Information loss in 2 bleb grading systems

A deep learning approach to investigate the filtration bleb functionality after glaucoma surgery: a preliminary study

PURPOSE

To distinguish functioning from failed filtration blebs (FBs) implementing a deep learning (DL) model on slit-lamp images.

METHODS

Retrospective, cross-sectional, multicenter study for development and validation of an artificial intelligence classification algorithm. The dataset consisted of 119 post-trabeculectomy FB images of whom we were aware of the surgical outcome. The ground truth labels were annotated and images splitted into three outcome classes: complete (C) or qualified success (Q), and failure (F). Images were prepared implementing various data cleaning and data transformations techniques. A set of DL models were trained using different ResNet architectures as the backbone. Transfer and ensemble learning were then applied to obtain a final combined model. Accuracy, sensitivity, specificity, area under the ROC curve, and area under the precision-recall curve were calculated to evaluate the final model. Kappa coefficient and P value on the accuracy measure were used to prove the statistical significance level.

RESULTS

The DL approach reached good results in unraveling FB functionality. Overall, the model accuracy reached a score of 74%, with a sensitivity of 74% and a specificity of 87%. The area under the ROC curve was 0.8, whereas the area under the precision-recall curve was 0.74. The P value was equal to 0.00307, and the Kappa coefficient was 0.58.

CONCLUSIONS

All considered metrics supported that the final DL model was able to discriminate functioning from failed FBs, with good accuracy. This approach could support clinicians in the patients' management after glaucoma surgery in absence of adjunctive clinical data.

Angiographic biomarkers of filtering bleb function after XEN gel implantation for glaucoma: an optical coherence tomography-angiography study

PURPOSE

To evaluate, using optical coherence tomography-angiography (OCT-A), the vascular features of good bleb function after XEN gel implantation (XGI) for uncontrolled glaucoma.

METHODS

Forty-three patients (43 eyes), who underwent XGI, were enrolled. According to the intraocular pressure (IOP) reduction, patients were classified into Group 1 (21 eyes; success) and Group 2 (22 eyes; failure). Optical coherence tomography-angiography (OCT-A) was performed to image the vascularization of the conjunctival bleb-wall. The main outcomes were as follows: vessel displacement areas (VDAs), major vessel displacement area (MVDA; mm2 ), non-flow whole area (NFWA; mm2 ) and bleb-wall vessel density (BVD; %). Co-registered B-scans were also considered to evaluate the bleb-wall cyst-like structure density and area (BCSD, cysts/mm2 ; BCSA, mm2 ), and the bleb-wall thickness (BT, µm).

RESULTS

Mean postoperative follow-up was 7.5 ± 0.14 months; Group 1 and 2 IOP were 14.0 ± 2.5 and 25.3 ± 2.1 mmHg, respectively (p < 0.001). Greater VDA (p < 0.001), MVDA (p = 0.046) and NFWA (p = 0.001) values, and lower BVD (p < 0.001) was found in Group 1 compared to Group 2. Group 1 showed higher BSCD, BSCA and BT values compared to Group 2 (p < 0.001). Postoperative IOP positively correlated with BVD (r = 0.567; p = 0.003), but negatively with VDAs, MVDA (r = -0.581, p = 0.002; r = -0.619, p = 0.001, respectively), BCSD, BCSA (r = -0.580; p = 0.002; r = -0.664; p < 0.001) and BT (r = -0.627, p = 0.001).

CONCLUSION

Successful filtration blebs after XGI present numerous and large areas of vessel displacement within the bleb-wall, along with a rarefied vascular network. These OCT-A features can be considered angiographic biomarkers of a good aqueous humour percolation through the bleb-wall layers.

Effect of subconjunctival injection with conbercept as an adjuvant to filtration surgery for open angle glaucoma: a prospective randomized interventional 6-month follow-up study

AIM

To compare the safety and efficacy of subconjunctival injection with conbercept and 5-fluorouracil (5-FU) for open angle glaucoma (OAG) patients after filtration surgery.

METHODS

As a prospective randomized interventional trial, 36 eyes from 36 patients after OAG surgery were collected and divided randomly into conbercept and 5-FU groups. All patients were subconjunctivally injected with either conbercept (0.2 mL) or 5-FU (0.2 mL) on the 5^th day post-operatively. The intraocular pressure (IOP), number of medications used, type of conjunctival bleb, and complications were recorded and analyzed pre-operatively and 1d, 1wk, 1, 3 and 6mo post-injection.

