Search

Long short-term memory (LSTM) neural networks and attention mechanism have been widely used in sentiment representation learning and detection of texts. However, most of the existing deep learning models for text sentiment analysis ignore emotion's modulation effect on sentiment feature extraction, and the attention mechanisms of these deep neural network architectures are based on word- or sentence-level abstractions. Ignoring higher level abstractions may pose a negative effect on learning text sentiment features and further degrade sentiment classification performance. To address this issue, in this article, a novel model named AEC-LSTM is proposed for text sentiment detection, which aims to improve the LSTM network by integrating emotional intelligence (EI) and attention mechanism. Specifically, an emotion-enhanced LSTM, named ELSTM, is first devised by utilizing EI to improve the feature learning ability of LSTM networks, which accomplishes its emotion modulation of learning system via the proposed emotion modulator and emotion estimator. In order to better capture various structure patterns in text sequence, ELSTM is further integrated with other operations, including convolution, pooling, and concatenation. Then, topic-level attention mechanism is proposed to adaptively adjust the weight of text hidden representation. With the introduction of EI and attention mechanism, sentiment representation and classification can be more effectively achieved by utilizing sentiment semantic information hidden in text topic and context. Experiments on real-world data sets show that our approach can improve sentiment classification performance effectively and outperform state-of-the-art deep learning-based methods significantly.

Person re-identification (PReID) has received increasing attention due to its significantly importance in intelligent video surveillance. However, most existing multi-scale feature learning methods embed the multi-scale feature extraction modules for PReID, which increases the complexity of inference network and reduces the timeliness. Moreover, jointly using the small-scale and large-scale features to learn feature representations may weaken the local detailed features extraction and spatial information learning. Besides, some attentive local features are often suppressed when introducing the attention mechanisms for deep PReID models. To address these issues, a deep model with Multi-scale Feature Representation Learning (MFRL) and Random Batch Feature Mask (RBFM) is proposed for PReID in this study. To ensure the feature representations discriminability and spatial information learning, two identity losses are adopted to supervise the small-scale and large-scale features learning in MFRL module, respectively. To alleviate the situation of local attentive features being suppressed by using attention mechanisms, RBFM branch with random feature block dropping strategy which can learn the attentive local feature representations. The proposed methods are only performed in the training phase and discarded in the testing phase, thus, enhancing the effectiveness of the model. Our model achieves the state-of-the-art on the popular benchmark datasets including Market-1501, DukeMTMC-reID and CUHK03. Besides, we conduct a set of ablation experiments to verify the effectiveness of the proposed methods.

In the real-world applications of machine learning and cybernetics, the data with imbalanced distribution of classes or skewed class proportions is very pervasive. When dealing with imbalanced data, traditional classification approaches might fail to learn a good classifier. In the phase of learning, these algorithms are greatly impacted by the skewed distribution of data. Consequently, the performance of classification drops drastically. In this study, we propose a novel two-in-one algorithm for classifying the imbalanced data by integrating metric learning and ensemble learning algorithms. Firstly, we design a new metric learning algorithm for imbalanced data, which is called Large Margin Nearest Neighbors Balance (called LMNNB). This method can minimize the distance between one sample and its similar neighbors which belong to the same class, and maximize the distance from its dissimilar neighbors which belong to different classes as well. Essentially, this beneficial effect can also be achieved even if the distribution of data is imbalanced. Through metric learning, the imbalance data can be used to learn a better classifier. Secondly, we propose an ensemble learning algorithm to further improve the performance of classification. This method combines multiple sub-classifiers and makes decisions by applying a soft voting strategy. Extensive experiments are conducted on real benchmark imbalanced datasets to demonstrate the effectiveness of LMNNB with ensemble algorithm (called LMNNB-E) in several evaluation measurements. The results show that LMNNB and LMNNB-E outperform the state-of-the-art methods in classifying imbalance data.

It is key to identify both sentiment and topic for well understanding and managing social media data such as online reviews and microblogs. This paper studies a robust and reliable solution for synchronous analysis of sentiment and topic in online reviews. Specifically, a probabilistic model is proposed for joint sentiment topic detection with multi-granular computation, named MgJST (multi-granular joint sentiment topic). The MgJST model introduces sentence level structural knowledge to detect sentiment and topic simultaneously from reviews based on latent Dirichlet allocation (LDA). The sets of experiments are conducted on seven sentiment analysis datasets. Experimental results demonstrate that the proposed model significantly outperforms state-of-the-art unsupervised approaches WSTM and STSM in terms of sentiment detection quality, and has powerful ability to extract topics from reviews.

