Search

• Multi-view Graph Neural Network captures the diversity of multi-sensor signals. • Two single views in the multi-view graph are fused by a view-attention block. • Multi-view Graph Neural Network shows robust diagnosis even under strong noise. Modern large-scale equipment is deployed with multiple sensors to monitor the operating state in real time, thus imposing higher requirements on intelligent fault diagnosis methods. However, current deep learning-based methods for multi-sensor information fusion often rely on features extracted from a single domain, which are incompetent to characterize the diversity and complexity of multi-sensor signals. To fully exploit the potential of multi-domain features, a Multi-view Graph Neural Network (MvGNN) combining time domain (TD) and frequency domain (FD) features is proposed for the fault diagnosis of a multi-sensor rotating machine system. Firstly, the normalized TD signals are modeled as graph-structured data in terms of the k nearest neighbor (k NN) algorithm. The nodes' initial features are transformed into two different feature spaces using Convolutional Neural Network (CNN) and Fast Fourier Transform (FFT) to form multi-view (TD and FD) graphs. Subsequently, single-view graphs are learned by independent graph convolution blocks separately to aggregate the multi-sensor information. Lastly, a view-attention block is designed to compute the unified representation of the multi-view graph, which is subsequently input into a classifier to diagnose the health state. To verify the capabilities of MvGNN, two case studies are performed on public datasets. Experimental results show that the proposed method has satisfactory diagnostic accuracy and surpasses comparative methods. In addition, the abundant information contained in the multi-view graph endows the proposed method with stronger robustness in a noisy environment. [ABSTRACT FROM AUTHOR]

Nowadays, deep learning-based models have been widely developed for atrial fibrillation (AF) detection in electrocardiogram (ECG) signals. However, owing to the inevitable over-fitting problem, classification accuracy of the developed models severely differed when applying on the independent test datasets. This situation is more significant for AF detection from dynamic ECGs. In this study, we explored two potential training strategies to address the over-fitting problem in AF detection. The first one is to use the Fast Fourier transform (FFT) and Hanning-window-based filter to suppress the influence from individual difference. Another is to train the model on the wearable ECG data to improve the robustness of model. Wearable ECG data from 29 patients with arrhythmia were collected for at least 24 h. To verify the effectiveness of the training strategies, a Long Short-Term Memory (LSTM) and Convolution Neural Network (CNN)-based model was proposed and tested. We tested the model on the independent wearable ECG data set, as well as the MIT-BIH Atrial Fibrillation database and PhysioNet/Computing in Cardiology Challenge 2017 database. The model achieved 96.23%, 95.44%, and 95.28% accuracy rates on the three databases, respectively. Pertaining to the comparison of the accuracy rates on each training set, the accuracy of the model trained in conjunction with the proposed training strategies only reduced by 2%, while the accuracy of the model trained without the training strategies decreased by approximately 15%. Therefore, the proposed training strategies serve as effective mechanisms for devising a robust AF detector and significantly enhanced the detection accuracy rates of the resulting deep networks. [ABSTRACT FROM AUTHOR]

Coral probiotics can improve the tolerance of corals to heat stress, thus mitigating the process of coral thermal bleaching. Sensitive and specific detection of coral probiotics at low abundances is highly desirable but remains challenging, especially for rapid and on-site detection of coral probiotics. Since the electrochemical biosensor has been recently used in the field of environmental DNA (eDNA) detection, herein, an efficient electrochemical biosensor was developed based on CoS 2 /CoSe 2 -NC HNCs electrode material with a specific DNA probe for the C. marina detection. After optimization, the lower limit of detection (LOD) values of such biosensors for the target DNA and genomic DNA were 1.58 fM and 6.5 pM, respectively. On this basis, a portable device was constructed for the practical detection of C. marina eDNA, and its reliability and accuracy were verified by comparison with the ddPCR method (P > 0.05). For each analysis, the average cost was only ∼ $1.08 and could be completed within 100 min with reliable sensitivity and specificity. Overall, the biosensor could reflect the protective effect of probiotics on coral heat stress, and the proposed technique will put new insights into the rapid and on-site detection of coral probiotics to assist corals against global warming. [ABSTRACT FROM AUTHOR]

