14,37–39} So, our MVMS model included whole vascular indexes and separated them into four principal factors: VD, nasal VDI, central VDI, and temporal VDI. In addition, the MVMS could assist clinicians in overall, standardized, and accurate evaluation of vascularization severity. 

We investigated a complemented grading system of quantitative parameters (MCS, BH, and MVMS) that could reflect the post-trabeculectomy IOP. We demonstrated that filtering blebs with overall highly vascularized blebs (MVMS ≥0), flat morphology (BH <1.33 mm), and non-MCS could reflect a higher uncontrolled IOP value. The effect of bleb vascularization evaluated by MVMS on postoperative IOP was almost twice that of morphological patterns, which suggested the significance of vascular evaluation. So far, researchers have dedicated to develop a more accurate approaches to assess the filtering blebs function involving bleb height, bleb range, central/peri-bleb vascularization, microcysts, and Seidel test.^{12–15,40,41} Notably, there has been no quantitative and objective grading methods combined for the morphologic and vascular signs. In addition, our comprehensive (MVMS-MCS-BH) grading system showed closed association with postoperative IOP, which could provide a standardized observational indicator of bleb function, especially for primary ophthalmologists (Fig. 6). 

Figure 6.

 

A case with failed filtering bleb by IOP = 26 mm Hg. (A) SLM showing avascular cystic appearance which could be misdiagnosed as functional filtering blebs. (B) AS-OCT showing low-level height filtering bleb accompanied MCS. (C) AS-OCTA image in Tenon's layer with more detailed detection of deep-layer vascularization (white triangle) especially on the scleral flap (white arrow). MVMS is high level, and MVMS-MCS-BH grading system diagnoses a failed filtering bleb matching postoperative IOP. IOP = intraocular pressure; SLM = slit-lamp microscopy; AS-OCT = anterior segment optical coherence tomography; MCS = microcyst-structure; AS-OCTA = anterior segment optical coherence tomography angiography; BH = bleb height; MVMS = multi-vascular modal score.

The diagnostic efficiency of MVMS-BH-MCS grading system was better than that of the slit-lamp classification and single indicators. MVMS-BH-MCS showed the highest AUC, sensitivity, and specificity as 0.92, 100%, and 80%, respectively. However, those of IBAGS and MBGS were at most 0.62, 72%, and 61%. The previous study showed a better correlation between VD and postoperative IOP results than that obtained with the IBAGS and MBGS.²⁸ The slit-lamp classifications were subjective, and their AUC was relatively unsatisfactory. MVMS-BH-MCS presented by the Loess plot, a value of 17.61 was used as an early warning biomarker of the functional change of the filtering blebs. When MVMS-BH-MCS reached the level of the classic failed criterion of 21 mm Hg, the postoperative IOP had already entered the linear rising phase. This indicated that MVMS-BH-MCS could be an early biomarker to foresight the uncontrollable increase in postoperative IOP. 

Our study has several limitations. First, projection artifacts are commonly observed in OCTA images. Previous studies have investigated methods for artifact elimination, including projection-resolved (PR) algorithms, slab-subtraction (SS) algorithms, and others.^42,43 However, there is no universal standard now, and current approaches may still have limitations, such as causing loss of capillary signals near the projected artifact.⁴⁴ Further research and discussion are needed to develop specific techniques for anterior segment angiography and address the problem of projection artifacts in AS-OCTA imaging. Second, the study was based on clinical data of a single research institution, the performance of AS-OCTA was needed to verify by multicenter trials. Third, our study was a cross-sectional study focused on a post-trabeculectomy patient, a prospective cohort study was needed to the broad significance of the combined morphological and vascular parameters. 

Conclusion

In conclusion, quantitative evaluation was essential for assessing the function of filtering blebs after trabeculectomy. The comprehensive system combining MVMS and morphological features was significant for precise monitoring of filtering blebs, and help ophthalmologists monitor patients accurately. 

Acknowledgments

Supported by the National Key Research and Development Program of China (2020YFA0112701); the National Natural Science Foundation of China (82171057 and 81870657); Science and Technology Program of Guangzhou, China (202206080005); and the Natural Science Foundation of Guangdong Province (2022A1515012168). 

Disclosure: M. Luo, None; H. Xiao, None; J. Huang, None; L. Jin, None; Z. Li, None; S. Tu, None; H. Huang, None; Y. Zhu, None; Y. Li, None; Y. Zhuo, None