Myeloproliferative neoplasms (MPNs) are characterized by upregulation of proinflammatory cytokines and immune dysregulation, which provide a reasonable basis for immunotherapy in patients. Megakaryocytes are crucial in the pathogenesis of primary myelofibrosis (PMF), the most clinically aggressive subtype of MPN. In this study, we aimed to explore PD‐L1 (programmed death‐ligand 1) expression in megakaryocytes and its clinical implications in PMF. We analyzed PD‐L1 expression on megakaryocytes in PMF patients by immunohistochemistry and correlated the results with clinicopathological features and molecular aberrations. We employed a two‐tier grading system considering both the proportion of cells positively stained and the intensity of staining. Among the 85 PMF patients, 41 (48%) showed positive PD‐L1 expression on megakaryocytes with the immune‐reactive score ranging from 1 to 12. PD‐L1 expression correlated closely with higher white blood cell count (p = 0.045), overt myelofibrosis (p = 0.010), JAK2V617F mutation (p = 0.011), and high‐molecular risk mutations (p = 0.045), leading to less favorable overall survival in these patients (hazard ratio 0.341, 95% CI 0.135–0.863, p = 0.023). Our study provides unique insights into the interaction between immunologic and molecular phenotypes in PMF patients. Future work to explore the translational potential of PD‐L1 in the clinical setting is needed.

Studies have reported a positive association between hepatitis B surface antigen (HBsAg)-positive hepatitis B virus (HBV) infection and follicular lymphoma (FL). Nevertheless, clinical information concerning chronic HBV infection in FL is sparse. This retrospective cohort study investigated the prognostic impact of HBsAg in immunocompetent patients with FL treated with frontline rituximab-containing chemoimmunotherapy in an HBV-endemic area between 2006 and 2016. Among the 149 analyzed patients, 32 (21.5%) were HBsAg-positive. HBsAg positivity was positively associated with symptomatic splenomegaly, significant serous effusions, and peritreatment hepatic dysfunction. HBsAg-positive patients had a trend of lower complete remission rate (59.4% vs. 76.9%, P = 0.07), significantly poorer overall survival (hazard ratio for death, 2.68; 95% confidence interval, 1.21–5.92), and shorter progression-free survival than had HBsAg-negative patients. Multivariate analysis revealed that HBsAg is an independent adverse prognostic factor for overall survival. Intriguingly, HBsAg-positive patients had a higher incidence of progression of disease within 24 months (POD24) than had HBsAg-negative patients (cumulative incidence rate, 25.8% vs. 12.4%, P = 0.045). This study revealed that patients with FL and chronic HBV infection represent a distinct subgroup with a markedly poor prognosis. HBsAg was positively associated with POD24 and might serve as a new prognostic predictor of the survival of FL patients in endemic regions for HBV infection.

Early recognition of adult-onset immunodeficiency associated with neutralizing anti-interferon gamma autoantibodies (anti-IFNγ Abs) remains difficult, and misdiagnoses have been reported. Although febrile lymphadenopathy is among the most common initial manifestations of this disorder, no comprehensive clinicopathologic analysis of lymphadenopathy in patients with anti-IFNγ Abs has been reported. Here, we describe 26 lymph node biopsy specimens from 16 patients. All patients exhibited concurrent disseminated nontuberculous mycobacterial infections, and 31% received a tentative diagnosis of lymphoma at initial presentation. We found 3 distinct histomorphologic patterns: well-formed granuloma (46%), suppurative inflammation or loose histiocytic aggregates (31%), and lymphoproliferative disorder (LPD, 23%). The latter shared some of the features of malignant T-cell lymphoma, IgG4-related disease, and multicentric Castleman disease. Half of the specimens with LPD had monoclonal T cells, and 33.3% were indistinguishable from angioimmunoblastic T-cell lymphoma as per current diagnostic criteria. All lymphadenopathy with LPD features regressed with antibiotics without administration of cytotoxic chemotherapy or immunotherapy. The median follow-up time was 4.3 years. Our study highlights the substantial challenge of distinguishing between lymphoma and other benign lymphadenopathy in the setting of neutralizing anti-IFNγ Abs. Increased vigilance and multidisciplinary discussion among clinicians and pathologists are required to achieve the most appropriate diagnosis and management.

BACKGROUND: CpG island methylator phenotype (CIMP) represents a carcinogenesis pathway of colorectal cancer (CRC) and the association between CIMP CRC, molecular features and risk factors in East Asian population is less studied. METHODS: We prospectively enrolled newly diagnosed CRC patients at the National Taiwan University Hospital. Clinicopathological data and risk factors for CRC were collected during interview. The tumour samples were subjected to CIMP, RAS/BRAF mutation and microsatellite instability tests. CIMP-high was determined when ≧3 methylated loci of p16, MINT1, MINT2, MINT31 and MLH1 were identified. Multivariate logistic regression was used to evaluate the association between risk factors and CIMP-high CRC. RESULTS: Compared with CIMP-low/negative CRC, CIMP-high CRC was associated with more stage IV disease, BRAF V600E mutation and high body mass index (BMI ≧ 27.5 kg/m(2)) in younger patients (age