P-Chlorophenols (4-CP) are known as persistent toxic compounds for their acute toxicity and poor biodegradability. When the concentration of phenolic substances in water reaches 0.1005 mg/L, it becomes undrinkable. As a result, the detection of this human-toxic compound was indispensable. Hence, the high sensitivity and absorption ability of MOFs@MOFs composites (HP-UIO66@Ni-BDC) was applied to detect the 4-CP, which was constructed by using the ligand defect NH 2 -UIO-66 and Ni-BDC via a ligand linking approach. The improved performance of the materials compared to bare MOFs was attributable to the enhanced catalytic ability by the interaction of the MOFs and the high affinity between the 4-CP and the MOFs@MOFs composites (the electron donor-acceptor interaction and π-π stacking interaction). The electrochemical results illustrated the satisfactory electrochemical performances of modified GCEs toward 4-CP, such as the low limit of detection (LOD = 0.328 μM), wider linear detection range from 10 to 160 μM, and good stability and selectivity. Besides, the recovery experiments' results have shown that the electrochemical sensor could be used in the actual sample detecting. All the results suggested that the prepared HP-UIO66@Ni-BDC composites could serve as a prospective sensor for the real-time detection of 4-CP. [Display omitted] • The double MOF composite of HP-UIO66@Ni-BDC electrode was successfully synthesized. • Ligand connecting mode can promote electron transfer and electrochemical sensing. • Low LOD and wide detection range for detecting P-Chlorophenol were realized. • The reliability, veracity, and practicality of the electrochemical sensor was verified. [ABSTRACT FROM AUTHOR]

[Display omitted] • Uniform double-shell N C@SnO 2 HNs were assembled by interfacial bonding. • Interface Sn-C bond was proved by XPS characterization and DFT calculation. • Sn-C synergistic effect promotes the electron transfer and electrochemical sensing. • Low LOD (41.5 nM) and wide detection range (0.5–25 μM) of DA were realized. • The portable microchip was developed in service to in-situ DA detection. Higher resistance is exceptionally detrimental to metal oxide-based electrochemical sensors' detection performances that depend on their electrical conductivity. Herein, guided by the synergistic effect of the heterophase boundary, a hollow core–shell combination with a double-layered structure was engineered via hydrolysis strategy, namely nitrogen-doped carbon hollow nano-microspheres (N C HNMs) coated with SnO 2 outer layer (N C@SnO 2 HNMs), then applied for high-performance electrochemical sensing. Advantages including large surface area and strong electronic coupling of Sn-C bond on the boundary facilitated the adsorption of target molecules and electron transferring across the composites, achieving efficient quantitative identification. Density functional theory (DFT) calculation demonstrated the motivation of the charge accumulation at the interface due to the formation of the Sn-C bond, which acted as a bridge for charge transfer. Unsurprisingly, the N C@SnO 2 HNMs-based electrode showed an excellent performance toward dopamine (DA), such as favorable DA selectivity, a lower limit of detection, and a wide linear detection range. Moreover, the electrode material could be applied to the microsensor for in-situ DA detection in artificial sweat. [ABSTRACT FROM AUTHOR]

Recent studies have indicated that coral mucus plays an important role in the bioaccumulation of a few organic pollutants by corals, but no relevant studies have been conducted on organochlorine pesticides (OCPs). Previous studies have also indicated that OCPs widely occur in a few coral reef ecosystems and have a negative effect on coral health. Therefore, this study focused on the occurrence and bioaccumulation of a few OCPs, such as dichlorodiphenyltrichloroethanes (DDTs), hexachlorobenzene (HCB) and p , p ′-methoxychlor (MXC), in the coral tissues and mucus as well as in plankton and seawater from a coastal reef ecosystem (Weizhou Island) in the South China Sea. The results indicated that DDTs were the predominant OCPs in seawater and marine biota. Higher concentrations of OCPs in plankton may contribute to the enrichment of OCPs by corals. The significantly higher total OCP concentration (∑ 8 OCPs) found in coral mucus than in coral tissues suggested that coral mucus played an essential role in resisting enrichment of OCPs by coral tissues. This study explored the different functions of coral tissues and mucus in OCP enrichment and biodegradation for the first time, highlighting the need for OCP toxicity experiments from both tissue and mucus perspectives. [Display omitted] • OCPs were firstly studied in coral mucus in the South China Sea. • Plankton may have an influence on coral enrichment of OCPs. • Obvious higher ∑ 8 OCPs occurred in coral mucus than in coral tissues. • Coral mucus showed lower metabolic capability of DDTs than coral tissues. • Coral mucus may resist the enrichment and metabolism of OCPs by coral tissues. [ABSTRACT FROM AUTHOR]

