Machine learning-based prediction of anatomical outcome after idiopathic macular hole surgery

Supervised machine learning statistical models for visual outcome prediction in macular hole surgery: a single-surgeon, standardized surgery study

PURPOSE

To evaluate the predictive accuracy of various machine learning (ML) statistical models in forecasting postoperative visual acuity (VA) outcomes following macular hole (MH) surgery using preoperative optical coherence tomography (OCT) parameters.

METHODS

This retrospective study included 158 eyes (151 patients) with full-thickness MHs treated between 2017 and 2023 by the same surgeon and using the same intraoperative surgical technique. Data from electronic medical records and OCT scans were extracted, with OCT-derived qualitative and quantitative MH characteristics recorded. Six supervised ML models-ANCOVA, Random Forest (RF) regression, K-Nearest Neighbor, Support Vector Machine, Extreme Gradient Boosting, and Lasso regression-were trained using an 80:20 training-to-testing split. Model performance was evaluated on an independent testing dataset using the XLSTAT software. In total, the ML statistical models were trained and tested on 14,652 OCT data points from 1332 OCT images.

RESULTS

Overall, 91% achieved MH closure post-surgery, with a median VA gain of -0.3 logMAR units. The RF regression model outperformed other ML models, achieving the lowest mean square error (MSE = 0.038) on internal validation. The most significant predictors of VA were postoperative MH closure status (variable importance = 43.078) and MH area index (21.328). The model accurately predicted the post-operative VA within 0.1, 0.2 and 0.3 logMAR units in 61%, 78%, and 87% of OCT images, respectively.

CONCLUSION

The RF regression model demonstrated superior predictive accuracy for forecasting postoperative VA, suggesting ML-driven approaches may improve surgical planning and patient counselling by providing reliable insights into expected visual outcomes based on pre-operative OCT features.

CLINICAL TRIAL REGISTRATION NUMBER

Not applicable.

Predicting macular hole surgery outcomes: Integrating preoperative OCT features with supervised machine learning statistical models

PURPOSE

To evaluate various supervised machine learning (ML) statistical models to predict anatomical outcomes after macular hole (MH) surgery using preoperative optical coherence tomography (OCT) features.

METHODS

This retrospective study analyzed OCT data from idiopathic MH eyes at baseline and at 1-month post-surgery. The dataset was split 80:20 between training and testing. XLSTAT® statistical software (Lumivero, USA) was used to train different ML models on 10°CT parameters: prefoveal posterior cortical vitreous status, epiretinal membrane, intraretinal cysts, foveal retinal pigment epithelium hyperreflectivity, MH basal diameter, MH area (MHA), hole-forming factor, MH index, tractional hole index, and diameter hole index. The most effective statistical model was identified and was further assessed for accuracy, sensitivity, and specificity on a separate testing dataset.

RESULTS

Six ML statistical models were trained on 33,260°CT data points from 3326°CT images of 308 operated MH (300 patients) eyes. Following training and internal validation, the random forest (RF) model achieved the highest accuracy (0.92), precision (0.94), recall (0.97), and F-score (0.96), and lowest misclassification rate. RF model identified the MHA index as the best predictor of post-surgical anatomical success. Following external testing, the RF model confirmed the highest accuracy and lowest misclassification rate (8.8%).

CONCLUSION

ML-based statistical models can be used to predict MH status after surgery. The RF model was the most accurate ML model, and the MHA index was the best predictor of postoperative hole closure after surgery based on preoperative OCT parameters. These predictions may help with future surgical planning for MH patients.

The role of artificial intelligence in macular hole management: A scoping review

We focus on the utility of artificial intelligence (AI) in the management of macular hole (MH). We synthesize 25 studies, comprehensively reporting on each AI model's development strategy, validation, tasks, performance, strengths, and limitations. All models analyzed ophthalmic images, and 5 (20 %) also analyzed clinical features. Study objectives were categorized based on 3 stages of MH care: diagnosis, identification of MH characteristics, and postoperative predictions of hole closure and vision recovery. Twenty-two (88 %) AI models underwent supervised learning, and the models were most often deployed to determine a MH diagnosis. None of the articles applied AI to guiding treatment plans. AI model performance was compared to other algorithms and to human graders. Of the 10 studies comparing AI to human graders (i.e., retinal specialists, general ophthalmologists, and ophthalmology trainees), 5 (50 %) reported equivalent or higher performance. Overall, AI analysis of images and clinical characteristics in MH demonstrated high diagnostic and predictive accuracy. Convolutional neural networks comprised the majority of included AI models, including those which were high performing. Future research may consider validating algorithms to propose personalized treatment plans and explore clinical use of the aforementioned algorithms.