With the development of location‐based services and Big data technology, vehicle map matching techniques are growing rapidly, which is the fundamental techniques in the study of exploring global positioning system (GPS) data. The pre‐processed GPS data can provide the guarantee of high‐quality data for the research of mining passenger’s points of interest and urban computing services. The existing surveys mainly focus on map‐matching algorithms, but there are few descriptions on the key phases of the acquisition of sampling data, floating car and road data preprocessing in vehicle map matching systems. To address these limitations, the contribution of this survey on map matching techniques lies in the following aspects: (i) the background knowledge, function and system framework of vehicle map matching techniques; (ii) description of floating car data and road network structure to understand the detailed phase of map matching; (iii) data preprocessing rules, specific methodologies, and significance of floating car and road data; (iv) map matching algorithms are classified by the sampling frequency and data information. The authors give the introduction of open‐source GPS sampling data sets, and the evaluation measurements of map‐matching approaches; (v) the suggestions on data preprocessing and map matching algorithms in the future work.

Background: 188Re-liposome has been used for evaluating the theranostic efficacy on human head and neck squamous cell carcinoma (HNSCC) at preclinical stages. Here we furthercompared the microRNA expressive profile in orthtopic HNSCC tumor model exposed to 188Re-liposome. Methods: A single dose or dual doses of 188Re-liposome was intravenously injected into tumor-bearing mice followed by the Cerenkov luminescent imaging (CLI) for monitoring the accumulation of 188Re-liposome in tumors. The microRNA expressive profile was generated using the Taqman® OpenArray® Human MicroRNA Panel followed by the DIANA mirPath analysis, KEGG signaling pathways prediction, and Kaplan-Meier survival analysis for predicting the prognostic role of 188Re-liposome affected microRNAs. Results: Dual doses of 188Re-liposome exhibited a better tumor suppression than a single dose of 188Re-liposome, including reduced tumor size, Ki-67 proliferative marker, and epithelial-mesenchymal transition (EMT) related factors. The microRNA expressive profiles showed that 22 microRNAs and 19 microRNAs were up-regulated and down-regulated by dual doses of 188Re-liposome, respectively. Concomitantly, these two groups of microRNAs were inversely regulated by a single dose of 188Re-liposome accordingly. These microRNAs influenced most downstream genes involved in cancer related signaling pathways. Further, miR-520e and miR-522-3p were down-regulated whereas miR-186-5p and miR-543 were up-regulated by dual doses of 188Re-liposome, and they separately affected most of genes involved in their corresponding pathways with high significance. Additionally, high expressions of miR-520e and miR-522-3p were associated with lower survival rate of HNSCC patients. Conclusion: MicroRNA expression could be used to evaluate the therapeutic efficacy and regarded prognostic factors using different doses of 188Re-liposome.

A robust morphomolecular classification system for gastric carcinoma is required. A 4-tier morphologic classification is proposed, including diffuse, intestinal, tubular, and lymphoid types. A tissue microarray for mismatch repair immunohistochemistry and Epstein-Barr virus (EBV) in situ hybridization were performed in 329 gastric carcinomas. DNA flow cytometry was used to detect aneuploidy in formalin-fixed paraffin-embedded samples. Lymphoid histology was the third most common histologic pattern at our institute and strongly associated with EBV infection and PMS2/MLH1-deficiency (both P

Intralymphatic spread is a rare finding and is associated with poor prognosis in diffuse large B cell lymphoma (DLBCL). Here, we report a case of relapsed DLBCL with mucosal intralymphatic spread. A 69-year-old man had been diagnosed with gastric DLBCL stage IIE at 57 years. He had a relapse with generalized lymphadenopathy and extranodal involvement at 61 years; then, second complete remission was achieved after salvage chemotherapy. He then had a second relapse with involvement of the terminal ileum, spinal cord, and left tonsil. The terminal ileum showed intralymphatic spread in the lamina propria of the intestinal villi, which was confirmed by D2–40 immunostaining. Eleven months later, another biopsy showed prominent intralymphatic spread in the mucosa of the terminal ileum. After salvage therapies, the spinal cord and tonsillar tumors resolved, but the intestinal tumors were refractory. The patient eventually died of progressive disease. In contrast to previously reported cases, the involved lymphatic vessels were observed in the mucosa, and the lymphoma cells expressed CD5 in the first colonoscopic biopsy. This rare case broadens the spectrum of intralymphatic spread in DLBCL.