Breast cancer (BC) is a serious threat to women's life and healthy. Increasing evidence indicated that blocking Warburg effect could attenuate the development of BC. Circular RNAs (circRNAs) has been found to be dysregulated in various carcinomas, including BC. Our study aims to illustrate the role and regulatory mechanism of circ_0039960 in BC development. RT-qPCR and western blotting were utilized to evaluate the expression of circ_0039960 in tissues recruited from 32 cases of BC patients and also BC cell lines. Circ_0039960 shRNA was transfected into cells to explore its function on cell processes. CCK-8, flow cytometry and ELISA were used to measure cell viability, cell cycle and apoptosis. Warburg effect was detected by using commercial kits. Besides, bioinformatic prediction, RIP and luciferase reporter assays were performed to validate the interactions between circ_0039960, miR-1178 and PRMT7. The results showed that circ_0039960 and PRMT7 were both up-regulated, while miR-1178 was down-regulated, in BC tissues and cells. Silencing circ_0039960 effectively inhibited cell viability and Warburg effect of BC cells, also, induced cell cycle arrest and apoptosis. Moreover, we validated that circ_0039960 positively mediated PRMT7 expression via directly targeting to miR-1178. The inhibition of miR-1178 and overexpression of PRMT7 reversed the effect of circ_0039960 knockdown on BC cell growth and Warburg effect. In general, our research demonstrated that circ_0039960 regulates cell growth and Warburg effect in BC cells via miR-1178/PRMT7 axis. This may provide new evidence for the exploration of BC diagnostic and therapeutic targets. [Display omitted] • Circ_0039960 was up-regulated and correlated with BC prognosis. • Inhibiting circ_0039960 suppressed BC cell processes. • Inhibiting circ_0039960depressed Warburg effect in BC cells. • Circ_0039960 targeted to PRMT7 via sponging miR-1178. • Circ_0039960regulated BC cell processes and Warburg effect via miR-1178/PRMT7. [ABSTRACT FROM AUTHOR]

Abstract: p-Type and n-type Cu2O thin films were controllably prepared using a simple solvothermal method by adjusting pH value of the copper (II) acetate aqueous solution. Photoelectrochemical experiments show that the Cu2O thin films synthesized in acid and alkaline (or neutral) media present n-type and p-type semiconductor character, respectively. Moreover, the films prepared at pH 5 have the best photoelectrochemical properties. The mechanism for the formation of these p-type and n-type Cu2O films is discussed. The Cu2O p–n homojunction fabricated in this study shows typical p–n junction character. This facile preparation method may be a promising way to prepare p–n homojunctions for semiconductor devices. [Copyright &y& Elsevier]

Abstract: CaSn(OH)6 nanotubes were fabricated by sonochemical precipitation method at room temperature. A direct rolling process from nanosheets to nanotubes was expected for the synthesis of CaSn(OH)6 nanotubes. The transient CaSn(OH)6 nanosheets are formed as intermediates produced by the spontaneous self-assembly and transformation of amorphous colloid clusters. During the crystallization process of intermediate nanosheets, the relaxation of surface strain in the nanosheet interfaces can induce the nanosheets to roll up to form nanotubes under ultrasonic conditions. In this synthesis, the addition of Na2CO3 seems to play an important role in the formation, size, and shape control of the nanotubes. Investigations into the stability performance of the nanotubes indicate that the morphologies are very sensitive to pH and temperature. The method suggests a general strategy for the design and fabrication of functional single-crystalline nanotubes through an intermediate nanosheet rolling process. The in vitro fabrication of such single crystal nanotubes could shed light on fundamental mechanisms for closed hollow nanostructures. Furthermore, nanotubes produced in high yield and at low cost are envisioned to have applications in areas ranging from medicine to pharmaceuticals through to materials science. [Copyright &y& Elsevier]

