The study aimed to develop machine learning models that have strong prediction power and interpretability for diagnosis of glaucoma based on retinal nerve fiber layer (RNFL) thickness and visual field (VF). We collected various candidate features from the examination of retinal nerve fiber layer (RNFL) thickness and visual field (VF). We also developed synthesized features from original features. We then selected the best features proper for classification (diagnosis) through feature evaluation. We used 100 cases of data as a test dataset and 399 cases of data as a training and validation dataset. To develop the glaucoma prediction model, we considered four machine learning algorithms: C5.0, random forest (RF), support vector machine (SVM), and k-nearest neighbor (KNN). We repeatedly composed a learning model using the training dataset and evaluated it by using the validation dataset. Finally, we got the best learning model that produces the highest validation accuracy. We analyzed quality of the models using several measures. The random forest model shows best performance and C5.0, SVM, and KNN models show similar accuracy. In the random forest model, the classification accuracy is 0.98, sensitivity is 0.983, specificity is 0.975, and AUC is 0.979. The developed prediction models show high accuracy, sensitivity, specificity, and AUC in classifying among glaucoma and healthy eyes. It will be used for predicting glaucoma against unknown examination records. Clinicians may reference the prediction results and be able to make better decisions. We may combine multiple learning models to increase prediction accuracy. The C5.0 model includes decision rules for prediction. It can be used to explain the reasons for specific predictions.

No abstract

The aim is to develop a machine learning prediction model for the diagnosis of glaucoma and an explanation system for a specific prediction. Clinical data of the patients based on a visual field test, a retinal nerve fiber layer optical coherence tomography (RNFL OCT) test, a general examination including an intraocular pressure (IOP) measurement, and fundus photography were provided for the feature selection process. Five selected features (variables) were used to develop a machine learning prediction model. The support vector machine, C5.0, random forest, and XGboost algorithms were tested for the prediction model. The performance of the prediction models was tested with 10-fold cross-validation. Statistical charts, such as gauge, radar, and Shapley Additive Explanations (SHAP), were used to explain the prediction case. All four models achieved similarly high diagnostic performance, with accuracy values ranging from 0.903 to 0.947. The XGboost model is the best model with an accuracy of 0.947, sensitivity of 0.941, specificity of 0.950, and AUC of 0.945. Three statistical charts were established to explain the prediction based on the characteristics of the XGboost model. Higher diagnostic performance was achieved with the XGboost model. These three statistical charts can help us understand why the machine learning model produces a specific prediction result. This may be the first attempt to apply “explainable artificial intelligence” to eye disease diagnosis.

Age, race, ocular surgery (cataract and refractive), and DM seem to significantly affect donor corneal ECD. Of these variables, age, a history of cataract surgery, and DM were found to be the greatest risk factors for inadequate donor cell density (less than 2000/mm).

The corneal sensitivity and ocular surface changes were significant in the subclinical KC and KC eyes compared with the control subjects. Ocular surface disease in KC was characterized by tear deficiency disorder and abnormal impression cytology results. However, no significant difference in tear osmolarity was found among the groups. The decrease in corneal sensitivity and ocular surface change may be associated with the pathogenesis of ocular surface changes in KC and the progression of the disease.

Purpose We sought to evaluate the distribution and characteristics of meibomian gland dysfunction (MGD) and the treatment patterns for symptomatic MGD patients in South Korea. Methods One hundred ninety-six right eyes of 196 MGD patients were enrolled. For each patient, meibum expressibility in the central eight glands in both the upper and lower eyelids was examined. Each upper and lower eyelid was separately classified into one of the following three subtypes: nonobvious obstructive (low-delivery without lid margin abnormality), obvious obstructive (low-delivery with lid margin abnormality), and hypersecretory (high-delivery with lid margin abnormality). All treatment plans were also recorded. Results The mean number of expressible glands of the central eight glands in the upper eyelids (3.9 ± 2.6) was significantly higher than that in the lower eyelids (2.2 ± 2.4, p < 0.001). Obvious obstructive MGD was the most common subtype, followed by the hypersecretory and nonobvious obstructive subtypes in both the upper and lower eyelids. Of the 196 subjects, 38 (19.4%) had upper and lower eyelids that were assigned to different categories. Eyelid hygiene was the most prescribed treatment (74.5%), followed by lubricant eye drop usage (71.5%). Physicians tended to determine treatment plans based on the subtype of the upper eyelid rather than that of the lower eyelid. Conclusions The majority of subjects were classified as having the obvious obstructive subtype of MGD, and 19.4% had upper and lower eyelids that were different subtypes. Eyelid hygiene was the most prescribed treatment for MGD patients, and treatment patterns were mostly determined based on the subtype of the upper eyelids.

PurposeTo evaluate choroidal structural changes in glaucoma using choroidal vascularity index (CVI) compared to healthy subjects.MethodsThis retrospective study included 56 patients with open angle glaucoma (OAG), 50 patients with preperimetric glaucoma (PPG) and 50 age-matched healthy eyes. Choroidal images were binarized into luminal area (LA) and stromal area. CVI was defined as the ratio of LA to total circumscribed choroid area (TCA). Mean choroidal thickness (CT) and mean CVI between glaucoma patients and healthy subjects were compared.ResultsOAG and PPG eyes showed smaller LA (0.45 ± 0.13 ㎟ vs. 0.47 ± 0.11 ㎟, p = 0.04). In multivariate regression analysis, CVI of both OAG (64.34±0.19%, p = 0.001) and PPG (65.37±0.15%, p = 0.001) were significantly lower than healthy eyes (68.81±0.14%).ConclusionEyes with glaucoma demonstrated reduced CVI compared with healthy eyes. CVI may be a potential noninvasive tool for studying vascular dysfunction in glaucoma.

PurposeTo describe surgical outcomes for transconjunctival anterior orbitotomy for intraconal cavernous hemangiomas.MethodsThe medical records of 9 consecutive patients with intraconal cavernous hemangiomas who underwent surgical removal by transconjunctival anterior orbitotomy were retrospectively reviewed. The conjunctiva was incised and retracted with a traction suture. For large tumors, a rectus muscle was temporarily disinserted. Tenon's capsule was separated and the tumor was removed with a cryoprobe or clamp. Surgical outcomes, positions of the tumors, methods of approach, and intra- and post-operative complications were evaluated.ResultsThe mean follow-up period was 33 ± 6.8 months. No bony orbitotomy was used in this technique and the cosmetic results were very satisfactory. All tumors were removed intact. In 4 patients, tumors were extirpated with the aid of a cryoprobe. No patients had residual proptosis or limitation of ocular movement. No signs of recurrence were noted in any cases at 33 months follow-up. No serious or permanent complications were observed during or after the operation.ConclusionsTransconjunctival anterior orbitotomy is an important surgical procedure in the treatment of intraconal cavernous hemangiomas. It can produce an excellent result, even if the posterior border of the tumor abuts the orbital apex.