Machine Learning and Optical-Coherence-Tomography-Derived Radiomics Analysis to Predict the Postoperative Anatomical Outcome of Full-Thickness Macular Hole

Full-thickness macular hole (FTMH) leads to central vision loss. It is essential to identify patients with FTMH at high risk of postoperative failure accurately to achieve anatomical closure. This study aimed to construct a predictive model for the anatomical outcome of FTMH after surgery. A retrospective study was performed, analyzing 200 eyes from 197 patients diagnosed with FTMH. Radiomics features were extracted from optical coherence tomography (OCT) images. Logistic regression, support vector machine (SVM), and backpropagation neural network (BPNN) classifiers were trained and evaluated. Decision curve analysis and survival analysis were performed to assess the clinical implications. Sensitivity, specificity, F1 score, and area under the receiver operating characteristic curve (AUC) were calculated to assess the model effectiveness. In the training set, the AUC values were 0.998, 0.988, and 0.995, respectively. In the test set, the AUC values were 0.941, 0.943, and 0.968, respectively. The OCT-omics scores were significantly higher in the "Open" group than in the "Closed" group and were positively correlated with the minimum diameter (MIN) and base diameter (BASE) of FTMH. Therefore, in this study, we developed a model with robust discriminative ability to predict the postoperative anatomical outcome of FTMH.

Preoperative OCT Characteristics Contributing to Prediction of Postoperative Visual Acuity in Eyes with Macular Hole

Objectives: To develop a machine learning logistic regression algorithm that can classify patients with an idiopathic macular hole (IMH) into those with good or poor vison at 6 months after a vitrectomy. In addition, to determine its accuracy and the contribution of the preoperative OCT characteristics to the algorithm. Methods: This was a single-center, cohort study. The classifier was developed using preoperative clinical information and the optical coherence tomographic (OCT) findings of 43 eyes of 43 patients who had undergone a vitrectomy. The explanatory variables were selected using a filtering method based on statistical significance and variance inflation factor (VIF) values, and the objective variable was the best-corrected visual acuity (BCVA) at 6 months postoperation. The discrimination threshold of the BCVA was the 0.15 logarithm of the minimum angle of the resolution (logMAR) units. Results: The performance of the classifier was 0.92 for accuracy, 0.73 for recall, 0.60 for precision, 0.74 for F-score, and 0.84 for the area under the curve (AUC). In logistic regression, the standard regression coefficients were 0.28 for preoperative BCVA, 0.13 for outer nuclear layer defect length (ONL_DL), -0.21 for outer plexiform layer defect length (OPL_DL) - (ONL_DL), and -0.17 for (OPL_DL)/(ONL_DL). In the IMH form, a stenosis pattern with a narrowing from the OPL to the ONL of the MH had a significant effect on the postoperative BCVA at 6 months. Conclusions: Our results indicate that (OPL_DL) - (ONL_DL) had a similar contribution to preoperative visual acuity in predicting the postoperative visual acuity. This model had a strong performance, suggesting that the preoperative visual acuity and MH characteristics in the OCT images were crucial in forecasting the postoperative visual acuity in IMH patients. Thus, it can be used to classify MH patients into groups with good or poor postoperative visual acuity, and the classification was comparable to that of previous studies using deep learning.

Artificial Intelligence, Digital Imaging, and Robotics Technologies for Surgical Vitreoretinal Diseases

OBJECTIVE

To review recent technological advancement in imaging, surgical visualization, robotics technology, and the use of artificial intelligence in surgical vitreoretinal (VR) diseases.

BACKGROUND

Technological advancements in imaging enhance both preoperative and intraoperative management of surgical VR diseases. Widefield imaging in fundal photography and OCT can improve assessment of peripheral retinal disorders such as retinal detachments, degeneration, and tumors. OCT angiography provides a rapid and noninvasive imaging of the retinal and choroidal vasculature. Surgical visualization has also improved with intraoperative OCT providing a detailed real-time assessment of retinal layers to guide surgical decisions. Heads-up display and head-mounted display utilize 3-dimensional technology to provide surgeons with enhanced visual guidance and improved ergonomics during surgery. Intraocular robotics technology allows for greater surgical precision and is shown to be useful in retinal vein cannulation and subretinal drug delivery. In addition, deep learning techniques leverage on diverse data including widefield retinal photography and OCT for better predictive accuracy in classification, segmentation, and prognostication of many surgical VR diseases.

CONCLUSION

This review article summarized the latest updates in these areas and highlights the importance of continuous innovation and improvement in technology within the field. These advancements have the potential to reshape management of surgical VR diseases in the very near future and to ultimately improve patient care.

FINANCIAL DISCLOSURE(S)

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Novel Artificial Intelligence-Based Assessment of Imaging Biomarkers in Full-Thickness Macular Holes: Preliminary Data from a Pivotal Trial