The ability to identify objects of interest from digital visual signals is critical for many applications of intelligent systems. For such object detection task, accuracy and computational efficiency are two important aspects, especially for applications with real-time requirement. In this paper, we study shuttlecock detection problem of a badminton robot, which is very challenging since the shuttlecock often moves fast in complex contexts, and must be detected precisely in real time so that the robot can plan and execute its following movements. To this end, we propose two novel variants of Tiny YOLOv2, a well-known deep learning based detector. We first modify the loss function to adaptively improve the detection speed for small objects such as shuttlecock. We then modify the architecture of Tiny YOLOv2 to retain more semantic information of small objects, so as to further improve the performance. Experimental results show that the proposed networks can achieve high detection accuracy with the fastest speed, compared with state-of-the-art deep detectors such as Faster R-CNN, SSD, Tiny YOLOv2, and YOLOv3. Our methods could be potentially applied to other tasks of detecting high-speed small objects. • Tackle the challenging problem of detecting shuttlecock in real-time. • Define a new loss function to better adapt to the task of detecting small objects. • Modify state-of-the-art deep architecture to retain more semantic information. • Results on real-world dataset show the effectiveness in both accuracy and speed. [ABSTRACT FROM AUTHOR]

• The CLUMondo model can effectively simulate future land use change trends. • In the three major functional zones, the carbon storage varies greatly. • The loss of forest and cropland leads to the reduction of carbon storage. In order to achieve the dual goals of territorial development optimization and protection, the Major Function Oriented Zone (MFOZ) has implemented differentiated spatial policies, which will lead to significant changes in future land use and ecosystem services. This study takes Nanning, China as the study area, and integrates the CLUMondo model and the InVEST model to assess the impact of land use change on carbon storage from 2010–2015 and 2015–2030 in three different scenarios. The results show that: (1) under the Natural Growth Scenario (NGS) and Economic Development Scenario (EDS), the overall carbon storage in the study area will decrease, while, under the Ecological Protection Scenario (EPS), the overall carbon storage will increase; (2) under the NGS, carbon storage in the key development zone will decrease, while it will increase in the agricultural production zone and in the key ecological function zone; (3) under the EDS, carbon storage will decrease in all the three major functional zones; (4) under the EPS, carbon storage will increase in all the three major functional zones. In the future, the three major functional zones should implement differentiated development policies to further optimize the land spatial development pattern and realize sustainable development. [ABSTRACT FROM AUTHOR]

Unsupervised Domain Adaptation (UDA) methods have been successful in reducing label dependency by minimizing the domain discrepancy between labeled source domains and unlabeled target domains. However, these methods face challenges when dealing with Multivariate Time-Series (MTS) data. MTS data typically originates from multiple sensors, each with its unique distribution. This property poses difficulties in adapting existing UDA techniques, which mainly focus on aligning global features while overlooking the distribution discrepancies at the sensor level, thus limiting their effectiveness for MTS data. To address this issue, a practical domain adaptation scenario is formulated as Multivariate Time-Series Unsupervised Domain Adaptation (MTS-UDA). In this paper, we propose SEnsor Alignment (SEA) for MTS-UDA, aiming to address domain discrepancy at both local and global sensor levels. At the local sensor level, we design endo-feature alignment, which aligns sensor features and their correlations across domains. To reduce domain discrepancy at the global sensor level, we design exo-feature alignment that enforces restrictions on global sensor features. We further extend SEA to SEA++ by enhancing the endo-feature alignment. Particularly, we incorporate multi-graph-based higher-order alignment for both sensor features and their correlations. Extensive empirical results have demonstrated the state-of-the-art performance of our SEA and SEA++ on six public MTS datasets for MTS-UDA.

Objective: To explore the interaction between family members and nursing home staff during the adjustment period of newly admitted elderly individuals in a nursing home., Design: A qualitative descriptive study based on semistructured interviews; data were analysed using a thematic topic analysis approach., Setting: Interviews were conducted face-to-face., Participants: An interview was undertaken with 15 nursing home staff and 12 family members of recently admitted elderly individuals in three nursing facilities from November 2022 to January 2023 in a major Chinese urban centre., Results: This study identified 5 main themes and 10 subthemes. The 5 themes identified are: Information transmission and sharing, pleasant interaction atmosphere, interaction dilemmas and challenges, inadequate organisational management and the necessity and anticipation of interaction., Conclusion: During the process of older persons transitioning to institutional care, we discovered issues and unresolved requirements in the interactions between family members and nursing home staff. In summary, there is a need to enhance the development of an interactive environment that supports the institutionalisation of older persons. The findings derived from this study are valuable for developing interactive programmes., Trial Registration Number: ChiCTR2100045590, date: 19 April 2021., Competing Interests: Competing interests: None declared., (© Author(s) (or their employer(s)) 2024. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.)

