Scalable Moment Propagation and Analysis of Variational Distributions for Practical Bayesian Deep Learning.

Bayesian deep learning is one of the key frameworks employed in handling predictive uncertainty. Variational inference (VI), an extensively used inference method, derives the predictive distributions by Monte Carlo (MC) sampling. The drawback of MC sampling is its extremely high computational cost compared to that of ordinary deep learning. In contrast, the moment propagation (MP)-based approach propagates the output moments of each layer to derive predictive distributions instead of MC sampling. Because of this computational property, it is expected to realize faster inference than MC-based approaches. However, the applicability of the MP-based method in deep models has not been explored sufficiently, even though some studies have demonstrated the effectiveness of MP only in small toy models. One of the reasons is that it is difficult to train deep models by MP because of the large variance in activations. To realize MP in deep models, some normalization layers are required but have not yet been studied. In addition, it is still difficult to design well-calibrated MP-based models, because the effectiveness of MP-based methods under various variational distributions has also not been investigated. In this study, we propose a fast and reliable MP-based Bayesian deep-learning method. First, to train deep-learning models using MP, we introduce a batch normalization layer extended to random variables to prevent increases in the variance of activations. Second, to identify the appropriate variational distribution in MP, we investigate the treatment of moments of several variational distributions and evaluate their uncertainty quality of predictions. Experiments with regression tasks demonstrate that the MP-based method provides qualitatively and quantitatively equivalent predictive performance to MC-based methods regardless of variational distributions. In the classification tasks, we show that we can train MP-based deep models by extended batch normalization. We also show that the MP-based approach realizes 2.0-2.8 times faster inference than the MC-based approach while maintaining the predictive performance. The results of this study can help realize a fast and well-calibrated uncertainty estimation method that can be deployed in a wider range of reliability-aware applications.

Assessing Autonomic Nervous System Function by Pulse Rate Variability in Patients with Chronic Obstructive Pulmonary Disease.

Heart rate variability (HRV) is measured to analyze autonomic nervous system function in humans, and pulse rate variability (PRV) assessed using the photoplethysmography method with a pulse oximeter has been proposed as a surrogate for HRV. To examine whether PRV is compatible with HRV in patients with chronic obstructive pulmonary disease (COPD), we simultaneously measured HRV with an electrocardiogram and PRV with a pulse oximeter in patients with COPD, and compared low-frequency and high-frequency components computed from HRV and PRV as indicators of autonomic nervous system function. In a Bland-Altman analysis, the low-frequency component computed from HRV exhibited good consistency with that computed from PRV. The high-frequency component showed a significant fixed error but relatively good consistency. Our results indicate that autonomic nervous system function may be estimated with the low-frequency component by measuring PRV with a pulse oximeter in patients with COPD.

Specialized Pro-Resolving Mediators Do Not Inhibit the Synthesis of Inflammatory Mediators Induced by Tumor Necrosis Factor-α in Synovial Fibroblasts.

Background: Tumor necrosis factor (TNF)-α, a proinflammatory cytokine, is involved in the pathogenesis of rheumatoid arthritis (RA). The omega-3 unsaturated fatty acid-derived metabolites resolvin (Rv) D1, RvE1, and maresin-1 (MaR1) have been reported as anti-inflammatory lipid mediators and are known as specialized pro-resolving mediators (SPMs). In this study, we aimed to investigate the anti-inflammatory effects of SPMs on TNF-α-induced responses in synovial fibroblasts. Methods: We investigated the effects of SPMs on gene expression and/or production of cyclooxygenase-2 (COX-2), microsomal prostaglandin E synthase-1 (mPGES-1), interleukin (IL)-6, and matrix metalloproteinase (MMP)-3, which are involved in TNF-α-induced synovitis in RA or OA synovial fibroblasts, by quantitative real-time PCR. We also investigated the effects of SPMs on the mitogen-activated protein kinase (MAPK) signaling pathway by western blotting. Anti-inflammatory effects of SPMs were evaluated by applying SPMs to cultured synovial fibroblasts, followed by TNF-α stimulation. Results: The induction of COX-2, mPGES-1, IL-6, and MMP-3 by TNF-α in synovial fibroblasts was not suppressed by omega 3-derived SPMs regardless of their origin such as RA or OA. SPMs had no effect on lipid mediator receptor gene expression induce by TNF-α and did not inhibit the TNF-α-activated MAPK signaling pathway. The production of COX-2 and IL-6 protein was significantly decreased by p38 inhibitor. Conclusion: Despite reports on the anti-inflammatory effect of omega 3-derived SPMs, its anti-inflammatory effect on TNF-α-induced responses was not observed in synovial fibroblasts. The reason may be that SPMs have no suppressive effect on p38 activation, which plays an important role in the production of inflammatory cytokines in synovial fibroblasts.

Health-Related Quality of Life in Patients on Home Oxygen Therapy with Telemonitoring.

Home oxygen therapy (HOT) is an important treatment for patients with chronic respiratory diseases. Recently, telemonitoring of HOT has been become available. In the present study, we examined whether telemonitoring of HOT could improve health-related quality of life (HRQOL). Twelve patients receiving HOT participated in this study. The oxygen flow rates, use of the oxygen concentrator, and the values of percutaneous arterial oxygen saturation measured by each patient with a pulse oximeter were checked using a telemonitoring system for a period of one month. Interventions based on the results obtained were carried out in order to optimize oxygen use in this patient cohort. We evaluated the results of the SF-36 questionnaire before the initiation of telemonitoring and at 3 months after completion of the study. We identified significant improvements in SF-36 sub-scores after completion of this intervention. We conclude that telemonitoring may be a useful method to improve HRQOL.

Functional characteristics of a renal H+/lipophilic cation antiport system in porcine LLC-PK1 cells and rats.

We have recently found an H+/quinidine (a lipophilic cation, QND) antiport system in Madin-Darby canine kidney (MDCK) cells. The primary aim of the present study was to evaluate whether the H+/lipophilic cation antiport system is expressed in porcine LLC-PK1 cells. That is, we investigated uptake and/or efflux of QND and another cation, bisoprolol, in LLC-PK1 cells. In addition, we studied the renal clearance of bisoprolol in rats. Uptake of QND into LLC-PK1 cells was decreased by acidification of the extracellular pH or alkalization of the intracellular pH. Cellular uptake of QND from the apical side was much greater than from the basolateral side. In addition, apical efflux of QND from LLC-PK1 cells was increased by acidification of the extracellular pH. Furthermore, lipophilic cationic drugs significantly reduced uptake of bisoprolol in LLC-PK1 cells. Renal clearance of bisoprolol in rats was approximately 7-fold higher than that of creatinine, and was markedly decreased by alkalization of the urine pH. The present study suggests that the H+/lipophilic cation antiport system is expressed in the apical membrane of LLC-PK1 cells. Moreover, the H+/lipophilic cation antiport system may be responsible for renal tubular secretion of bisoprolol in rats.