RESUMEN
BACKGROUND/AIMS: To design a deep learning (DL) model for the detection of glaucoma progression with a longitudinal series of macular optical coherence tomography angiography (OCTA) images. METHODS: 202 eyes of 134 patients with open-angle glaucoma with ≥4 OCTA visits were followed for an average of 3.5 years. Glaucoma progression was defined as having a statistically significant negative 24-2 visual field (VF) mean deviation (MD) rate. The baseline and final macular OCTA images were aligned according to centre of fovea avascular zone automatically, by checking the highest value of correlation between the two images. A customised convolutional neural network (CNN) was designed for classification. A comparison of the CNN to logistic regression model for whole image vessel density (wiVD) loss on detection of glaucoma progression was performed. The performance of the model was defined based on the confusion matrix of the validation dataset and the area under receiver operating characteristics (AUC). RESULTS: The average (95% CI) baseline VF MD was -3.4 (-4.1 to -2.7) dB. 28 (14%) eyes demonstrated glaucoma progression. The AUC (95% CI) of the DL model for the detection of glaucoma progression was 0.81 (0.59 to 0.93). The sensitivity, specificity and accuracy (95% CI) of DL model were 67% (34% to 78%), 83% (42% to 97%) and 80% (52% to 95%), respectively. The AUC (95% CI) for the detection of glaucoma progression based on the logistic regression model was lower than the DL model (0.69 (0.50 to 0.88)). CONCLUSION: The optimised DL model detected glaucoma progression based on longitudinal macular OCTA images showed good performance. With external validation, it could enhance detection of glaucoma progression. TRIAL REGISTRATION NUMBER: NCT00221897.
RESUMEN
PURPOSE: To use longitudinal optical coherence tomography (OCT) and OCT angiography (OCTA) data to detect glaucomatous visual field (VF) progression with a supervised machine learning approach. DESIGN: Prospective cohort study. METHODS: One hundred ten eyes of patients with suspected glaucoma (33.6%) and patients with glaucoma (66.4%) with a minimum of 5 24-2 VF tests and 3 optic nerve head and macula images over an average follow-up duration of 4.1 years were included. VF progression was defined using a composite measure including either a "likely progression event" on Guided Progression Analysis, a statistically significant negative slope of VF mean deviation or VF index, or a positive pointwise linear regression event. Feature-based gradient boosting classifiers were developed using different subsets of baseline and longitudinal OCT and OCTA summary parameters. The area under the receiver operating characteristic curve (AUROC) was used to compare the classification performance of different models. RESULTS: VF progression was detected in 28 eyes (25.5%). The model with combined baseline and longitudinal OCT and OCTA parameters at the global and hemifield levels had the best classification accuracy to detect VF progression (AUROC = 0.89). Models including combined OCT and OCTA parameters had higher classification accuracy compared with those with individual subsets of OCT or OCTA features alone. Including hemifield measurements significantly improved the models' classification accuracy compared with using global measurements alone. Including longitudinal rates of change of OCT and OCTA parameters (AUROCs = 0.80-0.89) considerably increased the classification accuracy of the models with baseline measurements alone (AUROCs = 0.60-0.63). CONCLUSIONS: Longitudinal OCTA measurements complement OCT-derived structural metrics for the evaluation of functional VF loss in patients with glaucoma.
