Search

Summary The present study aimed to examine the effectiveness of bi-level positive airway pressure (BiPAP) versus continuous positive airway pressure (CPAP) in preterm infants with birth weight less than 1500 g and respiratory distress syndrome (RDS) following intubation-surfactant-extubation (INSURE) treatment. A two-center randomized control trial was performed. The primary outcome was the reintubation rate of infants within 72 h of age after INSURE. Secondary outcomes included bronchopulmonary dysplasia (BPD), necrotizing enterocolitis (NEC), retinopathy of prematurity (ROP) and incidences of adverse events. Lung function at one year of corrected age was also compared between the two groups. There were 140 cases in the CPAP group and 144 in the BiPAP group. After INSURE, the reintubation rates of infants within 72 h of age were 15% and 11.1% in the CPAP group and the BiPAP group, respectively (P>0.05). Neonates in the BiPAP group was on positive airway pressure (PAP) therapy three days less than in the CPAP group (12.6 d and 15.3 d, respectively, P0.05). There was no difference in lung function at one year of age between the two groups (P>0.05). In conclusion, after INSURE, the reintubation rate of infants within 72 h of age was comparable between the BiPAP group and the CPAP group. BiPAP was superior to CPAP in terms of shorter durations (days) on PAP support and oxygen supplementation. There were no differences in the incidences of BPD and ROP, and lung function at one year of age between the two ventilation methods.

This study aimed to analyze counter-movement jump (CMJ) performance in time and frequency domains. Fortyfour Division I American football players participated in the study. Kinetic variables were collected from both dominant and non-dominant legs using two force plates. Normalized peak power, normalized net impulse, and normalized peak force significantly correlated with jump height (r = .960, r = .998, r = .725, respectively with p < .05). The mean frequency component was significantly correlated with CMJ performance (r = .355 with p < .05). The reliability of the frequency variables was higher than the time domain variables. Frequency domain variables showed weaker correlations with jump height compared with time domain variables. Frequency domain analysis provides frequency components, which represent the rate of energy transmission from the eccentric phase to the end of the push-off phase. Frequency component information may provide additional information for the analyses of CMJ performance for athletes.

Deep learning (DL) has attracted the attention of computational toxicologists as it offers a potentially greater power for in silico predictive toxicology than existing shallow learning algorithms. However, contradicting reports have been documented. To further explore the advantages of DL over shallow learning, we conducted this case study using two cell-based androgen receptor (AR) activity datasets with 10K chemicals generated from the Tox21 program. A nested double-loop cross-validation approach was adopted along with a stratified sampling strategy for partitioning chemicals of multiple AR activity classes (i.e., agonist, antagonist, inactive, and inconclusive) at the same distribution rates amongst the training, validation and test subsets. Deep neural networks (DNN) and random forest (RF), representing deep and shallow learning algorithms, respectively, were chosen to carry out structure-activity relationship-based chemical toxicity prediction. Results suggest that DNN significantly outperformed RF (p < 0.001, ANOVA) by 22–27% for four metrics (precision, recall, F-measure, and AUPRC) and by 11% for another (AUROC). Further in-depth analyses of chemical scaffolding shed insights on structural alerts for AR agonists/antagonists and inactive/inconclusive compounds, which may aid in future drug discovery and improvement of toxicity prediction modeling.

In recent years, there has been an increase in the association of technology in sports and live sports broadcasting networks. From score updates, broadcasting commercials, assisting referees for decision making, and minimizing errors, the adoption of technology has been used for fair play and improve results. This has been possible with the advancement in video analysis, classification techniques, and the availability of resources. This paper introduces a new labelled video dataset collected from a live basketball game broadcasted on live TV to determine the type of basket scored in the basketball game. Among different shots, the points the player can score are basically of three types: 3 points, 2 points, which depends on the range of shots taken and 1 point which is the free shots taken after a foul. This dataset consists of labelled video clips collected from the live broadcast of the game from the broadcasting medium to classify different scoring activities. This paper also gives the preliminary analysis of the dataset for different class labels using 3D ConvNet and two-stream 3D ConvNet methods to show the complexity of the dataset.

