Key Points Question Are deep learning techniques sufficiently accurate to classify presegmented phases in videos of cataract surgery for subsequent automated skill assessment and feedback? Findings In this cross-sectional study including videos from a convenience sample of 100 cataract procedures, modeling time series of labels of instruments in use appeared to yield greater accuracy in classifying phases of cataract operations than modeling cross-sectional data on instrument labels, spatial video image features, spatiotemporal video image features, or spatiotemporal video image features with appended instrument labels. Meaning Time series models of instruments in use may serve to automate the identification of phases in cataract surgery, helping to develop efficient and effective surgical skill training tools in ophthalmology.

IMPORTANCEAdherence to screening for vision-threatening proliferative sickle cell retinopathy is limited among patients with sickle cell hemoglobinopathy despite guidelines recommending dilated fundus examinations beginning in childhood. An automated algorithm for detecting sea fan neovascularization from ultra-widefield color fundus photographs could expand access to rapid retinal evaluations to identify patients at risk of vision loss from proliferative sickle cell retinopathy.OBJECTIVE To develop a deep learning system for detecting sea fan neovascularization from ultra-widefield color fundus photographs from patients with sickle cell hemoglobinopathy. DESIGN, SETTING, AND PARTICIPANTSIn a cross-sectional study conducted at a single-institution, tertiary academic referral center, deidentified, retrospectively collected, ultra-widefield color fundus photographs from 190 adults with sickle cell hemoglobinopathy were independently graded by 2 masked retinal specialists for presence or absence of sea fan neovascularization. A third masked retinal specialist regraded images with discordant or indeterminate grades. Consensus retinal specialist reference standard grades were used to train a convolutional neural network to classify images for presence or absence of sea fan neovascularization. Participants included nondiabetic adults with sickle cell hemoglobinopathy receiving care from a Wilmer Eye Institute retinal specialist; the patients had received no previous laser or surgical treatment for sickle cell retinopathy and underwent imaging with ultra-widefield color fundus photographs between January 1, 2012, and January 30, 2019.INTERVENTIONS Deidentified ultra-widefield color fundus photographs were retrospectively collected.MAIN OUTCOMES AND MEASURES Sensitivity, specificity, and area under the receiver operating characteristic curve of the convolutional neural network for sea fan detection.RESULTS A total of 1182 images from 190 patients were included. Of the 190 patients, 101 were women (53.2%), and the mean (SD) age at baseline was 36.2 (12.3) years; 119 patients (62.6%) had hemoglobin SS disease and 46 (24.2%) had hemoglobin SC disease. One hundred seventy-nine patients (94.2%) were of Black or African descent. Images with sea fan neovascularization were obtained in 57 patients (30.0%). The convolutional neural network had an area under the curve of 0.988 (95% CI, 0.969-0.999), with sensitivity of 97.4% (95% CI, 86.5%-99.9%) and specificity of 97.0% (95% CI, 93.5%-98.9%) for detecting sea fan neovascularization from ultra-widefield color fundus photographs.CONCLUSIONS AND RELEVANCE This study reports an automated system with high sensitivity and specificity for detecting sea fan neovascularization from ultra-widefield color fundus photographs from patients with sickle cell hemoglobinopathy, with potential applications for improving screening for vision-threatening proliferative sickle cell retinopathy.

No abstract

Objective: Computer-aided analysis of laryngoscopy images has potential to add objectivity to subjective evaluations. Automated classification of biomedical images is extremely challenging due to the precision required and the limited amount of annotated data available for training. Convolutional neural networks (CNNs) have the potential to improve image analysis and have demonstrated good performance in many settings. This study applied machine-learning technologies to laryngoscopy to determine the accuracy of computer recognition of known laryngeal lesions found in patients post-extubation. Methods: This is a proof of concept study that used a convenience sample of transnasal, flexible, distal-chip laryngoscopy images from patients post-extubation in the intensive care unit. After manually annotating images at the pixel-level, we applied a CNN-based method for analysis of granulomas and ulcerations to test potential machine-learning approaches for laryngoscopy analysis. Results: A total of 127 images from 25 patients were manually annotated for presence and shape of these lesions—100 for training, 27 for evaluating the system. There were 193 ulcerations (148 in the training set; 45 in the evaluation set) and 272 granulomas (208 in the training set; 64 in the evaluation set) identified. Time to annotate each image was approximately 3 minutes. Machine-based analysis demonstrated per-pixel sensitivity of 82.0% and 62.8% for granulomas and ulcerations respectively; specificity was 99.0% and 99.6%. Conclusion: This work demonstrates the feasibility of machine learning via CNN-based methods to add objectivity to laryngoscopy analysis, suggesting that CNN may aid in laryngoscopy analysis for other conditions in the future.

