Revision Surgery for Idiopathic Macular Hole after Failed Primary Vitrectomy

Non-Invasive Retinal Imaging Modalities for the Identification of Prognostic Factors in Vitreoretinal Surgery for Full-Thickness Macular Holes

In this review, we will focus on different non-invasive retinal imaging techniques that can be used to evaluate morphological and functional features in full-thickness macular holes with a prognostic purpose. Technological innovations and developments in recent years have increased the knowledge of vitreoretinal interface pathologies by identifying potential biomarkers useful for surgical outcomes prediction. Despite a successful surgery of full-thickness macular holes, the visual outcomes are often puzzling, so the study and the identification of prognostic factors is a current topic of interest. Our review aims to provide an overview of the current knowledge on prognostic biomarkers identified in full-thickness macular holes by means of different retinal imaging tools, such as optical coherence tomography, optical coherence tomography angiography, microperimetry, fundus autofluorescence, and adaptive optics.

Prognostic Factors for Visual Outcomes in Closed Idiopathic Macular Holes after Vitrectomy: Outcomes at 4 Years in a Monocentric Study

Purpose

To identify predictive factors of visual outcomes in the eyes after successful macular hole (MH) surgery.

Methods

It is a retrospective monocentric study of the eyes that underwent successful vitrectomy for full-thickness MH in an academic, tertiary care center (CHU de Québec–Université Laval, Québec, Canada) between 2014 and 2018. We included a single eye per patient and excluded the eyes with ocular comorbidities. Clinical and anatomical features of patients were collected, including demographics, MH duration, baseline MH size, baseline visual acuity (VA), and final VA. Multiple logistic regressions were performed to determine predictive factors of VA ≥70 ETDRS letters (Snellen equivalent: 20/40) and VA gain ≥15 ETDRS letters at final follow-up. Areas under the receiver operating characteristic curve (AUC) were used to determine the performance of each model and identify the Youden index maximizing performance at a given threshold.

Results

A total of 460 eyes were included in this study; 274/460 eyes (60%) achieved final VA ≥70 ETDRS letters and 304/460 eyes (66%) had a VA gain ≥15 ETDRS letters at 24 months follow-up. Multiple logistic regression analyses showed that the main predictive factors for final VA ≥70 ETDRS letters (model AUC = 0.716) were baseline VA (OR = 1.064; p < 0.001), MH duration (OR = 0.950; p=0.005), and age (OR = 0.954; p=0.004). Predictors of VA gain ≥15 ETDRS letters at final follow-up (model AUC = 0.615) were baseline VA (OR = 0.878; p < 0.001), MH duration (OR = 0.940; p < 0.001), and MH size (OR = 0.998; p=0.036). Thresholds for the final VA ≥70 ETDRS letters model and the VA gain ≥15 ETDRS letters model were VA ≥55.5 ETDRS letters (Snellen equivalent: 6/30) and MH size of 237 μm, respectively.

Conclusion

The eyes with shorter MH duration, smaller MH size, and higher preoperative VA achieved better visual outcomes after successful MH surgery.

Predicting Visual Improvement After Macular Hole Surgery: A Combined Model Using Deep Learning and Clinical Features

Purpose

The purpose of this study was to assess the feasibility of deep learning (DL) methods to enhance the prediction of visual acuity (VA) improvement after macular hole (MH) surgery from a combined model using DL on high-definition optical coherence tomography (HD-OCT) B-scans and clinical features.

Methods

We trained a DL convolutional neural network (CNN) using pre-operative HD-OCT B-scans of the macula and combined with a logistic regression model of pre-operative clinical features to predict VA increase ≥15 Early Treatment Diabetic Retinopathy Study (ETDRS) letters at 6 months post-vitrectomy in closed MHs. A total of 121 MHs with 242 HD-OCT B-scans and 484 clinical data points were used to train, validate, and test the model. Prediction of VA increase was evaluated using the area under the receiver operating characteristic curve (AUROC) and F1 scores. We also extracted the weight of each input feature in the hybrid model.

Results

All performances are reported on the held-out test set, matching results obtained with cross-validation. Using a regression on clinical features, the AUROC was 80.6, with an F1 score of 79.7. For the CNN, relying solely on the HD-OCT B-scans, the AUROC was 72.8 ± 14.6, with an F1 score of 61.5 ± 23.7. For our hybrid regression model using clinical features and CNN prediction, the AUROC was 81.9 ± 5.2, with an F1 score of 80.4 ± 7.7. In the hybrid model, the baseline VA was the most important feature (weight = 59.1 ± 6.9%), while the weight of HD-OCT prediction was 9.6 ± 4.2%.

Conclusions

Both the clinical data and HD-OCT models can predict postoperative VA improvement in patients undergoing vitrectomy for a MH with good discriminative performances. Combining them into a hybrid model did not significantly improve performance.

Translational Relevance

OCT-based DL models can predict postoperative VA improvement following vitrectomy for MH but fusing those models with clinical data might not provide improved predictive performance.