PURPOSE: To determine if smartphone photography could be a useful adjunct to blindness prevalence surveys by providing an accurate diagnosis of corneal opacity. METHODS: A total of 174 patients with infectious keratitis who had undergone corneal culturing over the past 5 years were enrolled in a diagnostic accuracy study at an eye hospital in South India. Both eyes had an ophthalmologist-performed slit lamp examination, followed by anterior segment photography with a handheld digital single lens reflex (SLR) camera and a smartphone camera coupled to an external attachment that provided magnification and illumination. The diagnostic accuracy of photography was assessed relative to slit lamp examination. RESULTS: In total, 90 of 174 enrolled participants had a corneal opacity in the cultured eye and no opacity in the contralateral eye, and did not have a penetrating keratoplasty or missing photographs. Relative to slit lamp examination, the sensitivity of corneal opacity diagnosis was 68% (95%CI 58-77%) using the smartphones default settings and 59% (95%CI 49-69%) using the SLR, and the specificity was 97% (95%CI 93-100%) for the smartphone and 97% (95%CI 92-100%) for the SLR. The sensitivity of smartphone-based corneal opacity diagnosis was higher for larger scars (81% for opacities 2 mm in diameter or larger), more visually significant scars (100% for eyes with visual acuity worse than 20/400), and more recent scars (85% for eyes cultured in the past 12 months). CONCLUSION: The diagnostic performance of a smartphone coupled to an external attachment, while somewhat variable, demonstrated high specificity and high sensitivity for all but the smallest opacities.

PURPOSE: Automated machine learning (AutoML) has emerged as a novel tool for medical professionals lacking coding experience, enabling them to develop predictive models for treatment outcomes. This study evaluated the performance of AutoML tools in developing models predicting the success of pneumatic retinopexy (PR) in treatment of rhegmatogenous retinal detachment (RRD). These models were then compared with custom models created by machine learning (ML) experts. DESIGN: Retrospective multicenter study. PARTICIPANTS: Five hundred and thirty nine consecutive patients with primary RRD that underwent PR by a vitreoretinal fellow at 6 training hospitals between 2002 and 2022. METHODS: We used 2 AutoML platforms: MATLAB Classification Learner and Google Cloud AutoML. Additional models were developed by computer scientists. We included patient demographics and baseline characteristics, including lens and macula status, RRD size, number and location of breaks, presence of vitreous hemorrhage and lattice degeneration, and physicians experience. The dataset was split into a training (n = 483) and test set (n = 56). The training set, with a 2:1 success-to-failure ratio, was used to train the MATLAB models. Because Google Cloud AutoML requires a minimum of 1000 samples, the training set was tripled to create a new set with 1449 datapoints. Additionally, balanced datasets with a 1:1 success-to-failure ratio were created using Python. MAIN OUTCOME MEASURES: Single-procedure anatomic success rate, as predicted by the ML models. F2 scores and area under the receiver operating curve (AUROC) were used as primary metrics to compare models. RESULTS: The best performing AutoML model (F2 score: 0.85; AUROC: 0.90; MATLAB), showed comparable performance to the custom model (0.92, 0.86) when trained on the balanced datasets. However, training the AutoML model with imbalanced data yielded misleadingly high AUROC (0.81) despite low F2-score (0.2) and sensitivity (0.17). CONCLUSIONS: We demonstrated the feasibility of using AutoML as an accessible tool for medical professionals to develop models from clinical data. Such models can ultimately aid in the clinical decision-making, contributing to better patient outcomes. However, outcomes can be misleading or unreliable if used naively. Limitations exist, particularly if datasets contain missing variables or are highly imbalanced. Proper model selection and data preprocessing can improve the reliability of AutoML tools. FINANCIAL DISCLOSURES: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Objective

To evaluate the outcomes of patients undergoing pneumatic retinopexy (PR) performed by vitreoretinal fellows at 6 academic centers in the United States.

Design

Retrospective, multicenter, consecutive case series.

Participants

483 patients with primary retinal detachments who underwent PR by 49 vitreoretinal fellows from 6 U.S. training sites between 2002 and 2016.

Methods

We reviewed medical records of patients and recorded baseline clinical characteristics (age, sex, baseline visual acuity, lens status, presence of lattice degeneration, presence of vitreous hemorrhage, location of retinal breaks, macular status, and size of detachment), visual and anatomic outcomes at 3-months after PR, as well as training level and PR experience of the fellow at the time of the procedure.

Main outcome measures

Single-procedure anatomic success and visual acuity at 3-months follow-up, and association with clinical and training-related factors.

Results

Vitreoretinal fellows performed a variable number of PR, with a median of 7 cases per fellow (range 1-24). Single-procedure anatomic success was 66.8%, and mean LogMAR visual outcome was 0.43 (Snellen equivalent 20/54) at 3 months. Factors that were independently associated with single-procedure success include phakic lens status (P = 0.01), smaller size of retinal detachment (P = 0.02), and the fellow's procedure experience (P = 0.01). The only factor associated with worse visual outcome was baseline visual acuity (P < 0.001).

Conclusion

Vitreoretinal fellows perform variably few pneumatic retinopexies but have outcomes comparable to reported rates by experienced specialists. Procedure experience of individual fellows may impact anatomic outcomes.