Artificial intelligence (AI)- and deep learning (DL)-based systems have shown significant progress in the field of macular disorders, demonstrating high performance in detecting retinal fluid and assessing anatomical changes during disease progression. This study aimed to validate an AI algorithm for identifying and quantifying prognostic factors in visual recovery after macular hole (MH) surgery by analyzing major optical coherence tomography (OCT) biomarkers. This study included 20 patients who underwent vitrectomy for a full-thickness macular hole (FTMH). The mean diameter of the FTMH was measured at 285.36 ± 97.4 μm. The preoperative best-corrected visual acuity (BCVA) was 0.76 ± 0.06 logMAR, improving to 0.38 ± 0.16 postoperatively, with a statistically significant difference (p = 0.001). AI software was utilized to assess biomarkers, such as intraretinal fluid (IRF) and subretinal fluid (SRF) volume, external limiting membrane (ELM) and ellipsoid zone (EZ) integrity, and retinal hyperreflective foci (HRF). The AI analysis showed a significant decrease in IRF volume, from 0.08 ± 0.12 mm3 preoperatively to 0.01 ± 0.01 mm3 postoperatively. ELM interruption improved from 79% ± 18% to 34% ± 37% after surgery (p = 0.006), whereas EZ interruption improved from 80% ± 22% to 40% ± 36% (p = 0.007) postoperatively. Additionally, the study revealed a negative correlation between preoperative IRF volume and postoperative BCVA recovery, suggesting that greater preoperative fluid volumes may hinder visual improvement. The integrity of the ELM and EZ was found to be essential for postoperative visual acuity improvement, with their disruption negatively impacting visual recovery. The study highlights the potential of AI in quantifying OCT biomarkers for managing MHs and improving patient care.

A feasibility study for objective evaluation of visual acuity based on pattern-reversal visual evoked potentials and other related visual parameters with machine learning algorithm

BACKGROUND

To develop machine learning models for objectively evaluating visual acuity (VA) based on pattern-reversal visual evoked potentials (PRVEPs) and other related visual parameters.

METHODS

Twenty-four volunteers were recruited and forty-eight eyes were divided into four groups of 1.0, 0.8, 0.6, and 0.4 (decimal vision). The relationship between VA, peak time, or amplitude of P100 recorded at 5.7°, 2.6°, 1°, 34', 15', and 7' check sizes were analyzed using repeated-measures analysis of variance. Correlations between VA and P100, contrast sensitivity (CS), refractive error, wavefront aberrations, and visual field were analyzed by rank correlation. Based on meaningful P100 peak time, P100 amplitude, and other related visual parameters, four machine learning algorithms and an ensemble classification algorithm were used to construct objective assessment models for VA. Receiver operating characteristic (ROC) curves were used to compare the efficacy of different models by repeated sampling comparisons and ten-fold cross-validation.

RESULTS

The main effects of P100 peak time and amplitude between different VA and check sizes were statistically significant (all P < 0.05). Except amplitude at 2.6° and 5.7°, VA was negatively correlated with peak time and positively correlated with amplitude. The peak time initially shortened with increasing check size and gradually lengthened after the minimum value was reached at 1°. At the 1° check size, there were statistically significant differences when comparing the peak times between the vision groups with each other (all P < 0.05), and the amplitudes of the vision reduction groups were significantly lower than that of the 1.0 vision group (all P < 0.01). The correlations between peak time, amplitude, and visual acuity were all highest at 1° (r_s = - 0.740, 0.438). VA positively correlated with CS and spherical equivalent (all P < 0.001). There was a negative correlation between VA and coma aberrations (P < 0.05). For different binarization classifications of VA, the classifier models with the best assessment efficacy all had the mean area under the ROC curves (AUC) above 0.95 for 500 replicate samples and above 0.84 for ten-fold cross-validation.

CONCLUSIONS

Machine learning models established by meaning visual parameters related to visual acuity can assist in the objective evaluation of VA.

The Importance of Anatomic Configuration and Cystic Changes in Macular Hole: Predicting Surgical Success with a Different Approach

PURPOSE

This study aimed to define a novel metric for the area of the macular hole (MH) and cysts located around the hole using an optical coherence tomography (OCT) device.

METHODS

This study was conducted with 58 eyes of 56 patients. The patients were divided into two groups according to anatomic closure after surgery. Using the metrics of macular hole index (MHI), tractional hole index (THI), hole forming factor (HFF), macular hole area (HA), the cystoid space areas in the inner retinal layers (CA), and our novel metric, the cyst hole area index (CHAI) was calculated. The correlation of the CA, the HA, and the CHAI with other indexes were assessed. Receiver operating characteristic (ROC) curves and cut-off values were derived for indexes predicting type 1 or type 2 closures.

RESULTS

The CA showed a strong positive correlation with the base MH size and the maximum MH height (r = 0.624, p < 0.001; r = 0.722, p < 0.001, respectively). The HA showed a strong positive correlation with basal MH size and minimum MH size (r = 0.934, p < 0.001; r = 0.765, p < 0.001). The HA showed a moderate positive correlation with maximum MH height (r = 0.483, p < 0.001, respectively). CHAI showed a moderate positive correlation with minimum MH size (r = 0.297, p = 0.02). CHAI and HA showed a moderate negative correlation with post-operative BCVA (r = -0.39, p = 0.003; r = -0.357, p = 0.006; respectively). ROC curve analysis showed that MHI (0.823), THI (0.750), and HFF (0.722) predicted type 1 closure and that CHAI (0.769) and HA (0.709) predicted type 2 closures.

CONCLUSION

MHI and our novel index CHAI, which can be calculated without any additional software, could successfully predict type 1 and type 2 closures, respectively.

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.