DLformer: A Dynamic Length Transformer-Based Network for Efficient Feature Representation in Remaining Useful Life Prediction

Representation learning-based remaining useful life (RUL) prediction plays a crucial role in improving the security and reducing the maintenance cost of complex systems. Despite the superior performance, the high computational cost of deep networks hinders deploying the models on low-compute platforms. A significant reason for the high cost is the computation of representing long sequences. In contrast to most RUL prediction methods that learn features of the same sequence length, we consider that each time series has its characteristics and the sequence length should be adjusted adaptively. Our motivation is that an "easy" sample with representative characteristics can be correctly predicted even when short feature representation is provided, while "hard" samples need complete feature representation. Therefore, we focus on sequence length and propose a dynamic length transformer (DLformer) that can adaptively learn sequence representation of different lengths. Then, a feature reuse mechanism is developed to utilize previously learned features to reduce redundant computation. Finally, in order to achieve dynamic feature representation, a particular confidence strategy is designed to calculate the confidence level for the prediction results. Regarding interpretability, the dynamic architecture can help human understand which part of the model is activated. Experiments on multiple datasets show that DLformer can increase up to 90% inference speed, with less than 5% degradation in model accuracy.

Fine-Grained Recognition With Learnable Semantic Data Augmentation

Fine-grained image recognition is a longstanding computer vision challenge that focuses on differentiating objects belonging to multiple subordinate categories within the same meta-category. Since images belonging to the same meta-category usually share similar visual appearances, mining discriminative visual cues is the key to distinguishing fine-grained categories. Although commonly used image-level data augmentation techniques have achieved great success in generic image classification problems, they are rarely applied in fine-grained scenarios, because their random editing-region behavior is prone to destroy the discriminative visual cues residing in the subtle regions. In this paper, we propose diversifying the training data at the feature-level to alleviate the discriminative region loss problem. Specifically, we produce diversified augmented samples by translating image features along semantically meaningful directions. The semantic directions are estimated with a covariance prediction network, which predicts a sample-wise covariance matrix to adapt to the large intra-class variation inherent in fine-grained images. Furthermore, the covariance prediction network is jointly optimized with the classification network in a meta-learning manner to alleviate the degenerate solution problem. Experiments on four competitive fine-grained recognition benchmarks (CUB-200-2011, Stanford Cars, FGVC Aircrafts, NABirds) demonstrate that our method significantly improves the generalization performance on several popular classification networks (e.g., ResNets, DenseNets, EfficientNets, RegNets and ViT). Combined with a recently proposed method, our semantic data augmentation approach achieves state-of-the-art performance on the CUB-200-2011 dataset. Source code is available at https://github.com/LeapLabTHU/LearnableISDA.

Multidimensional Refinement Graph Convolutional Network With Robust Decouple Loss for Fine-Grained Skeleton-Based Action Recognition

Graph convolutional networks (GCNs) have been widely used in skeleton-based action recognition. However, existing approaches are limited in fine-grained action recognition due to the similarity of interclass data. Moreover, the noisy data from pose extraction increase the challenge of fine-grained recognition. In this work, we propose a flexible attention block called channel-variable spatial-temporal attention (CVSTA) to enhance the discriminative power of spatial-temporal joints and obtain a more compact intraclass feature distribution. Based on CVSTA, we construct a multidimensional refinement GCN (MDR-GCN) that can improve the discrimination among channel-, joint-, and frame-level features for fine-grained actions. Furthermore, we propose a robust decouple loss (RDL) that significantly boosts the effect of the CVSTA and reduces the impact of noise. The proposed method combining MDR-GCN with RDL outperforms the known state-of-the-art skeleton-based approaches on fine-grained datasets, FineGym99 and FSD-10, and also on the coarse NTU-RGB + D 120 dataset and NTU-RGB + D X-view version. Our code is publicly available at https://github.com/dingyn-Reno/MDR-GCN.

Multiparameter spectral CT-based radiomics in predicting the expression of programmed death ligand 1 in non-small-cell lung cancer

AIM

To explore the value of radiomics for predicting the expression of programmed death ligand 1 (PD-L1) in non-small-cell lung cancer (NSCLC) based on multiparameter spectral computed tomography (CT) images.