As the cellular wireless communication techniques grow rapidly, the cells become smaller than the traditional communication system, then the handover events are very frequent and need to support a large number of users, and handover detection has become a very active research direction in a mobile computing environment. In order to copy with the problem of frequent handover operations between base stations in current cellular communication networks as cybernetic systems, we propose a novel handover detection approach by integrating frequent trajectory patterns mining and location prediction techniques. The proposed model contains the following essential steps: (1) mining frequent trajectory patterns from large-scale historical trajectory databases by applying an improved Apriori-like rule-based machine learning algorithm, which finds the intersection of candidate items by applying the trajectory connection operation instead of calculating the support count of each trajectory patterns and the candidate items are considerably reduced; (2) discovering movement rules based on the frequent trajectory pattern set by finding the postfix items of given prefix items satisfying the minimum support threshold; (3) inferring the future locations of moving objects by applying the movement rules matching strategy; (4) determining whether or not to perform handover detection across base stations in a cellular network beyond the discovered prediction results, according to the coverage area of cellular networks. Extensive experiments were conducted on the mobile datasets and the experimental results demonstrate the advantages of the proposed algorithm from the following four aspects: (1) the accuracy of handover detection is above 95% at average which is a very satisfactory result in a mobile computing environment; (2) the time cost is less than 20 s when the number of movement rules and handover detection is 1000, which further shows the merit of the runtime performance of the proposed method; (3) the frequent-trajectory-patterns based handover detection algorithm can successfully avoid the ping-pong effect due to unnecessary handover operations; (4) and lastly significantly reduce the error rate of frequent handover decisions and the average unnecessary handover rate is lower than 0.05 when compared with the state-of-the-art methods.

Most existing unsupervised approaches to detect topic sentiment in social texts consider only the text sequences in corpus and put aside social dynamics, as leads to algorithm’s disability to discover true sentiment of social users. To address the issue, a probabilistic graphical model LDTSM (Long-term Dependence Topic-Sentiment Mixture) is proposed, which introduces dependency distance and uses the dynamics of social media to achieve the perfect combination of inheriting historical topic sentiment and fitting topic sentiment distribution underlying in current social texts. Extensive experiments on real-world SinaWeibo datasets show that LDTSM significantly outperforms JST, TUS-LDA and dNJST in terms of sentiment classification accuracy, with better inference convergence, and topic and sentiment evolution analysis results demonstrate that our approach is promising.

Person re-identification (PReID), which aims to re-identity a pedestrian from multiple non-overlapping cameras, has been significantly improved by deep learning system. There exist two popular deep frameworks used for PReID, i.e., identification and triplet models. Since these two frameworks have different loss functions, they have their own advantages and disadvantages. To combine the both advantages of two frameworks, in this paper, we propose using the triplet and Online Instance Matching (OIM) losses to train the carefully designed network. Given a triplet input images, the combined model can output the identities of the input images and learn a corresponding similarity measurement simultaneously. Experiments on CUHK01, CUHK03, Market-1501, and DukeMTMC-reID datasets demonstrate that the proposed model outperforms the compared state-of-the-art methods in most cases.

In recent years, person re-identification (ReID) has received much attention since it is a fundamental task in intelligent surveillance systems and has widespread application prospects in numerous fields. Given an image of a pedestrian captured from one camera, the task is to identify this pedestrian from the gallery set captured by other multiple cameras. It is a challenging issue since the appearance of a pedestrian may suffer great changes across different cameras. The task has been greatly boosted by deep learning technology. There are mainly six types of deep learning-based methods designed for this issue, i.e. identification deep model, verification deep model, distance metric-based deep model, part-based deep model, video-based deep model and data augmentation-based deep model. In this paper, we first give a comprehensive review of current six types of deep learning methods. Second, we present the detailed descriptions of existing person ReID datasets. Then, some state-of-the-art performances of methods over recent years on several representative ReID datasets are summarized. Finally, we conclude this paper and discuss the future directions of the person ReID.

Person re-identification (PReID) has received increasing attention due to it being an important role in intelligent surveillance. Many state-of-the-art PReID methods are part-based deep models. Most of these models focus on learning the part feature representation of a person's body from the horizontal direction. However, the feature representation of the body from the vertical direction is usually ignored. In addition, the relationships between these part features and different feature channels are not considered. In this paper, we introduce a multi-branch deep model for PReID. Specifically, the model consists of five branches. Among the five branches, two branches learn the part features with spatial information from horizontal and vertical orientations; one branch aims to learn the interdependencies between different feature channels generated by the last convolution layer of the backbone network; the remaining two branches are identification and triplet sub-networks in which the discriminative global feature and a corresponding measurement can be learned simultaneously. All five branches can improve the quality of representation learning. We conduct extensive comparison experiments on three benchmarks, including Market-1501, CUHK03, and DukeMTMC-reID. The proposed deep framework outperforms other competitive state-of-the-art methods. The code is available at https://github.com/caojunying/person-reidentification.

The purpose of Person re-identification (PReID) is to identify the same individual from the non-overlapping cameras, the task has been greatly promoted by the deep learning system. In this study, we review two widely-used CNN frameworks in the PReID community: identification model and triplet model. We provide a comprehensive overview of the advantages and limitations of the two models and present a hybrid model that combines the advantages of both identification and triplet models. Specifically, the proposed model employs triplet loss, identification loss and center loss to simultaneously train the carefully designed network. Furthermore, the dropout scheme is adopted by its identification subnetwork. Given a triplet unit images, the model can output the identities of the three input images and force the Euclidean distance between the mismatched pairs to be larger than those between the matched pairs as well as reduce the variance of the same class at the same time. Extensive comparative experiments on three PReID benchmark datasets (CUHK01, CUHK03, Market-1501) show that our proposed architecture outperforms many state of the art methods in most cases.