Objective The perinatal consequences of neonates born to severe acute respiratory syndrome-associated coronavirus-2 (SARS-CoV-2) infected mothers are uncertain. This study aimed to compare the differences in clinical manifestation, laboratory results, and outcomes of neonates born to mothers with or without coronavirus disease 2019 (COVID-19). Study Design A total of 48 neonates were admitted to Tongji Hospital and HuangShi Maternal and Child Healthcare Hospital from January 17 to March 4, 2020. The neonates were divided into three groups according to the mothers' conditions: neonates born to mothers with confirmed COVID-19, neonates born to mothers with clinically diagnosed COVID-19, and neonates born to mothers without COVID-19. The clinical data of mothers and infants in the three groups were collected, compared, and analyzed. Results The deliveries occurred in a negative pressure isolation room, and the neonates were separated from their mothers immediately after birth for further observation and treatment. None of the neonates showed any signs of fever, cough, dyspnea, or diarrhea. SARS-CoV-2 reverse transcriptase-polymerase chain reaction of the throat swab and feces samples from the neonates in all three groups was negative. No differences were detected in the whole blood cell, lymphocytes, platelet, and liver and renal function among the three groups. All mothers and their infants showed satisfactory outcomes, including a 28-week preterm infant. Conclusion The clinical manifestations, radiological, and biochemical results did not show any difference between the three groups. No evidence of vertical transmission was found in this study whether the pregnant women developed coronavirus infection in the third (14 cases) or second trimester (1 case). Key points Characteristics of neonates born to mothers with and without COVID-19 have been compared. All the 48 cases presented in the study had good outcomes. A 28-week preterm born to COVID-19 mother presented to be clear of SARS-COV-2 infection.

The aim of this study was to investigate the clinical characteristics of neonates born to SARS-CoV-2 infected mothers and increase the current knowledge on the perinatal consequences of COVID-19. Nineteen neonates were admitted to Tongji Hospital from January 31 to February 29, 2020. Their mothers were clinically diagnosed or laboratory-confirmed with COVID-19. We prospectively collected and analyzed data of mothers and infants. There are 19 neonates included in the research. Among them, 10 mothers were confirmed COVID-19 by positive SARS-CoV-2 RT-PCR in throat swab, and 9 mothers were clinically diagnosed with COVID-19. Delivery occurred in an isolation room and neonates were immediately separated from the mothers and isolated for at least 14 days. No fetal distress was found. Gestational age of the neonates was 38.6 ± 1.5 weeks, and average birth weight was 3293 ± 425 g. SARS-CoV-2 RT-PCR in throat swab, urine, and feces of all neonates were negative. SARS-CoV-2 RT-PCR in breast milk and amniotic fluid was negative too. None of the neonates developed clinical, radiologic, hematologic, or biochemical evidence of COVID-19. No vertical transmission of SARS-CoV-2 and no perinatal complications in the third trimester were found in our study. The delivery should occur in isolation and neonates should be separated from the infected mothers and care givers.

Additional file 1: Text S1. SMOTEENN algorithm. Figure S1. Illustration of SMOTE and ENN techniques. (a) The original imbalanced data; (b) Synthetic samples are generated for the minority class using SMOTE. (c) Using ENN, those mislabeled synthetic samples were removed from the minority class. (d) The rebalanced data after the application of SMOTEENN. Table S1. Evaluation metrics derived for four classification methods (RF, RUS, SMO and SMN) with twelve Tox21 qHTS assay datasets. Specificity and two other metrics (sensitivity and SSG, both appearing in Table 2) are shown.