MATERIALS AND METHODS

A total of 220 patients with NSCLC were enrolled retrospectively and divided into the training (n=176) and testing (n=44) cohorts. The radiomics features were extracted from the conventional CT images, mono-energy 40 keV images, iodine density (ID) maps, Z-effective maps, and electron density maps. The logistic regression (LR) and support vector machine (SVM) algorithms were employed to build models based on radiomics signatures. The prediction abilities were qualified by the area under the curve (AUC) obtained from the receiver operating characteristic (ROC) curve. Internal validation was performed on the independent testing dataset.

RESULTS

The combined model for PD-L1 ≥1%, which consisted of the radiomics score (rad-score; p<0.0001), white blood cell (WBC; p=0.027) counts, and air bronchogram (p=0.003), reached the highest performance with the AUCs of 0.873 and 0.917 in the training and testing dataset, respectively, which was better than the radiomics model with the AUCs of 0.842 and 0.886. The combined model for PD-L1 ≥50%, which consisted of rad-score (p<0.0001) and WBC counts (p=0.027), achieved the highest performance in the training and testing dataset with AUCs of 0.932 and 0.903, respectively, which was better than the radiomics model with AUCs of 0.920 and 0.892, respectively.

CONCLUSION

The radiomics model based on the multiparameter images of spectral CT can predict the expression level of PD-L1 in NSCLC. The combined model can obtain higher prediction efficiency and serves as a promising method for immunotherapy selection.

An Energy Matching Vessel Segmentation Framework in 3-D Medical Images

Accurate vascular segmentation from High Resolution 3-Dimensional (HR3D) medical scans is crucial for clinicians to visualize complex vasculature and diagnose related vascular diseases. However, a reliable and scalable vessel segmentation framework for HR3D scans remains a challenge. In this work, we propose a High-resolution Energy-matching Segmentation (HrEmS) framework that utilizes deep learning to directly process the entire HR3D scan and segment the vasculature to the finest level. The HrEmS framework introduces two novel components. Firstly, it uses the real-order total variation operator to construct a new loss function that guides the segmentation network to obtain the correct topology structure by matching the energy of the predicted segment to the energy of the manual label. This is different from traditional loss functions such as dice loss, which matches the pixels between predicted segment and manual label. Secondly, a curvature-based weight-correction module is developed, which directs the network to focus on crucial and complex structural parts of the vasculature instead of the easy parts. The proposed HrEmS framework was tested on three in-house multi-center datasets and three public datasets, and demonstrated improved results in comparison with the state-of-the-art methods using both topology-relevant and volumetric-relevant metrics. Furthermore, a double-blind assessment by three experienced radiologists on the critical points of the clinical diagnostic processes provided additional evidence of the superiority of the HrEmS framework.

Effect of SIRT1 gene single-nucleotide polymorphisms on susceptibility to type 1 diabetes in a Han Chinese population

AIMS

SIRT1 deficiency has been associated with diabetes, and a variant of the SIRT1 gene has been found to be involved in human autoimmune diabetes; however, it is unclear whether this genetic variation exists in Han Chinese with type 1 diabetes (T1D) and whether it contributes to development of T1D. Therefore, we aimed to explore the association of the SIRT1 gene single-nucleotide polymorphisms (SNPs) rs10997866 and rs3818292 in a Han Chinese population with T1D.

METHODS

This study recruited 2653 unrelated Han Chinese individuals, of whom 1289 had T1D and 1364 were healthy controls. Allelic and genotypic distributions of SIRT1 polymorphisms (rs10997866 and rs3818292) were determined by MassARRAY. Basic characteristics, genotype and allele frequencies of selected SNPs were compared between the T1D patients and healthy controls. Further genotype-phenotype association analysis of the SNPs was performed on the T1D patients divided into three groups according to genotype. Statistical analyses included the chi-square test, Mann‒Whitney U test, Kruskal‒Wallis H test and logistic regression.