Automatic emotion recognition based on multichannel Electroencephalography (EEG) holds great potential in advancing human-computer interaction. However, several significant challenges persist in existing research on algorithmic emotion recognition. These challenges include the need for a robust model to effectively learn discriminative node attributes over long paths, the exploration of ambiguous topological information in EEG channels and effective frequency bands, and the mapping between intrinsic data qualities and provided labels. To address these challenges, this study introduces the distribution-based uncertainty method to represent spatial dependencies and temporal-spectral relativeness in EEG signals based on Graph Convolutional Network (GCN) architecture that adaptively assigns weights to functional aggregate node features, enabling effective long-path capturing while mitigating over-smoothing phenomena. Moreover, the graph mixup technique is employed to enhance latent connected edges and mitigate noisy label issues. Furthermore, we integrate the uncertainty learning method with deep GCN weights in a one-way learning fashion, termed Connectivity Uncertainty GCN (CU-GCN). We evaluate our approach on two widely used datasets, namely SEED and SEEDIV, for emotion recognition tasks. The experimental results demonstrate the superiority of our methodology over previous methods, yielding positive and significant improvements. Ablation studies confirm the substantial contributions of each component to the overall performance.

Background: Human placental mesenchymal stromal cells (hPMSCs) are known to limit graft-versus-host disease (GVHD). CD8 + CD122 + PD-1 + Tregs have been shown to improve the survival of GVHD mice. However, the regulatory roles of hPMSCs in this subgroup remain unclear. Here, the regulatory mechanism of hPMSCs in reducing liver fibrosis in GVHD mice by promoting CD8 + CD122 + PD-1 + Tregs formation and controlling the balance of IL-6 and IL-10 were explored., Methods: A GVHD mouse model was constructed using C57BL/6J and BALB/c mice and treated with hPMSCs. LX-2 cells were explored to study the effects of IL-6 and IL-10 on the activation of hepatic stellate cells (HSCs). The percentage of CD8 + CD122 + PD-1 + Tregs and IL-10 secretion were determined using FCM. Changes in hepatic tissue were analysed by HE, Masson, multiple immunohistochemical staining and ELISA, and the effects of IL-6 and IL-10 on LX-2 cells were detected using western blotting., Results: hPMSCs enhanced CD8 + CD122 + PD-1 + Treg formation via the CD73/Foxo1 and promoted IL-10, p53, and MMP-8 levels, but inhibited IL-6, HLF, α-SMA, Col1α1, and Fn levels in the liver of GVHD mice through CD73. Positive and negative correlations of IL-6 and IL-10 between HLF were found in liver tissue, respectively. IL-6 upregulated HLF, α-SMA, and Col1α1 expression via JAK2/STAT3 pathway, whereas IL-10 upregulated p53 and inhibited α-SMA and Col1α1 expression in LX-2 cells by activating STAT3., Conclusions: hPMSCs promoted CD8 + CD122 + PD-1 + Treg formation and IL-10 secretion but inhibited HSCs activation and α-SMA and Col1α1 expression by CD73, thus controlling the balance of IL-6 and IL-10, and alleviating liver injury in GVHD mice., (Copyright © 2024. Published by Elsevier B.V.)

Unsupervised domain adaptation (UDA) is becoming a prominent solution for the domain-shift problem in many time-series classification tasks. With sequence properties, time-series data contain both local and sequential features, and the domain shift exists in both features. However, conventional UDA methods usually cannot distinguish those two features but mix them into one variable for direct alignment, which harms the performance. To address this problem, we propose a novel virtual-label-based hierarchical domain adaptation (VLH-DA) approach for time-series classification. Specifically, we first slice the original time-series data and introduce virtual labels to represent the type of each slice (called local patterns). With the help of virtual labels, we decompose the end-to-end (i.e., signal to time-series label) time-series task into two parts, i.e., signal sequence to local pattern sequence and local pattern sequence to time-series label. By decomposing the complex time-series UDA task into two simpler subtasks, the local features and sequential features can be aligned separately, making it easier to mitigate distribution discrepancies. Experiments on four public time-series datasets demonstrate that our VLH-DA outperforms all state-of-the-art (SOTA) methods.