The specificity of toxicant-target biomolecule interactions lends to the very imbalanced nature of many toxicity datasets, causing poor performance in Structure–Activity Relationship (SAR)-based chemical classification. Undersampling and oversampling are representative techniques for handling such an imbalance challenge. However, removing inactive chemical compound instances from the majority class using an undersampling technique can result in information loss, whereas increasing active toxicant instances in the minority class by interpolation tends to introduce artificial minority instances that often cross into the majority class space, giving rise to class overlapping and a higher false prediction rate. In this study, in order to improve the prediction accuracy of imbalanced learning, we employed SMOTEENN, a combination of Synthetic Minority Over-sampling Technique (SMOTE) and Edited Nearest Neighbor (ENN) algorithms, to oversample the minority class by creating synthetic samples, followed by cleaning the mislabeled instances. We chose the highly imbalanced Tox21 dataset, which consisted of 12 in vitro bioassays for > 10,000 chemicals that were distributed unevenly between binary classes. With Random Forest (RF) as the base classifier and bagging as the ensemble strategy, we applied four hybrid learning methods, i.e., RF without imbalance handling (RF), RF with Random Undersampling (RUS), RF with SMOTE (SMO), and RF with SMOTEENN (SMN). The performance of the four learning methods was compared using nine evaluation metrics, among which F1 score, Matthews correlation coefficient and Brier score provided a more consistent assessment of the overall performance across the 12 datasets. The Friedman’s aligned ranks test and the subsequent Bergmann-Hommel post hoc test showed that SMN significantly outperformed the other three methods. We also found that a strong negative correlation existed between the prediction accuracy and the imbalance ratio (IR), which is defined as the number of inactive compounds divided by the number of active compounds. SMN became less effective when IR exceeded a certain threshold (e.g., > 28). The ability to separate the few active compounds from the vast amounts of inactive ones is of great importance in computational toxicology. This work demonstrates that the performance of SAR-based, imbalanced chemical toxicity classification can be significantly improved through the use of data rebalancing.

Deep learning (DL) has attracted the attention of computational toxicologists as it offers a potentially greater power for in silico predictive toxicology than existing shallow learning algorithms. However, contradicting reports have been documented. To further explore the advantages of DL over shallow learning, we conducted this case study using two cell-based androgen receptor (AR) activity datasets with 10K chemicals generated from the Tox21 program. A nested double-loop cross-validation approach was adopted along with a stratified sampling strategy for partitioning chemicals of multiple AR activity classes (i.e., agonist, antagonist, inactive, and inconclusive) at the same distribution rates amongst the training, validation and test subsets. Deep neural networks (DNN) and random forest (RF), representing deep and shallow learning algorithms, respectively, were chosen to carry out structure-activity relationship-based chemical toxicity prediction. Results suggest that DNN significantly outperformed RF (p < 0.001, ANOVA) by 22–27% for four metrics (precision, recall, F-measure, and AUPRC) and by 11% for another (AUROC). Further in-depth analyses of chemical scaffolding shed insights on structural alerts for AR agonists/antagonists and inactive/inconclusive compounds, which may aid in future drug discovery and improvement of toxicity prediction modeling.

In silico toxicity prediction plays an important role in the regulatory decision making and selection of leads in drug design as in vitro/vivo methods are often limited by ethics, time, budget, and other resources. Many computational methods have been employed in predicting the toxicity profile of chemicals. This review provides a detailed end-to-end overview of the application of machine learning algorithms to Structure-Activity Relationship (SAR)-based predictive toxicology. From raw data to model validation, the importance of data quality is stressed as it greatly affects the predictive power of derived models. Commonly overlooked challenges such as data imbalance, activity cliff, model evaluation, and definition of applicability domain are highlighted, and plausible solutions for alleviating these challenges are discussed.

The challenge of developing facial recognition systems has been the focus of many research efforts in recent years and has numerous applications in areas such as security, entertainment, and biometrics. Recently, most progress in this field has come from training very deep neural networks on massive datasets which is computationally intensive and time consuming. Here, we propose a deep transfer learning (DTL) approach that integrates transfer learning techniques and convolutional neural networks and apply it to the problem of facial recognition to fine-tune facial recognition models. Transfer learning can allow for the training of robust, high-performance machine learning models that require much less time and resources to produce than similarly performing models that have been trained from scratch. Using a pre-trained face recognition model, we were able to perform transfer learning to produce a network that is capable of making accurate predictions on much smaller datasets. We also compare our results with results produced by a selection of classical algorithms on the same datasets to demonstrate the effectiveness of the proposed DTL approach.