RESULTS

The allelic (G vs. A) and genotypic (GA vs. AA) distributions of SIRT1 rs10997866 were significantly different in T1D patients and healthy controls (P = 0.039, P = 0.027), and rs10997866 was associated with T1D susceptibility under dominant, overdominant and additive models (P = 0.026, P = 0.030 and P = 0.027, respectively). Moreover, genotype-phenotype association analysis showed the GG genotype of rs10997866 and the GG genotype of rs3818292 to be associated with higher titers of IA-2A (P = 0.013 and P = 0.038, respectively).

CONCLUSION

SIRT1 rs10997866 is significantly associated with T1D susceptibility, with the minor allele G conferring a higher risk of T1D. Moreover, SIRT1 gene rs10997866 and rs3818292 correlate with the titer of IA-2A in Han Chinese individuals with T1D.

Probabilistic Contrastive Learning for Long-Tailed Visual Recognition

Long-tailed distributions frequently emerge in real-world data, where a large number of minority categories contain a limited number of samples. Such imbalance issue considerably impairs the performance of standard supervised learning algorithms, which are mainly designed for balanced training sets. Recent investigations have revealed that supervised contrastive learning exhibits promising potential in alleviating the data imbalance. However, the performance of supervised contrastive learning is plagued by an inherent challenge: it necessitates sufficiently large batches of training data to construct contrastive pairs that cover all categories, yet this requirement is difficult to meet in the context of class-imbalanced data. To overcome this obstacle, we propose a novel probabilistic contrastive (ProCo) learning algorithm that estimates the data distribution of the samples from each class in the feature space, and samples contrastive pairs accordingly. In fact, estimating the distributions of all classes using features in a small batch, particularly for imbalanced data, is not feasible. Our key idea is to introduce a reasonable and simple assumption that the normalized features in contrastive learning follow a mixture of von Mises-Fisher (vMF) distributions on unit space, which brings two-fold benefits. First, the distribution parameters can be estimated using only the first sample moment, which can be efficiently computed in an online manner across different batches. Second, based on the estimated distribution, the vMF distribution allows us to sample an infinite number of contrastive pairs and derive a closed form of the expected contrastive loss for efficient optimization. Other than long-tailed problems, ProCo can be directly applied to semi-supervised learning by generating pseudo-labels for unlabeled data, which can subsequently be utilized to estimate the distribution of the samples inversely. Theoretically, we analyze the error bound of ProCo. Empirically, extensive experimental results on supervised/semi-supervised visual recognition and object detection tasks demonstrate that ProCo consistently outperforms existing methods across various datasets.

Critical Values of Daily Sedentary Time and Its Longitudinal Association with Mild Cognitive Impairment Considering APOE ε4: A Prospective Cohort Study

BACKGROUND

Long sedentary time and physical inactivity are negatively related to cognition, but the cut-off value remains unclear, and apolipoprotein E polymorphism ε4 (APOE ε4) is a known genetic risk factor of mild cognitive impairment (MCI).

OBJECTIVES

To explore longitudinal association of sedentary time and MCI, and to identify a cutoff value that increases the risk of developing MCI, taking into account APOE ε4 stratification and its interactions.

DESIGN

A prospective cohort study.

SETTING

Population-based study.

PARTICIPANTS

We included 4932 older adults from Tianjin Elderly Nutrition and Cognition (TENC) cohort study recruited from March 2018 to June 2021 with 3.11 years of median follow-up time.

MEASUREMENTS

The primary outcome was newly diagnosed MCI, which was diagnosed by a modified version of the Petersen's criteria. The information of sedentary time (hours/day) and physical activity (MET-h/week) were obtained by questionnaire. Cox proportional hazard regression models and restricted spline curve were conducted.

RESULTS

A total of 4932 participants were included (mean [SD] age, 67.85 [4.96] years; 2627 female [53.3%] and 2305 male [46.7%]), 740 newly onset MCI patients were identified. Longer sedentary time was associated with higher risk of MCI for all participants (HR:1.069, 95%CI: 1.034, 1.105), especially in APOE ε4 non-carriers (HR:1.083, 95%CI: 1.045, 1.123) whether adjusted potential confounders. Sedentary time had synergistic interactions with APOE ε4 (β:1.503, 95%CI: 1.163, 1.942) and physical activities (β: 1.495, 95%CI: 1.210, 1.846). Restricted spline curve showed a cut-off value of 3.03 hours/day.