Deep neural networks (DNNs) have been widely used in many artificial intelligence (AI) tasks. However, deploying them brings significant challenges due to the huge cost of memory, energy, and computation. To address these challenges, researchers have developed various model compression techniques such as model quantization and model pruning. Recently, there has been a surge in research on compression methods to achieve model efficiency while retaining performance. Furthermore, more and more works focus on customizing the DNN hardware accelerators to better leverage the model compression techniques. In addition to efficiency, preserving security and privacy is critical for deploying DNNs. However, the vast and diverse body of related works can be overwhelming. This inspires us to conduct a comprehensive survey on recent research toward the goal of high-performance, cost-efficient, and safe deployment of DNNs. Our survey first covers the mainstream model compression techniques, such as model quantization, model pruning, knowledge distillation, and optimizations of nonlinear operations. We then introduce recent advances in designing hardware accelerators that can adapt to efficient model compression approaches. In addition, we discuss how homomorphic encryption can be integrated to secure DNN deployment. Finally, we discuss several issues, such as hardware evaluation, generalization, and integration of various compression approaches. Overall, we aim to provide a big picture of efficient DNNs from algorithm to hardware accelerators and security perspectives.

Learning time-series representations when only unlabeled data or few labeled samples are available can be a challenging task. Recently, contrastive self-supervised learning has shown great improvement in extracting useful representations from unlabeled data via contrasting different augmented views of data. In this work, we propose a novel Time-Series representation learning framework via Temporal and Contextual Contrasting (TS-TCC) that learns representations from unlabeled data with contrastive learning. Specifically, we propose time-series-specific weak and strong augmentations and use their views to learn robust temporal relations in the proposed temporal contrasting module, besides learning discriminative representations by our proposed contextual contrasting module. Additionally, we conduct a systematic study of time-series data augmentation selection, which is a key part of contrastive learning. We also extend TS-TCC to the semi-supervised learning settings and propose a Class-Aware TS-TCC (CA-TCC) that benefits from the available few labeled data to further improve representations learned by TS-TCC. Specifically, we leverage the robust pseudo labels produced by TS-TCC to realize a class-aware contrastive loss. Extensive experiments show that the linear evaluation of the features learned by our proposed framework performs comparably with the fully supervised training. Additionally, our framework shows high efficiency in few labeled data and transfer learning scenarios.

Remaining useful life (RUL) prediction is an essential component for prognostics and health management of a system. Due to the powerful ability of nonlinear modeling, deep learning (DL) models have emerged as leading solutions by capturing temporal dependencies within time series sensory data. However, in RUL prediction tasks, data are typically collected from multiple sensors, introducing spatial dependencies in the form of sensor correlations. Existing methods are limited in effectively modeling and capturing the spatial dependencies, restricting their performance to learn representative features for RUL prediction. To overcome the limitations, we propose a novel LOcal-GlObal correlation fusion-based framework (LOGO). Our approach combines both local and global information to model sensor correlations effectively. From a local perspective, we account for local correlations that represent dynamic changes of sensor relationships in local ranges. Simultaneously, from a global perspective, we capture global correlations that depict relatively stable relations between sensors. An adaptive fusion mechanism is proposed to automatically fuse the correlations from different perspectives. Subsequently, we define sequential micrographs for each sample to effectively capture the fused correlations. Graph neural network (GNN) is introduced to capture the spatial dependencies within each micrograph, and the temporal dependencies between these sequential micrographs are then captured. This approach allows us to effectively model and capture the dependency information within the data for accurate RUL prediction. Extensive experiments have been conducted, verifying the effectiveness of our method.