Background Multi-label classification of data remains to be a challenging problem. Because of the complexity of the data, it is sometimes difficult to infer information about classes that are not mutually exclusive. For medical data, patients could have symptoms of multiple different diseases at the same time and it is important to develop tools that help to identify problems early. Intelligent health risk prediction models built with deep learning architectures offer a powerful tool for physicians to identify patterns in patient data that indicate risks associated with certain types of chronic diseases. Results Physical examination records of 110,300 anonymous patients were used to predict diabetes, hypertension, fatty liver, a combination of these three chronic diseases, and the absence of disease (8 classes in total). The dataset was split into training (90%) and testing (10%) sub-datasets. Ten-fold cross validation was used to evaluate prediction accuracy with metrics such as precision, recall, and F-score. Deep Learning (DL) architectures were compared with standard and state-of-the-art multi-label classification methods. Preliminary results suggest that Deep Neural Networks (DNN), a DL architecture, when applied to multi-label classification of chronic diseases, produced accuracy that was comparable to that of common methods such as Support Vector Machines. We have implemented DNNs to handle both problem transformation and algorithm adaption type multi-label methods and compare both to see which is preferable. Conclusions Deep Learning architectures have the potential of inferring more information about the patterns of physical examination data than common classification methods. The advanced techniques of Deep Learning can be used to identify the significance of different features from physical examination data as well as to learn the contributions of each feature that impact a patient’s risk for chronic diseases. However, accurate prediction of chronic disease risks remains a challenging problem that warrants further studies.

The challenge of developing facial recognition systems has been the focus of many research efforts in recent years and has numerous applications in areas such as security, entertainment, and biometrics. Recently, most progress in this field has come from training very deep neural networks on massive datasets. Here, we use a pre-trained face recognition model and perform transfer learning to produce a network that is capable of making accurate predictions on a much smaller dataset. We also compare our results with results produced by a selection of classical algorithms on the same dataset.

In this paper, we introduce an experiment evaluating performance of football players in countermovement jumps (CMJs). Three methods including time domain, frequency domain, and machine learning algorithms are proposed for performance evaluation. Correlation coefficients and p-values are given for time domain and frequency domain methods, and prediction errors are given for different machine learning algorithms.

For the first time, uniform α- and β-Bi 2 O 3 hollow microspheres were successfully synthesized through a facile template-free solvothermal method followed by calcination of the precursor. A mixture of glycerol (Gly) and ethanol (EtOH) (volume ratio of Gly–EtOH = 1:1) was selected as the solvent, Bi(NO 3 ) 3 ⋅5H 2 O as the bismuth source. The resulting samples were characterized by thermogravimetry (TG), Fourier transform infrared spectra (FT-IR), X-ray powder diffraction (XRD), scanning and transmission electron microscopy (SEM, TEM), Brunauer–Emmett–Teller (BET) surface area and UV–Vis diffuse reflectance spectra (DRS). Based on the electron microscope observations, the formation of such hollow spheres structure was proposed as a self-assembly process followed by Ostwald ripening. The sample calcined at 270 °C exhibited β-Bi 2 O 3 phase, whereas the sample calcined at 350 °C exhibited α phase. The band gaps of the as-prepared materials were estimated by DRS to be 3.50, 2.82, and 2.36 eV for the precursor, α- and β-Bi 2 O 3 , respectively. Upon visible light irradiation, the β-Bi 2 O 3 exhibits much higher photocatalytic activities than the α-Bi 2 O 3 and precursor for the degradation of rhodamine B (RhB) in aqueous solution, which could be attributed to its narrower band gap and efficient visible light absorption.