CONCLUSIONS

Long sedentary time (≥3.03 hours/day) could increase MCI risk, especially in APOE ε4 non-carriers, people with higher PA, aged 65 and above.

[Evaluating the efficiency of endoscope-assisted septo-rhinoplasty via intranasal approach]

Objective: To observe the functional and esthetic results of endoscope-assisted septo-rhinoplasty via intranasal approach. Methods: The clinical data of 12 patients with septal deviation and nasal deformity in the University of Hong Kong-Shenzhen Hospital from June 2021 to June 2022 were analyzed, including 8 males and 4 females, aging 28 to 58 years. All patients were operated under general anesthesia. The septal deviation was corrected by hemitransfixion incision, combined with intercartilaginous incision and other intranasal or extranasal incisions to perform osteotomy, hump resection and saddle nose correction. Patient satisfaction with postoperative functional and aesthetic improvements was assessed through Nasal Obstruction Symptom Evaluation (NOSE), Rhinoplasty Outcome Evaluation (ROE), and Visual Analogue Scale (VAS).SPSS 25 software was used for statistical analysis. Results: The correction of nasal septum deviation was satisfactory in all 12 patients. Nasal obstruction was relieved, with NOSE score and VAS score of nasal ventilation decreased [(21.67±10.30) vs (70.83±14.12), (1.83±1.03) vs (8.33±0.89), t value was 9.49 and 16.30, respectively, both P<0.05]. The nasal appearance of 10 patients with crooked nose deformity was improved, with ROE score and VAS score of nasal appearance increased [(21.30±2.31) vs (8.10±3.31), (8.90±0.99) vs (3.80±1.62), t value was -11.85 and -9.33, respectively, both P<0.05];The nasal vestibule morphology of 2 patients with abnormal nasal vestibule was improved. During the follow-up of 12 to 24 months, no postoperative complication such as nasal septum perforation, nasal cavity adhesion or nasal dorsal collapse occurred in all patients. Conclusion: Endoscope-assisted septo-rhinoplasty via intranasal approach can resolve both functional and esthetic problems at the same time, improving outcomes while reducing surgical trauma.

A review on nondiabetic hypoglycemia from various causes: Case series report

RATIONALE

Hypoglycemia is common in patients with glucose regulation disorders and related diabetic treatments but is rare in nondiabetic patients. Severe hypoglycemia can cause harm to patients' cognition, consciousness, central nervous system, cardiovascular and cerebrovascular system, and even death. However, the most fundamental way to control hypoglycemia is to identify the cause and deal with the primary disease. This article introduces 3 cases of nondiabetic hypoglycemia with different causes, aiming to improve our understanding of nondiabetic hypoglycemia and improve the ability of early diagnosis and differential diagnosis.

PATIENT CONCERNS

Case 1 is a 19-year-old female with a history of recurrent coma, and magnetic resonance imaging and endoscopic ultrasound of the pancreas suggest insulinoma. Case 2 is a 74-year-old male with a history of viral hepatitis, and computerized tomography shows multiple nodules in the liver, which is diagnosed as liver cancer. Case 3 is a 39-year-old female with a history of taking methimazole, who tested positive for insulin antibodies, and was diagnosed with insulin autoimmune syndrome.

DIAGNOSIS

All 3 patients were diagnosed with nondiabetic hypoglycemia, but the causes varied, and included insulinoma, non-islet cell tumor-induced hypoglycemia, and insulin autoimmune syndrome.

INTERVENTIONS

Case 1 underwent pancreatic tail resection; case 2 refused anti-tumor treatment and received glucose injections for palliative treatment only; and case 3 stopped taking methimazole.

OUTCOMES

After surgery, the blood sugar in case 1 returned to normal, and the blood sugar in case 2 was maintained at about 6.0 mmol/L. The symptoms of hypoglycemia gradually improved in case 3 after stopping the medication.