Volatile anesthetics, including isoflurane, have been reported to have negative effects on cognitive dysfunction characterized by cognitive deficits following anesthesia. The aim of the current study was to investigate the effects involved with disintegrin and metallopeptidase domain 2 (ADAM2) silencing on isoflurane-induced cognitive dysfunction via the P13 K/Akt signaling pathway in immature rats. One week old healthy Sprague-Dawley (SD) rats were recruited and administered isoflurane anesthesia. The rats were then subjected to shADAM2 or wortmannin (PI3K/Akt signaling pathway inhibitor) to identify the effects of ADAM2 and the PI3K/Akt signaling pathway on the cognitive function of rats. Morris water maze and passive-avoidance tests were performed to examine the cognitive function of the rats. TUNEL staining was conducted to detect neuronal apoptosis in the hippocampal CA1 region. The obtained experimental results demonstrated that isoflurane anesthesia led to increased escape latency, reaction time, number of errors and TUNEL-positive neurons, along with a decreased latency time. In response to treatment with shADAM2, escape latency, reaction time, number of errors and TUNEL-positive cells were all noted to have decreased, in addition to elevated latency time, while contrasting trends were observed in regard to treatment with wortmannin. Taken together, the key findings of the present study revealed that shADAM2 activated the PI3K/Akt signaling pathway, resulting in elevated expressions of PI3K and Akt. Our study ultimately identified that ADAM2 silencing alleviates isoflurane-induced cognitive dysfunction by activating the P13 K/Akt signaling pathway in immature rats.

DNA-binding proteins (DBPs) have an important role in various regulatory tasks. In recent years, with developing of deep learning, many fields like natural language processing, computer vision and so on have achieve great success. Some great model, for example DeepBind, brought deep learning to motif discovery and also achieve great success in predicting DNA-transcription factor binding, aka motif discovery. But these methods required integrating multiple features with raw DNA sequences such as secondary structure and their performances could be further improved. In this paper, we propose an efficient and simple neural network-based architecture, DBPCNN, integrating conservation scores and epigenomic data to raw DNA sequences for predicting in-vitro DNA protein binding sequence. We show that conservation scores and epigenomic data for raw DNA sequences can significantly improve the overall performance of the proposed model. Moreover, the automatic extraction of the DBA-binding proteins can enhance our understanding of the binding specificities of DBPs. We verify the effectiveness of our model on 20 motif datasets from in-vitro protein binding microarray data. More specifically, the average area under the receiver operator curve (AUC) was improved by 0.58% for conservation scores, 1.29% for MeDIP-seq, 1.20% for histone modifications respectively, and 2.19% for conservation scores, MeDIP-seq and histone modifications together. And the mean average precision (AP) was increased by 0.62% for conservation scores, 1.46% for MeDIP-seq, 1.27% for histone modifications respectively, and 2.29% for conservation scores, MeDIP-seq and histone modifications together.

Plant leaf recognition is a computer vision task used to identify plant species. To address the problem that current plant leaf recognition algorithms have difficulty in recognizing fine-grained leaf classification between classes, this paper proposes a DMSNet (Deep Multi-Scale Network) model, a plant leaf classification algorithm based on multi-scale feature extraction. In order to improve the extraction ability of different fine-grained features of the model, the model is improved on the basis of Multi-scale Backbone Architecture model. In order to achieve better plant leaf classification, a visual attention mechanism module to DMSNet is added and ADMSNet (Attention-based Deep Multi-Scale Network), which makes the model focus more on the plant leaf itself, is proposed, essential features are enhanced, and useless features are suppressed. Experiments on real datasets show that the classification accuracy of the DMSNet model reaches 96.43%. In comparison, the accuracy of ADMSNet with the addition of the attention module reaches 97.39%, and the comparison experiments with ResNet-50, ResNext, Res2Net-50 and Res2Net-101 models on the same dataset show that DMSNet improved the accuracy by 4.6%, 18.57%, 3.72% and 3.84%, respectively. The experimental results confirm that the DMSNet and ADMSNet plant leaf recognition models constructed in this paper can accurately recognize plant leaves and have better performance than the traditional models.

Infrared-Visible person re-identification is a kind of cross-modality person re-identification. The purpose of the task is that given a person image we need to find another image on the same person from gallery. The query images and gallery images are not only in RGB modality but in infrared modality as well. The cross-modality person ReID task can deal with the limitation of single modality because we usually can get images in more than one modality. In our work, we take advantage of both global feature and local feature. We use a dual-path structure to extract features from RGB images and infrared images respectively. Besides, we add the LSTM structure in each path to learn the serialized local features. The loss function consists of cross-entropy loss and hetero-center loss so that the model can bridge the cross-modality and intra-modality gaps to capture the modality-shared features and improve the cross-modality similarity. Finally, we do experiments on two datasets including SYSU-MM01 and RegDB, then compare with other methods in recent studies.