The value of Electrocardiogram (ECG) monitoring in early cardiovascular disease (CVD) detection is undeniable, especially with the aid of intelligent wearable devices. Despite this, the requirement for expert interpretation significantly limits public accessibility, underscoring the need for advanced diagnosis algorithms. Deep learning-based methods represent a leap beyond traditional rule-based algorithms, but they are not without challenges such as small databases, inefficient use of local and global ECG information, high memory requirements for deploying multiple models, and the absence of task-to-task knowledge transfer. In response to these challenges, we propose a multi-resolution model adept at integrating local morphological characteristics and global rhythm patterns seamlessly. We also introduce an innovative ECG continual learning (ECG-CL) approach based on parameter isolation, designed to enhance data usage effectiveness and facilitate inter-task knowledge transfer. Our experiments, conducted on four publicly available databases, provide evidence of our proposed continual learning method's ability to perform incremental learning across domains, classes, and tasks. The outcome showcases our method's capability in extracting pertinent morphological and rhythmic features from ECG segmentation, resulting in a substantial enhancement of classification accuracy. This research not only confirms the potential for developing comprehensive ECG interpretation algorithms based on single-lead ECGs but also fosters progress in intelligent wearable applications. By leveraging advanced diagnosis algorithms, we aspire to increase the accessibility of ECG monitoring, thereby contributing to early CVD detection and ultimately improving healthcare outcomes.

Percutaneous puncture is a common medical procedure that involves accessing an internal organ or tissue through the skin. Image guidance and surgical robots have been increasingly used to assist with percutaneous procedures, but the challenges and benefits of these technologies have not been thoroughly explored. The aims of this systematic review are to furnish an overview of the challenges and benefits of image-guided, surgical robot-assisted percutaneous puncture and to provide evidence on this approach. We searched several electronic databases for studies on image-guided, surgical robot-assisted percutaneous punctures published between January 2018 and December 2022. The final analysis refers to 53 studies in total. The results of this review suggest that image guidance and surgical robots can improve the accuracy and precision of percutaneous procedures, decrease radiation exposure to patients and medical personnel and lower the risk of complications. However, there are many challenges related to the use of these technologies, such as the integration of the robot and operating room, immature robotic perception, and deviation of needle insertion. In conclusion, image-guided, surgical robot-assisted percutaneous puncture offers many potential benefits, but further research is needed to fully understand the challenges and optimize the utilization of these technologies in clinical practice.

Recent studies have demonstrated the success of using the channel state information (CSI) from the WiFi signal to analyze human activities in a fixed and well-controlled environment. Those systems usually degrade when being deployed in new environments. A straightforward solution to solve this limitation is to collect and annotate data samples from different environments with advanced learning strategies. Although workable as reported, those methods are often privacy sensitive because the training algorithms need to access the data from different environments, which may be owned by different organizations. We present a practical method for the WiFi-based privacy-preserving cross-environment human activity recognition (HAR). It collects and shares information from different environments, while maintaining the privacy of individual person being involved. At the core of our approach is the utilization of the Johnson-Lindenstrauss transform, which is theoretically shown to be differentially private. Based on that, we further design an adversarial learning strategy to generate environment-invariant representations for HAR. We demonstrate the effectiveness of the proposed method with different data modalities from two real-life environments. More specifically, on the raw CSI dataset, it shows 2.18% and 1.24% improvements over challenging baselines for two environments, respectively. Moreover, with the discrete wavelet transform features, it further yields 5.71% and 1.55% improvements, respectively.

The past few years have witnessed a remarkable advance in deep learning for EEG-based sleep stage classification (SSC). However, the success of these models is attributed to possessing a massive amount of labeled data for training, limiting their applicability in real-world scenarios. In such scenarios, sleep labs can generate a massive amount of data, but labeling can be expensive and time-consuming. Recently, the self-supervised learning (SSL) paradigm has emerged as one of the most successful techniques to overcome labels' scarcity. In this paper, we evaluate the efficacy of SSL to boost the performance of existing SSC models in the few-labels regime. We conduct a thorough study on three SSC datasets, and we find that fine-tuning the pretrained SSC models with only 5% of labeled data can achieve competitive performance to the supervised training with full labels. Moreover, self-supervised pretraining helps SSC models to be more robust to data imbalance and domain shift problems.

