Neonatal Catheter and Tube Placement and Radiographic Assessment Statistics in Relation to Important Anatomic Landmarks.

OBJECTIVE: Neonatal catheters and tubes are commonly used for monitoring and support for intensive care and must be correctly positioned to avoid complications. Position assessment is routinely done by radiography. The objective of this study is to characterize neonatal catheter and tube placement in terms of the proportion of those devices that are malpositioned. STUDY DESIGN: Using an institutional dataset of 723 chest/abdominal radiographs of neonatal intensive care unit (ICU) patients (all within 60 days of birth), we assessed the proportion of catheters that are malpositioned. Many radiographs contained multiple catheter types. Umbilical venous catheters (UVCs; 448 radiographs), umbilical arterial catheters (UACs; 259 radiographs), endotracheal tubes (ETTs; 451 radiographs), and nasogastric tubes (NGTs; 603 radiographs) were included in our analysis. RESULTS: UVCs were malpositioned in 90% of radiographs, while UACs were malpositioned in 36%, ETTs in 30%, and NGTs in just 5%. The most common locations in which UVCs were malpositioned were in the right atrium (31%) and umbilical vein (21%), and for UACs the most common malpositioned tip location was the aortic arch (8%). For the remaining tubes, 5% of ETTs were found to be in the right main bronchus and 4% of NGTs were found in the esophagus. CONCLUSION: A substantial proportion of catheters and tubes are malpositioned, suggesting that optimizing methods of catheter placement and assessment ought to be areas of focus for future work. KEY POINTS: · Neonatal catheters are frequently malpositioned.. · Most umbilical venous catheters need readjustment.. · X-ray and ultrasound are important for assessment.. · Catheter tips should be assessed in all X-rays..

Learning temporal attention in dynamic graphs with bilinear interactions.

Reasoning about graphs evolving over time is a challenging concept in many domains, such as bioinformatics, physics, and social networks. We consider a common case in which edges can be short term interactions (e.g., messaging) or long term structural connections (e.g., friendship). In practice, long term edges are often specified by humans. Human-specified edges can be both expensive to produce and suboptimal for the downstream task. To alleviate these issues, we propose a model based on temporal point processes and variational autoencoders that learns to infer temporal attention between nodes by observing node communication. As temporal attention drives between-node feature propagation, using the dynamics of node interactions to learn this key component provides more flexibility while simultaneously avoiding issues associated with human-specified edges. We also propose a bilinear transformation layer for pairs of node features instead of concatenation, typically used in prior work, and demonstrate its superior performance in all cases. In experiments on two datasets in the dynamic link prediction task, our model often outperforms the baseline model that requires a human-specified graph. Moreover, our learned attention is semantically interpretable and infers connections similar to actual graphs.

Dynamic contact networks of swine movement in Manitoba, Canada: Characterization and implications for infectious disease spread.

We use swine shipping data from Manitoba to construct one-mode dynamic contact networks of swine locations and two-mode location-to-truck networks at four time scales: daily, weekly, monthly and for the entire two-year study period. We provide measures of graph evolution and graph characterization for each, useful in the development of statistical models related to infectious disease transmission. We find that Manitoba shipping practices differ from those in other Canadian regions, and particularly that truck sharing is more common in Manitoba than elsewhere in the country.

Deep learning for supervised classification of spatial epidemics.

In an emerging epidemic, public health officials must move quickly to contain the spread. Information obtained from statistical disease transmission models often informs the development of containment strategies. Inference procedures such as Bayesian Markov chain Monte Carlo allow researchers to estimate parameters of such models, but are computationally expensive. In this work, we explore supervised statistical and machine learning methods for fast inference via supervised classification, with a focus on deep learning. We apply our methods to simulated epidemics through two populations of swine farms in Iowa, and find that the random forest performs well on the denser population, but is outperformed by a deep learning model on the sparser population.

A hierarchical Bayesian model to predict APOE4 genotype and the age of Alzheimer's disease onset.

In this work we use a hierarchical Bayesian paradigm to introduce a theoretical framework to determine an individual's Apolipoprotein ε4 (APOE4) genotype, which heavily influences both the age of onset and probability of acquiring Alzheimer's disease (AD). This calculation is based solely on an individual's family history. This APOE4 genotype estimation is then combined with a number of known factors that influence AD onset to produce a function that estimates the onset of AD as a function of age. We disseminated our Alzheimer's predictive tool online at http://www.alzheimerspredictor.com.