Plant classification and recognition research is the basic research work of botany research and agricultural production. It is of great significance to identify and distinguish plant species and explore the relationship between plants. In recent years, most of the research methods focus on feature extraction and feature engineering related aspects. In this paper, a plant leaf recognition method based on fine-grained image classification is proposed, which can better find the regional block information of different species of plant leaves. In this study, the hierarchical and progressive training strategy is adopted, the method of cutting and generating jigsaw is used to force the model to find information of different granularity levels. The experiment proves that the model trained by the fine-grained classification method can better solve the problems of large intra-class spacing and small inter-class spacing of plant slices.

The various presentations of osseous Langerhans cell histiocytosis (LCH) make it difficult to distinguish from other bone diseases. In addition, there is no universally accepted protocol for managing osseous LCH for single non-central nervous system-risk lesions. Here, the rare cases of two paediatric patients, aged 1 and 2 years, who presented with a solitary tibial lesion at time of LCH diagnosis, are reported. One patient progressed to multiple lesions after curettage of the original lesion. Subsequently, both patients received preventive chemotherapy using the Taiwan Paediatric Oncology Group (TPOG) revised protocol for treating low risk patients with LCH, namely, TPOG LCH2002-LR. After receiving this treatment, which included a schedule of prednisolone and vincristine for 6 weeks, followed by prednisolone, vincristine and 6-mercaptopurine for a further 48 weeks, both patients are free from recurrence or progression.

Objective: The primary objective of this study was to investigate the prevalence of intensive care unit (ICU) pediatric delirium in Shanghai, China. Secondary objectives were to determine the association of hypoxia and infection with ICU pediatric delirium, and the impact between different age. Design: Prospective single-center observational study. Setting: Two pediatric intensive care unit (PICU) within a tertiary-A general hospital. Patients: Patients age between 1 month to 7 years in PICU who stayed at least 1 day were included. Convenance sampling was used. Interventions: None. Measurements and Main Results: Pediatric patients (n=639) were screened twice a day for the prevalence of ICU pediatric delirium by Cornell Assessment of Pediatric Delirium, 300 (46.95%) of them had infection and 213 (33.33%) had hypoxia in PICU. Children who suffered hypoxia remained more than three times likely to be delirious during their hospitalization compared with children who were not hypoxia, after controlling other covariates, the odds of pediatric delirium for subjects with hypoxia was 3.26 times (95% CI, 1.98-5.38) the odds without hypoxia. Also, the odds of pediatric delirium for subjects with infection was 2.55 times (95% CI, 1.58-4.11) the odds without infection adjusting for other covariates. After subgrouping by age, the occurrence of ICU pediatric delirium with infection for children younger than two years old was 5.37 times (95% CI, 3.09-9.33) compared with patients who were never infection, while that for the children equal to or older than two years old was no statistically significant relationship. Conclusions: The prevalence of ICU pediatric delirium was 31.30%, while there is an independent association of infection and hypoxia with ICU pediatric delirium. Furthermore, children younger than two years old took more risks on pediatric delirium when they were infected in this study, while there was no relationship between infection and pediatric delirium who aged 2 years or older.

Primary pulmonary lymphoma is an uncommon disease, and extranodal marginal zone lymphoma of mucosa‐associated lymphoid tissue (MALT) is the most common type of pulmonary lymphoma. The most frequent pattern observed in chest computed tomography (CT) is consolidation, followed by nodules and mass. The differentiation of pulmonary MALT lymphoma from other lung diseases is critical for disease management and treatment. However, pulmonary MALT lymphoma with isolated endobronchial manifestation has seldomly been reported. Here, we report a case of an elderly woman who presented with a four‐month history of cough, dyspnoea, and haemoptysis. Chest CT scan revealed left main bronchus narrowing without lung parenchymal lesion. Bronchoscopic examination was performed, and the diagnosis of primary pulmonary MALT lymphoma with isolated endobronchial involvement was made. She has been successfully treated with rituximab., We describe a case of primary pulmonary mucosa‐associated lymphoid tissue (MALT) lymphoma presented as an endobronchial tumor without lung parenchymal involvement.

With the development of network services and location-based systems, many mobile applications begin to use users’ geographical location to provide better services. In terms of social networks, geographical location is actively shared by users. In some applications with recommendation services, before the geographical location recommendation is provided, the authors have to obtain user's permission. This kind of social network integrated with geographical location information is called location-based social networks (abbreviate for LBSNs). In the LBSN, each user has location information when he or she checked in hotels or feature spots. Based on this information, they can identify user's trajectory of movement behaviour and activity patterns. In general, if there is friendship between two users, their trajectories in reality are likely to be similar. In this study, according to user's geographical location information over a period of time, they explore whether there exists friendly relationship between two users based on trajectory similarity and the structure theory of graphs. In particular, they propose a new factor function and a factor graph model based on user's geographical location to predict the friendship between two users in the real LBSN.