LESSONS

Non-diabetic hypoglycemia requires further examination to clarify the cause, and the correct differential diagnosis can provide timely and effective treatment, improving the patient's prognosis.

[Effect of information-knowledge-attitude-practice theory-based health education on clonorchiasis control among community residents and primary school students in Zhongshan City]

OBJECTIVE

To evaluate the effect of information-knowledge-attitude-practice (IKAP) theory-based health education interventions on clonorchiasis control among community residents and primary school students in Torch High-tech Development Zone of Zhongshan City, so as to provide insights into formulation of clonorchiasis control strategies among different types of populations.

METHODS

Residents were randomly sampled from two communities and students from two primary schools using a random cluster sampling method in Torch High-tech Development Zone, Zhongshan City from 2019 to 2021, and health education pertaining to clonorchiasis control was performed based on the IKAP theory. The changes in the awareness of basic clonorchiasis control knowledge and belief, self-reported risk of Clonorchis sinensis infections and dietary behaviors were compared among community residents and primary school students before and after health education interventions.

RESULTS

The participants included 146 male and 151 female community residents and 158 male and 153 female primary school students, with no significant difference detected in gender distribution (χ2 = 0.16, P > 0.05). The mean awareness of basic clonorchiasis control knowledge increased by 44.71% and 43.28% among primary school students and community residents 6 months following health education, and there were significant differences in the awareness of each item of basic clonorchiasis control knowledge before and after health education (χ2 = 41.53 to 284.44, all P values < 0.05). The proportions of primary school students and community residents thinking very high and high risks of C. sinensis infections increased from 9.35% and 6.71% before health education to 22.15% and 37.75% after health education, but only the difference of the attitudes in community residents achieve statistical significance (χ2 = 92.18, P < 0.05). The frequency of separation of raw and cooked foods with chopping board and knife significantly increased among community residents and primary school students following health education (χ2 = 16.04 to 62.65, all P values < 0.05). The frequency of eating raw freshwater fish (χ2 = 32.85, P < 0.05), fish congee (χ2 = 7.02, P < 0.05) and fish fillet hot pot (χ2 = 4.88, P < 0.05) significantly reduced among primary school students following health education, while only the frequency and proportions of eating raw freshwater fish have significantly reduced in community residents (χ2 = 11.77, P < 0.05).

CONCLUSIONS

Health education interventions based on the IKAP theory are effective to increase the awareness of clonorchiasis control knowledge and improve dietary behaviors associated with C. sinensis infections among community residents and primary school students in Zhongshan City.

Domain Adaptation via Prompt Learning

Unsupervised domain adaptation (UDA) aims to adapt models learned from a well-annotated source domain to a target domain, where only unlabeled samples are given. Current UDA approaches learn domain-invariant features by aligning source and target feature spaces through statistical discrepancy minimization or adversarial training. However, these constraints could lead to the distortion of semantic feature structures and loss of class discriminability. In this article, we introduce a novel prompt learning paradigm for UDA, named domain adaptation via prompt learning (DAPrompt). In contrast to prior works, our approach learns the underlying label distribution for target domain rather than aligning domains. The main idea is to embed domain information into prompts, a form of representation generated from natural language, which is then used to perform classification. This domain information is shared only by images from the same domain, thereby dynamically adapting the classifier according to each domain. By adopting this paradigm, we show that our model not only outperforms previous methods on several cross-domain benchmarks but also is very efficient to train and easy to implement.

Hundreds Guide Millions: Adaptive Offline Reinforcement Learning With Expert Guidance

Offline reinforcement learning (RL) optimizes the policy on a previously collected dataset without any interactions with the environment, yet usually suffers from the distributional shift problem. To mitigate this issue, a typical solution is to impose a policy constraint on a policy improvement objective. However, existing methods generally adopt a "one-size-fits-all" practice, i.e., keeping only a single improvement-constraint balance for all the samples in a mini-batch or even the entire offline dataset. In this work, we argue that different samples should be treated with different policy constraint intensities. Based on this idea, a novel plug-in approach named guided offline RL (GORL) is proposed. GORL employs a guiding network, along with only a few expert demonstrations, to adaptively determine the relative importance of the policy improvement and policy constraint for every sample. We theoretically prove that the guidance provided by our method is rational and near-optimal. Extensive experiments on various environments suggest that GORL can be easily installed on most offline RL algorithms with statistically significant performance improvements.