Wandering pattern classification is important for early recognition of cognitive deterioration and other health conditions in people with dementia (PWD). In this paper, we leverage the orientation data available on mobile devices to recognize dementia-related wandering patterns. In particular, we propose to use deep learning (DL) with long short-term memory networks (LSTM) as classifiers for detecting travel patterns including direct, pacing, lapping and random. Experimental results on a real dataset collected from 14 subjects show that deep LSTM classifiers perform better than traditional machine learning (ML) classifiers. Our proposed method can thus be potentially used in healthcare applications for dementia related wandering monitoring and management.Clinical Relevance- This demonstrates the potential of using readily available yet non-privacy information to detect dementia-related wandering patterns with high accuracy.

Human mesenchymal stromal cells (MSCs) harbor immunomodulatory properties to induce the generation of suppressive T cells. MSCs have been successfully used in treating graft-versus-host disease (GVHD) accompanied by abundant inflammatory cytokines such as IL-27. This study investigated the effects of IL-27 on the human placenta-derived MSCs (hPMSCs) to induce generation of CD4 + IL-10 + IFN-γ + T cells in vitro and in the humanized xenogenic GVHD NOD/SCID model. The results showed that the percentages of CD4 + IL-10 + IFN-γ + T cells were significantly increased in activated human PBMC from both healthy donors and GVHD patients with hPMSCs and in the liver and spleen of hPMSC-treated GVHD mice, and the level of CD4 + IL-10 + IFN-γ + T cells in the liver was greater than that in the spleen in hPMSC-treated GVHD mice. The serum level of IL-27 decreased and the symptoms abated in hPMSC-treated GVHD. Further, in vitro results showed that IL-27 promoted the regulatory effects of hPMSCs by enhancing the generation of CD4 + IL-10 + IFN-γ + T cells from activated PBMC. Activation occurred through increases in the expression of programmed death ligand 2 (PDL2) in hPMSCs via the JAK/STAT signaling pathway. These findings indicated that hPMSCs could alleviate GVHD mice symptoms by upregulating the production of CD4 + IL-10 + IFN-γ + T cells in the spleen and liver and downregulating serum levels of IL-27. In turn, the ability of hPMSCs to induce the generation of CD4 + IL-10 + IFN-γ + T cells could be promoted by IL-27 through increases in PDL2 expression in hPMSCs. The results of this study will be of benefit for the application of hPMSCs in clinical trials., (Copyright © 2019 by The American Association of Immunologists, Inc.)

Purpose: To test the safe clinical application of sufentanil as topical ophthalmic drops by examining treated rabbit eyes for ophthalmic irritation signs or short-time toxic reactions., Methods: Twenty-four rabbits were randomly divided into 8 groups (n = 3): The ocular toxicity at 14 d after eye drop ad- ministration was evaluated in groups 1 to 4, and at 30 d post- administration in groups 5 to 8. Groups 1 and 5 were treated with blank vehicle and served as normal controls. The left eyes of rabbits in groups 2 and 6 were exposed to low-dose sufentanil.(5 μg, 2 drops within 5 min), groups 3 and 7 received moderate-dose sufentanil (7.5 μg, 3 drops within 10 min), and groups 4 and 8 received high-dose sufentanil.(10 μg, 4 drops within 15 min). As self-controls, the right eyes of each rabbit were administered an equivalent amount of sodium chloride (9 g/L) at the same drop intervals. At 14 and 30 d after exposure to sufentanil, ophthalmic irritation signs were evaluated and corneas were stained with fluorescein and observed by slit-lamp microscopy. Corneal endothelial counts were performed and toxic reactions were evaluated., Results: Multiple parameters were compared in the control and experimental groups by visual inspection and slit-lamp examination at 14 and 30 d after sufentanil administration. No evidence of irritation signs (including corneal opacity, conjunctival congestion, or edema), eye secretions, iris abnormalities, or temporal eye closure were noted. Corneal en- dothelial cell counts did not significantly differ between the control and experimental groups. Light microscopy revealed no pathological or morphological injury to the cornea, conjunctiva, iris, ciliary body, retina, or optic nerve in either group. The same observation outcomes were noted at 14 and 30 d after administration., Conclusion: Single ocular administration of sufentanil at a dose of 5-10 μg in rabbits yields no ocular irritation or toxic responses at 14 or 30 d following eye drop delivery.