Identifying transcription factor (TF) binding sites (TFBSs) has play an important role in the computational inference of gene regulation. With the development of high-throughput technologies, there have been many conventional methods and deep learning models used in the identification of TFBSs. However, most methods are designed to predict TFBSs only based on raw DNA sequence leads to low accuracy. Therefore, we propose a Dual-channel Convolutional neural network (CNN) model combining DNA sequences and DNA Shape features to predict TFBSs, named DCDS. In the DCDS model, the convolution layer captures low-level features from input data and parallel pooling operations are used to find the most significant activation signal in a sequence for each filter to improve the prediction accuracy of TFBSs. We conduct a series of experiments on 66 in vitro datasets and experimental results show that proposed model DCDS is superior to some state-of-the-art methods.

Person re-identification, as the basic task of a multi-camera surveillance system, plays an important role in a variety of surveillance applications. However, the current mainstream person re-identification model based on deep learning requires a lot of labeled data, which takes a lot of time and manpower. In this study, we proposed a person re-identification method based on random occlusion recovery with noise channel. We add random occlusion blocks to the original image, use the GAN model for repair, and use the repaired image to expand the original training set. After that, the generated image is adjusted through the noise channel. Finally, we use the enhanced data set to train the baseline model. Our model achieves the state-of-the-art on Market-1501 dataset, proving that the method is effective.

Plant image recognition is an important thing for protecting plants, protecting the environment, and protecting nature. Recently, most models in the field of plant leaf recognition make classification after extracting global features. In this paper, we propose a plant leaf recognition model based on metric learning. Metric learning calculates the similarity of the extracted feature vectors to obtain the distance between different sample features, so as to determine whether similar pictures belong to the same category, and then achieve the classification effect. In this study, feature triplet are used for metric learning, and the loss function we used is triplet-loss.

Classification of functional genomic regions (such as promoters or enhancers) based on sequence data alone is a very important problem. Various data mining algorithms can be used well to apply to predict the promoter region. For example, association and clustering algorithms like Classification And Regression Tree (CART), machine learning algorithms like Simple Logistic, BayesNet, Random forest, or the most popular deep learning like Recurrent Neural Network (RNN), Convolutional Neural Networks (CNN). However, due to large amount of genetic data are unlabeled, these methods cannot directly solve this challenge. Therefore, we present a three-layer dynamic transfer learning language model (TLDTLL) for E. coli promoter classification problems. TLDTLL is an effective algorithm for inductive transfer learning that utilizes pre-training on large unlabeled genomic corpuses. This is particularly advantageous in the context of genomics data, which tends to contain significant volumes of unlabeled data. TLDTLL shows improved results over existing methods for classification of E. coli promoters using only sequence data.

With the rapid development of deep learning, convolution neural network achieve great success in predicting DNA-transcription factor binding, aka motif discovery, In this paper, we propose a novel neural network based architecture i.e. eDeepCNN, combining multi-layer convolution network and embedding layer for predicting in-vitro DNA protein binding sequence. Our model fully utilize fitting capacity of deep convolution neural network and is well designed to capture the interaction pattern between motifs in neighboring sequence. Meanwhile continuous embedding vector serves as a better description of nucleotides than one-hot image-like representation owing to its superior expressive ability. We verify the effectiveness of our model on 20 motif datasets from in-vitro protein binding microarray data (PBMs) and present promising results compared with well-established DeepBind model. In addition, we emphasis the significance of dropout strategy in our model to fight against the overfitting problem along with growing model complexity.

Infrared-Visible person re-identification is a challenging and fundamental task of associating the same person across visible and thermal cameras. Most of the studies focus on improving the global features to address the cross-modality issue, thus, some discriminative local and salient features are ignored by the deep models. A novel deep architecture named Dual-path Local Information Structure (DLIS) with Position Attention-guided Learning Module (PALM) is proposed to address the cross-modality issue for Infrared-Visible PReID task. The DLIS has two individual branches which contains a visible stream and an infrared stream to extract modality sharable features. The PALM can capture long-range dependencies and enhance the discriminative local feature representations to form the final feature descriptors. To supervise the network extracting discriminative features to shrink the margin of different modalities, the proposed model is conducted the joint supervision of cross-entropy loss function and hetero-center loss function. Compared with the recent studies, the proposed methods achieve the state-of-the-art on the two benchmark datasets including SYSU-MM01 and RegDB dataset.