Changes and Relationship in Nutrition Impact Symptoms, Malnutrition during Esophageal Cancer Treatment

PURPOSE/OBJECTIVE(S)

The aim of this study was to assess the changes and relationship between nutritional impact symptoms (NIS) and malnutrition incidence during radiotherapy in patients with esophageal cancer.

MATERIALS/METHODS

A prospective observational study recruited hospitalized patients with esophageal cancer who received radiotherapy or did not receive chemotherapy. 283 patients with esophageal carcinoma were followed up before and during the fourth week of radiotherapy. Nutritional parameters were collected during radiotherapy.

RESULTS

According to the patient 's assessment of NIS and subjective global assessment (PG-SGA), at the mid-term of radiotherapy, the proportion of patients with NIS≥3 increased from 20.8% to 61.13%. Inappetence (37.1%) and abdominal distension (28.6%) were the most common nutritional symptoms. Severe malnutrition increased from 39% to 58.1%.NIS (odds ratio (OR) 30.93, 95% CI 15.92, 60.10, p <0.001) and weight loss of ≥5% (odds ratio (OR) 24.1, 95% CI 11.98, 48.47, p <0.001) were independently associated with severe malnutrition during radiotherapy.

CONCLUSION

Strengthen the nutritional support therapy during mid-radiotherapy for esophageal cancer patient, and NIS can directly predict malnutrition.PG-SGA and NIS can be used for nutritional monitoring in esophageal cancer patients.

Nomogram predicts risk of perineural invasion based on serum biomarkers for pancreatic cancer

BACKGROUND

Pancreatic cancer is a fatal tumor, and the status of perineural invasion (PNI) of pancreatic cancer was positively related to poor prognosis including overall survival and recurrence-free survival. This study aims to develop and validate a predictive model based on serum biomarkers to accurately predict the perineural invasion.

MATERIALS AND METHODS

The patients from No.924 Hospital of PLA Joint Logistic Support Force were included. The predictive model was developed in the training cohort using logistic regression analysis, and then tested in the validation cohort. The area under curve (AUC), calibration curves and decision curve analysis were used to validate the predictive accuracy and clinical benefits of nomogram.

RESULTS

A nomogram was developed using preoperative total bilirubin, preoperative blood glucose, preoperative CA19-9. It achieved good AUC values of 0.753 and 0.737 in predicting PNI in training and validation cohorts, respectively. Calibration curves showed nomogram had good uniformity of the practical probability of PNI. Decision curve analyses revealed that the nomogram provided higher diagnostic accuracy and superior net benefit compared to single indicators.

CONCLUSION

The present study constructed and validate a novel nomogram predicted the PNI of resectable PHAC patients with high stability and accuracy. Besides, it could better screen high-risk probability of PNI in these patients, and optimize treatment decision-making.

Detection of Cd2+ based on Nano-Fe3O4/MoS2/Nafion/GCE sensor

It is necessary to detect cadmium ions in seawater with high sensitivity because the pollution of cadmium ions seriously endangers the health and life of human beings. Nano-Fe3O4/MoS2/Nafion modified glassy carbon electrode was prepared by a drop coating method. The electrocatalytic properties of Nano-Fe3O4/MoS2/Nafion were measured by Cyclic Voltammetry (CV). Differential Pulse Voltammetry (DPV) was used to study the stripping Voltammetry response of the modified electrode to Cd2+. The optimal conditions were determined: In 0.1 mol/L HAc-NaAc solution, the solution pH was 4.2, the deposition potential was - 1.0 V, and the deposition time was 720 s, the membrane thickness was 8 μL. Under the optimum condition, the linear relation of Cd2+ concentration was found in the range of 5-300 μg/L, and the detection limit was 0.053 μg/L. The recovery of Cd2+ in seawater ranged from 99.2 to 102.9%. A composite material with simple operation, rapid response and high sensitivity was constructed for the determination of Cd2+ in seawater.