RESULTS

There were significant differences in IOP between the conbercept and 5-FU groups 1mo (conbercept group: 12.17±1.04 mm Hg; 5-FU group: 13.50±2.33 mm Hg, t=2.214, P=0.037), 3mo (conbercept group: 13.00±1.88 mm Hg; 5-FU group: 14.50±2.28 mm Hg, t=2.153, P=0.039), and 6mo post-injection (conbercept group: 13.28±2.95 mm Hg; 5-FU group: 15.22±2.49 mm Hg, t=2.140, P=0.040); however, in the number of medications, a prominent difference was not shown between groups on post-injection 6mo (t=1.312, P=0.200). Moreover, there was mild vascularity observed in the conbecept group than the 5-FU group 1d (3a, 3b, 3c: t=8.497, 6.693, 4.515, P=0.000), 1wk (3a, 3b, 3c: t=3.431, 6.408, 3.984, P=0.002, 0.000, 0.000), and 1mo post-injection (3a, 3b, 3c: t=2.466, 2.466, 2.503, P=0.019, 0.019, 0.017). Simultaneously, differences from other indicators between the two groups were not demonstrated. Also, there was a lower probability of corneal epithelial stripping in the conbercept group than the 5-FU group (χ ²=4.500, P=0.034).

CONCLUSION

Subconjunctival injection of conbercept has a safe, effective, and tolerable profile for open angle glaucoma patients with distinct conjunctival congestion after filtration surgery.

Comparison of limbus-based and fornix-based trabeculectomy: success, bleb-related complications, and bleb morphology

PURPOSE

To compare the success and complications of trabeculectomy performed with limbus-based and fornix-based conjunctival approaches.

DESIGN

Retrospective case series with some prospective data collection.

PARTICIPANTS

Consecutive patients undergoing trabeculectomy by 2 surgeons between May 2000 and October 2008.

INTERVENTION

We performed limbus-based operations during the first 4 years and fornix-based operations during the last 4 years. We collected data by chart review and by examination at the most recent visit. For each follow-up visit, we defined success as undergoing no further glaucoma procedure and achieving one of our intraocular pressure (IOP) criteria. We used Kaplan-Meier survival analysis, Cox proportional hazards models, and generalized estimating equation (GEE) analysis. During 2009, 439 trabeculectomy sites of 347 patients were quantitatively assessed by the Indiana bleb grading system.

MAIN OUTCOME MEASURES

(1) Success rate of trabeculectomy, as determined by the achievement of each of our different IOP goals, with or without IOP-lowering medications; and (2) incidence of surgical complications.

RESULTS

During the 4 years after surgery, the success rates of limbus-based and fornix-based trabeculectomy were not statistically different for any of our IOP criteria. Blebs after limbus-based surgery were more likely to be graded as higher and to be avascular (GEE model, both P < 0.0001). Four percent of eyes experienced late-onset bleb leaks within 4 years after both limbus- and fornix-based operations; however, limbus-based cases developed bleb leaks significantly later than did fornix-based cases (2.1 vs. 1.0 years; P=0.002, GEE model). Late bleb-associated infection during the first 4 years after surgery occurred more often in limbus-based operations, although statistical significance was borderline (P=0.054, Cox model). Symptomatic hypotony during all available follow-up was more common with fornix-based operations (P=0.01, GEE model). Eyes undergoing the fornix-based operation had a greater risk of cataract surgery in the 4-year period after surgery (P=0.02, Cox model), and fornix-based cases requiring cataract surgery had the operation earlier than limbus-based cases (P=0.002, GEE model).

CONCLUSIONS

Success rates are similar between limbus-based and fornix-based trabeculectomy. Limbus-based procedures produce higher, more avascular blebs, with a greater risk of infection. Fornix-based procedures have more symptomatic hypotony and more and earlier cataract development.

Restricted post-trabeculectomy bleb formation by conjunctival scarring

PURPOSE

To evaluate the effects of radial conjunctival incision during fornix-based trabeculectomy, and conjunctival scarring associated with previous ocular surgeries, for the restriction of bleb formation.

METHODS

Digital photo-slit lamp images of filtering blebs, which had undergone fornix-based trabeculectomy with mitomycin C previously, were analysed. The area and the height of the bleb in the conjunctivae with radial incision (Ain and Hin, respectively) were compared with the area and the height of the bleb in the conjunctivae without radial incision (Ano and Hno, respectively). We compared the parameters of bleb area (total bleb area, Ain and Ano) and bleb height (the bleb height at the centre of the scleral flap; Hc, Hin and Hno) in eyes without previous ocular surgeries, with those in eyes that underwent previous ocular surgeries.

RESULTS

The study population consisted of 51 eyes. The Ano and the Hno were significantly larger than the Ain (p < 0.001) and the Hin (p < 0.001) respectively. Subgroup analyses in eyes without previous ocular surgeries and eyes with phacoemulsification cataract surgery also showed that the Ano was significantly larger than the Ain. However, there was no significant difference between the Ano and the Ain in eyes with previous trabeculectomy. A comparison of bleb formation between eyes without previous ocular surgeries and eyes with cataract surgery showed no significant differences in each of the bleb area and bleb height parameters. By contrast, the total bleb area (p = 0.021), the Ano (p = 0.017) and the Hc (p = 0.045) were significantly smaller in eyes with previous trabeculectomy than in eyes without previous ocular surgeries.

CONCLUSIONS

Our data demonstrate that bleb formation depends on radial conjunctival incision during fornix-based trabeculectomy and surgical conjunctival scarring from previous trabeculectomy.