Asunto(s)
Glaucoma , Campos Visuales , Humanos , Tomografía de Coherencia Óptica/métodos , Estudios Prospectivos , Presión Intraocular , Glaucoma/diagnóstico , Pruebas del Campo Visual , Angiografía con Fluoresceína/métodosRESUMEN
Wastewater-based surveillance has gained prominence and come to the forefront as a leading indicator of forecasting COVID-19 (coronavirus disease 2019) infection dynamics owing to its cost-effectiveness and its ability to inform early public health interventions. A university campus could especially benefit from wastewater surveillance, as universities are characterized by largely asymptomatic populations and are potential hot spots for transmission that necessitate frequent diagnostic testing. In this study, we employed a large-scale GIS (geographic information systems)-enabled building-level wastewater monitoring system associated with the on-campus residences of 7,614 individuals. Sixty-eight automated wastewater samplers were deployed to monitor 239 campus buildings with a focus on residential buildings. Time-weighted composite samples were collected on a daily basis and analyzed on the same day. Sample processing was streamlined significantly through automation, reducing the turnaround time by 20-fold and exceeding the scale of similar surveillance programs by 10- to 100-fold, thereby overcoming one of the biggest bottlenecks in wastewater surveillance. An automated wastewater notification system was developed to alert residents to a positive wastewater sample associated with their residence and to encourage uptake of campus-provided asymptomatic testing at no charge. This system, integrated with the rest of the "Return to Learn" program at the University of California (UC) San Diego-led to the early diagnosis of nearly 85% of all COVID-19 cases on campus. COVID-19 testing rates increased by 1.9 to 13× following wastewater notifications. Our study shows the potential for a robust, efficient wastewater surveillance system to greatly reduce infection risk as college campuses and other high-risk environments reopen. IMPORTANCE Wastewater-based epidemiology can be particularly valuable at university campuses where high-resolution spatial sampling in a well-controlled context could not only provide insight into what affects campus community as well as how those inferences can be extended to a broader city/county context. In the present study, a large-scale wastewater surveillance was successfully implemented on a large university campus enabling early detection of 85% of COVID-19 cases thereby averting potential outbreaks. The highly automated sample processing to reporting system enabled dramatic reduction in the turnaround time to 5 h (sample to result time) for 96 samples. Furthermore, miniaturization of the sample processing pipeline brought down the processing cost significantly ($13/sample). Taken together, these results show that such a system could greatly ameliorate long-term surveillance on such communities as they look to reopen.
RESUMEN
Large-scale wastewater surveillance has the ability to greatly augment the tracking of infection dynamics especially in communities where the prevalence rates far exceed the testing capacity. However, current methods for viral detection in wastewater are severely lacking in terms of scaling up for high throughput. In the present study, we employed an automated magnetic-bead-based concentration approach for viral detection in sewage that can effectively be scaled up for processing 24 samples in a single 40-min run. The method compared favorably to conventionally used methods for viral wastewater concentrations with higher recovery efficiencies from input sample volumes as low as 10 ml and can enable the processing of over 100 wastewater samples in a day. The sensitivity of the high-throughput protocol was shown to detect 1 asymptomatic individual in a building of 415 residents. Using the high-throughput pipeline, samples from the influent stream of the primary wastewater treatment plant of San Diego County (serving 2.3 million residents) were processed for a period of 13 weeks. Wastewater estimates of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) viral genome copies in raw untreated wastewater correlated strongly with clinically reported cases by the county, and when used alongside past reported case numbers and temporal information in an autoregressive integrated moving average (ARIMA) model enabled prediction of new reported cases up to 3 weeks in advance. Taken together, the results show that the high-throughput surveillance could greatly ameliorate comprehensive community prevalence assessments by providing robust, rapid estimates.IMPORTANCE Wastewater monitoring has a lot of potential for revealing coronavirus disease 2019 (COVID-19) outbreaks before they happen because the virus is found in the wastewater before people have clinical symptoms. However, application of wastewater-based surveillance has been limited by long processing times specifically at the concentration step. Here we introduce a much faster method of processing the samples and show its robustness by demonstrating direct comparisons with existing methods and showing that we can predict cases in San Diego by a week with excellent accuracy, and 3 weeks with fair accuracy, using city sewage. The automated viral concentration method will greatly alleviate the major bottleneck in wastewater processing by reducing the turnaround time during epidemics.
RESUMEN
Optical coherence tomography (OCT) is a non-invasive optical imaging modality capable of high resolution imaging of internal tissue structures. It is widely believed that the high axial resolution in OCT systems requires a wide-bandwidth light source. As a result, often the potential advantages of narrow-bandwidth sources (in terms of cost and/or imaging speed) are understood to come at the cost of significant reduction in imaging resolution. In this paper, we argue that this trade-off between resolution and speed is a shortcoming imposed by the-state-of-the-art A-scan reconstruction algorithm, Fast Fourier Transform, and can be circumvented through use of alternative processing methods. In particular, we investigate the shortcomings of the FFT as well as previously proposed alternatives and demonstrate the first application of an iterative regularized re-weighted l(2) norm method to improve the axial resolution of fast scan rate OCT systems in the narrow-bandwidth imaging conditions. We validate our claims via experimental results generated from a home-built OCT system used to image layered phantom and in vivo data. Our results rely on new, sophisticated signal processing algorithms to generate higher precision (i.e., higher resolution) OCT images at correspondingly fast scan rates. In other words, our work demonstrates the feasibility of simultaneously more reliable and more comfortable medical imaging systems for patients by reducing the overall scan time, without sacrificing image quality.