Multivariate time series data in practical applications, such as health care, geoscience, and biology, are characterized by a variety of missing values. In time series prediction and other related tasks, it has been noted that missing values and their missing patterns are often correlated with the target labels, a.k.a., informative missingness. There is very limited work on exploiting the missing patterns for effective imputation and improving prediction performance. In this paper, we develop novel deep learning models, namely GRU-D, as one of the early attempts. GRU-D is based on Gated Recurrent Unit (GRU), a state-of-the-art recurrent neural network. It takes two representations of missing patterns, i.e., masking and time interval, and effectively incorporates them into a deep model architecture so that it not only captures the long-term temporal dependencies in time series, but also utilizes the missing patterns to achieve better prediction results. Experiments of time series classification tasks on real-world clinical datasets (MIMIC-III, PhysioNet) and synthetic datasets demonstrate that our models achieve state-of-the-art performance and provide useful insights for better understanding and utilization of missing values in time series analysis.

Iron oxide nanoparticles (IONPs) have been increasingly used in biomedical applications, but the comprehensive understanding of their interactions with biological systems is relatively limited. In this study, we systematically investigated the in vitro cell uptake, cytotoxicity, in vivo distribution, clearance and toxicity of commercially available and well-characterized IONPs with different sizes and coatings. Polyethylenimine (PEI)-coated IONPs exhibited significantly higher uptake than PEGylated ones in both macrophages and cancer cells, and caused severe cytotoxicity through multiple mechanisms such as ROS production and apoptosis. 10 nm PEGylated IONPs showed higher cellular uptake than 30 nm ones, and were slightly cytotoxic only at high concentrations. Interestingly, PEGylated IONPs but not PEI-coated IONPs were able to induce autophagy, which may play a protective role against the cytotoxicity of IONPs. Biodistribution studies demonstrated that all the IONPs tended to distribute in the liver and spleen, and the biodegradation and clearance of PEGylated IONPs in these tissues were relatively slow (>2 weeks). Among them, 10 nm PEGylated IONPs achieved the highest tumor uptake. No obvious toxicity was found for PEGylated IONPs in BALB/c mice, whereas PEI-coated IONPs exhibited dose-dependent lethal toxicity. Therefore, it is crucial to consider the size and coating properties of IONPs in their applications.

Abstract-We previously showed that hydrogen peroxide (H 2 O 2 ) contributes to flow-induced dilation in human coronary resistance arteries (HCRAs); however, the source of this H 2 O 2 is not known. We hypothesized that the H 2 O 2 is derived from superoxide (O 2 •Ϫ ) generated by mitochondrial respiration. HCRAs were dissected from right atrial appendages obtained from patients during cardiac surgery and cannulated with micropipettes. H 2 O 2 -derived radicals and O 2•Ϫ were detected by electron spin resonance (ESR) using BMPO as the spin trap and by histofluorescence using hydroethidine (HE, 5 mol/L) and dichlorodihydrofluorescein (DCFH, 5 mol/L). Diameter changes to increases in pressure gradients (20 and 100 cm H 2 O) were examined in the absence and the presence of rotenone (1 mol/L), myxothiazol (100 nmol/L), cyanide (1 mol/L), mitochondrial complex I, III, and IV inhibitors, respectively, and apocynin (3 mmol/L), a NADPH oxidase inhibitor. At a pressure gradient of 100 cm H 2 O, ubisemiquinone and hydroxyl radicals were detected from effluents of vessels. Including superoxide dismutase and catalase in the perfusate reduced the ESR signals. Relative ethidium and DCFH fluorescence intensities in HCRAs exposed to flow were enhanced (1.45Ϯ0.15 and 1.57Ϯ0.12, respectively compared with no-flow) and were inhibited by rotenone (0.87Ϯ0.17 and 0.95Ϯ0.07). Videomicroscopic studies showed that rotenone and myxothiazol blocked flow-induced dilation (% max. dilation at 100 cm H 2 O: rotenone, 74Ϯ3% versus 3Ϯ13%; myxothiazol, 67Ϯ3% versus 28Ϯ4%; PϽ0.05). Neither cyanide nor apocynin altered flow-induced dilation. These results suggest that shear stress induced H 2 O 2 formation, and flow-induced dilation is derived from O 2•Ϫ originating from mitochondrial respiration.

Deep learning models (aka Deep Neural Networks) have revolutionized many fields including computer vision, natural language processing, speech recognition, and is being increasingly used in clinical healthcare applications. However, few works exist which have benchmarked the performance of the deep learning models with respect to the state-of-the-art machine learning models and prognostic scoring systems on publicly available healthcare datasets. In this paper, we present the benchmarking results for several clinical prediction tasks such as mortality prediction, length of stay prediction, and ICD-9 code group prediction using Deep Learning models, ensemble of machine learning models (Super Learner algorithm), SAPS II and SOFA scores. We used the Medical Information Mart for Intensive Care III (MIMIC-III) (v1.4) publicly available dataset, which includes all patients admitted to an ICU at the Beth Israel Deaconess Medical Center from 2001 to 2012, for the benchmarking tasks. Our results show that deep learning models consistently outperform all the other approaches especially when the 'raw' clinical time series data is used as input features to the models.