Blackberry-like CdS microspheres were synthesized by a facile hydrothermal method. Subsequently, through a simple ion-exchange reaction between the CdS microsphere and Bi(NO3)3·5H2O, novel urchin-shaped CdS/Bi2S3 core/shell heterostructures have successfully been fabricated. The photocatalytic activity and photostability were evaluated by degradation of rhodamine B (RhB) under visible light irradiation. The results show that the content of Bi2S3 had important effects on the photocatalytic activity, and the CdS/Bi2S3 heterostructures with the theoretical molar ratio of Bi/Cd 0.25 exhibited the best photocatalytic activity. The improved photocatalytic activity and photostability were attributed to the visible light absorption enhanced by Bi2S3 and the formation of a heterojunction between CdS and Bi2S3, which can effectively accelerate the charge separation and transfer corrosive holes from CdS to robust Bi2S3. Experimentally and theoretically, Bi2S3 anchored on CdS acts as a corrosion protection agent for CdS and also acts as a co-catalyst.

An adaptive, implicit, multiresolution time-domain (MRTD) algorithm is applied to solve the integro-differential equation arising from the canonical straight thin wire scattering problem. The Haar wavelet system is adopted for the expansion of the induced current along the wire responding to a Gaussian pulsed plane wave electrical field. The results from our marching-on-in-time (MOT) method are compared with those from time-domain moment method where pulse basis functions are utilized. The wavelet system provides an adaptive solution to the transient problem, reducing computational cost, both in time and storage, with a controllable level of accuracy.

Heterojunction construction is an exciting direction to pursue for highly active photocatalysts. In this study, novel core/shell β-Bi2O3/Bi2S3 hollow heterostructures were successfully synthesized through a simple and economical ion exchange method between β-Bi2O3 hollow microspheres and thioacetamide (CH3CSNH2, TAA), and characterized by multiform techniques, such as XRD, XPS, SEM, TEM, BET, DRS and PL. The results indicated that the core/shell β-Bi2O3/Bi2S3 hollow heterostructures exhibited strong absorption in visible light region and excellent photocatalytic performance for decomposing rhodamine B (RhB) compared with pure β-Bi2O3 under visible light irradiation. Among the β-Bi2O3/Bi2S3 photocatalysts with different molar percentage of Bi2S3 to initial β-Bi2O3, the β-Bi2O3/Bi2S3 (10%) heterostructures exhibited the highest photocatalytic activity, which was about 3.3 times higher than that of pure β-Bi2O3 sample. Moreover, the study on the mechanism suggested that the enhanced photocatalytic activity mainly resulted from the role of β-Bi2O3–Bi2S3 heterojunction formed in the β-Bi2O3/Bi2S3, which could lead to efficient separation of photoinduced carriers. Additionally, the photosensitization of Bi2S3 and the hollow nature of β-Bi2O3 were also responsible for the high photocatalytic activity.

The distinction between electrical engineering (EE) and electronic engineering technology (EET) programs is obvious. While an EE curriculum is more theoretical, EET curriculum is generally focused on an experience-building learning style and appeals to students who learn best by their hands-on experience. Electronic engineering technology programs prepare students for practical design and production work.

A multiresolution time-domain algorithm is applied to solve the integro-differential equation arising from the canonical straight thin wire scattering problem. The Haar wavelet system is adopted for the expansion of the induced current along the wire responding to a Gaussian input. The results from our explicit marching-on-time method are compared with those from frequency-domain moment method and time-domain moment method where pulse basis functions are utilized. The wavelet system provides an adaptive solution to the transient problem. Thus reducing the required memory and computation time.

An adaptive, implicit, multi-resolution time-domain method is applied to solve the integro-differential equations arising from the canonical thin wire scattering problems. The Haar wavelet system is adopted for the basis functions to decompose the induced current along the wire due to an input electric field. Both algorithms are discussed for straight thin wire and arbitrary thin wire problems. Results from our implicit marching-on-time algorithm using Haar wavelet basis functions are compared with those from pulse basis functions.