SummaryBackgroundInfectious disease modeling can serve as a powerful tool for science-based management of outbreaks, providing situational awareness and decision support for policy makers. Predictive modeling of an emerging disease is challenging due to limited knowledge on its epidemiological characteristics. For COVID-19, the prediction difficulty was further compounded by continuously changing policies, varying behavioral responses, poor availability and quality of crucial datasets, and the variable influence of different factors as the pandemic progresses. Due to these challenges, predictive modeling for COVID-19 has earned a mixed track record.MethodsWe provide a systematic review of prospective, data-driven modeling studies on population-level dynamics of COVID-19 in the US and conduct a quantitative assessment on crucial elements of modeling, with a focus on the aspects of modeling that are critical to make them useful for decision-makers. For each study, we documented the forecasting window, methodology, prediction target, datasets used, geographic resolution, whether they expressed quantitative uncertainty, the type of performance evaluation, and stated limitations. We present statistics for each category and discuss their distribution across the set of studies considered. We also address differences in these model features based on fields of study.FindingsOur initial search yielded 2,420 papers, of which 119 published papers and 17 preprints were included after screening. The most common datasets relied upon for COVID-19 modeling were counts of cases (93%) and deaths (62%), followed by mobility (26%), demographics (25%), hospitalizations (12%), and policy (12%). Our set of papers contained a roughly equal number of short-term (46%) and long-term (60%) predictions (defined as a prediction horizon longer than 4 weeks) and statistical (43%) versus compartmental (47%) methodologies. The target variables used were predominantly cases (89%), deaths (52%), hospitalizations (10%), and Rt (9%). We found that half of the papers in our analysis did not express quantitative uncertainty (50%). Among short-term prediction models, which can be fairly evaluated against truth data, 25% did not conduct any performance evaluation, and most papers were not evaluated over a timespan that includes varying epidemiological dynamics. The main categories of limitations stated by authors were disregarded factors (39%), data quality (28%), unknowable factors (26%), limitations specific to the methods used (22%), data availability (16%), and limited generalizability (8%). 36% of papers did not list any limitations in their discussion or conclusion section.InterpretationPublished COVID-19 models were found to be consistently lacking in some of the most important elements required for usability and translation, namely transparency, expressing uncertainty, performance evaluation, stating limitations, and communicating appropriate interpretations. Adopting the EPIFORGE 2020 guidelines would address these shortcomings and improve the consistency, reproducibility, comparability, and quality of epidemic forecasting reporting. We also discovered that most of the operational models that have been used in real-time to inform decision-making have not yet made it into the published literature, which highlights that the current publication system is not suited to the rapid information-sharing needs of outbreaks. Furthermore, data quality was identified to be one of the most important drivers of model performance, and a consistent limitation noted by the modeling community. The US public health infrastructure was not equipped to provide timely, high-quality COVID-19 data, which is required for effective modeling. Thus, a systematic infrastructure for improved data collection and sharing should be a major area of investment to support future pandemic preparedness.

The alignment strength of a graph matching is a quantity that gives the practitioner a measure of the correlation of the two graphs, and it can also give the practitioner a sense for whether the graph matching algorithm found the true matching. Unfortunately, when a graph matching algorithm fails to find the truth because of weak signal, there may be “phantom alignment strength” from meaningless matchings that, by random noise, have fewer disagreements than average (sometimes substantially fewer); this alignment strength may give the misleading appearance of significance. A practitioner needs to know what level of alignment strength may be phantom alignment strength and what level indicates that the graph matching algorithm obtained the true matching and is a meaningful measure of the graph correlation. The Phantom Alignment Strength Conjecture introduced here provides a principled and practical means to approach this issue. We provide empirical evidence for the conjecture, and explore its consequences.

No abstract

We introduce the Weighted Contextual Interval Score (WCIS), a new method for evaluating the performance of short-term interval-form forecasts. The WCIS provides a pragmatic utility-based characterization of probabilistic predictions, developed in response to the challenge of evaluating forecast performances in the turbulent context of the COVID-19 pandemic. Current widely-used scoring techniques generally fall into two groups: those that generate an individually interpretable metric, and those that generate a comparable and aggregable metric. The WCIS harmonizes these attributes, resulting in a normalized score that is nevertheless intuitively representative of the in-situ quality of individual forecasts. This method is expressly intended to enable practitioners and policy-makers who may not have expertise in forecasting but are nevertheless essential partners in epidemic response to use and provide insightful analysis of predictions. In this paper, we detail the methodology of the WCIS and demonstrate its utility in the context of US state-level COVID-19 predictions.