GPS data often have position deviations in precision and are apt to be affected by noise; hence, it is essential to extract features from trajectories before performing large-scale data mining. A GeoHash-based spatial coding technique called GeoHashTree was used to index spatiotemporal trajectory points in order to improve the efficiency of nearest-neighbor search. The GeoHashTree was applied in trajectory clustering and an improved density-based clustering algorithm was proposed to reduce the time complexity of nearest-neighbor search from $O(n^2)$ to $O(n\\log{n})$. After extracting trajectory points with changing angles, the proposed clustering approach was employed to achieve deep-level feature extraction on trajectory points with changing angles, which aims to accurately identify feature points. Extensive experiments are conducted on real GPS data and the results demonstrate that the proposed trajectory-clustering algorithm based on the GeoHashTree spatial index structure can improve time performance by an average of 90.89% as well as guarantee the accuracy of clustering compared with the traditional clustering method. The visualization results show that the trajectory feature extraction approach can effectively find trajectory points with changing angles and discover a varying types of feature points from large-scale data sets. In addition, the proposed approach does not depend on road network data and can dynamically update with new incoming trajectory data as road networks change in real time.

Objective To explore the influence of applying educational animated film as continuous post-discharge rehabilitation guidance for patients after total hip arthroplasty (THA). Methods Sixty patients discharged after THA were randomly divided into two groups. Traditional methods, such as distributing manuals of rehabilitation guidance on THA and phone call follow-ups, were adopted in the control group, whereas educational animated film was used as continuous rehabilitation guidance after discharge in the experimental group. Differences in recovery of hip joint function, accuracy of functional exercise, mastery of rehabilitation knowledge and quality of life between the two groups were compared. Results The experimental group had superior performance on all indicators compared to the control group ( P Conclusions The use of educational animated film as continuous post-discharge rehabilitation guidance in patients after THA achieved better effects than traditional education methods.

Background Patients with diffuse large B-cell lymphoma (DLBCL) with concurrent hepatitis B surface antigen (HBsAg)-positive hepatitis B virus (HBV) infection have distinct clinical features. Nevertheless, the prognostic value of HBsAg in DLBCL in the rituximab era remains unclear. Materials and Methods We conducted a retrospective cohort study to investigate the clinical relevance of HBsAg in immunocompetent patients with DLBCL treated with homogeneous rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone between 2002 and 2016. Results Among 416 analyzed patients, 98 (23.6%) were HBsAg positive. HBsAg positivity was associated with a younger age and more advanced stage at diagnosis, more frequent hepatic impairment during perichemotherapy, and a trend of higher National Comprehensive Cancer Network-International Prognostic Index (NCCN-IPI) score at diagnosis. Compared with the HBsAg-negative patients, the HBsAg-positive patients had a lower overall response rate (76.5% vs. 85.5%, p = .043), poorer 5-year overall survival (OS) rate (57.2% vs. 73.5%, p < .001), and shorter 5-year progression-free survival (PFS) rate (47.2% vs. 60.7%, p = .013). Multivariate analyses showed that HBsAg positivity was an independent unfavorable prognostic indicator for OS and PFS. A scoring system incorporating HBsAg positivity, the NCCN-IPI score, and serum albumin levels proved to be useful for stratifying prognostically relevant subgroups of patients with DLBCL. Conclusion This study demonstrated that HBV infection is uniquely relevant to DLBCL. HBsAg might serve as a novel biomarker to improve clinical risk stratification of patients with DLBCL in areas with high prevalence of HBV infection. Further research investigating the etiopathogenesis of HBV infection in DLBCL is imperative. Implications for Practice A considerable disparity exists regarding the prognostic relevance of hepatitis B surface antigen (HBsAg)-positive hepatitis B virus (HBV) infection in patients with diffuse large B-cell lymphoma (DLBCL). In this large, retrospective cohort study from an area with high prevalence of HBV infection, the authors demonstrated that HBsAg was an independent unfavorable factor significantly associated with survival, highlighting its potential as a novel prognostic indicator to improve the risk stratification of patients with DLBCL in the rituximab era.

Discovering DNA-protein binding sites, also known as motif discovery, is the foundation for further analyses of transcription factors (TFs). Deep learning algorithms such as convolutional neural networks (CNN) and recurrent neural networks (RNN) are introduced to motif discovery task and have achieved state-of–art performance. However, these methods still have limitations such as neglecting the context information in large-scale sequencing data. Thus, inspired by the similarity between DNA sequence and human language, in this paper we propose a hierarchical attention network for predicting DNA-protein binding sites which is based on a natural language processing method for document classification. The proposed method is tested on real ChIP-seq datasets and the experimental results show a considerable improvement compared with two well-tested deep learning-based sequence model, DeepBind and Deepsea.

With the rapid development of deep learning, some discriminative motif discovery methods based on deep neural network are gradually becoming the mainstream, which also bringing huge improvement of prediction accuracy. In this paper, we propose a convolutional neural network based architecture (eCNN), combining embedding layer with GloVe. Firstly, eCNN divides each single sequence of ChIP-seq datasets into multiple subsequences called k-mers by a sliding window, and then encoding k-mers into a relatively low dimension vectors by GloVe, and finally scores each vector using multiple convolutional networks. The experiment shows that our architecture can get good results on the task of motif discovery.