Ethnicity-specific association between TERT rs2736100 (A > C) polymorphism and lung cancer risk: a comprehensive meta-analysis

The rs2736100 (A > C) polymorphism of the second intron of Telomerase reverse transcriptase (TERT) has been confirmed to be closely associated with the risk of Lung cancer (LC), but there is still no unified conclusion on the results of its association with LC. This study included Genome-wide association studies (GWAS) and case-control studies reported so far on this association between TERT rs2736100 polymorphism and LC to clarify such a correlation with LC and the differences in it between different ethnicities and different types of LC. Relevant literatures published before May 7, 2022 on 'TERT rs2736100 polymorphism and LC susceptibility' in PubMed, EMbase, CENTRAL, MEDLINE databases were searched through the Internet, and data were extracted. Statistical analysis of data was performed in Revman5.3 software, including drawing forest diagrams, drawing funnel diagrams and so on. Sensitivity and publication bias analysis were performed in Stata 12.0 software. The C allele of TERT rs2736100 was associated with the risk of LC (Overall population: [OR] = 1.21, 95%CI [1.17, 1.25]; Caucasians: [OR] = 1.11, 95%CI [1.06, 1.17]; Asians: [OR] = 1.26, 95%CI [1.21, 1.30]), and Asians had a higher risk of LC than Caucasians (C vs. A: Caucasians: [OR] = 1.11 /Asians: [OR]) = 1.26). The other gene models also showed similar results. The results of stratified analysis of LC patients showed that the C allele was associated with the risk of Non-small-cell lung carcinoma (NSCLC) and Lung adenocarcinoma (LUAD), and the risk of NSCLC and LUAD in Asians was higher than that in Caucasians. The C allele was associated with the risk of Lung squamous cell carcinoma (LUSC) and Small cell lung carcinoma(SCLC) in Asians but not in Caucasians. NSCLC patients ([OR] = 1.27) had a stronger correlation than SCLC patients ([OR] = 1.03), and LUAD patients ([OR] = 1.32) had a stronger correlation than LUSC patients ([OR] = 1.09).In addition, the C allele of TERT rs2736100 was associated with the risk of LC, NSCLC and LUAD in both smoking groups and non-smoking groups, and the risk of LC in non-smokers of different ethnic groups was higher than that in smokers. In the Asians, non-smoking women were more at risk of developing LUAD. The C allele of TERT rs2736100 is a risk factor for LC, NSCLC, and LUAD in different ethnic groups, and the Asian population is at a more common risk. The C allele is a risk factor for LUSC and SCLC in Asians but not in Caucasians. And smoking is not the most critical factor that causes variation in TERT rs2736100 to increase the risk of most LC (NSCLC, LUAD). Therefore, LC is a multi-etiological disease caused by a combination of genetic, environmental and lifestyle factors.

SePiCo: Semantic-Guided Pixel Contrast for Domain Adaptive Semantic Segmentation

Domain adaptive semantic segmentation attempts to make satisfactory dense predictions on an unlabeled target domain by utilizing the supervised model trained on a labeled source domain. One popular solution is self-training, which retrains the model with pseudo labels on target instances. Plenty of approaches tend to alleviate noisy pseudo labels, however, they ignore the intrinsic connection of the training data, i.e., intra-class compactness and inter-class dispersion between pixel representations across and within domains. In consequence, they struggle to handle cross-domain semantic variations and fail to build a well-structured embedding space, leading to less discrimination and poor generalization. In this work, we propose emantic-Guided Pixel Contrast (SePiCo), a novel one-stage adaptation framework that highlights the semantic concepts of individual pixels to promote learning of class-discriminative and class-balanced pixel representations across domains, eventually boosting the performance of self-training methods. Specifically, to explore proper semantic concepts, we first investigate a centroid-aware pixel contrast that employs the category centroids of the entire source domain or a single source image to guide the learning of discriminative features. Considering the possible lack of category diversity in semantic concepts, we then blaze a trail of distributional perspective to involve a sufficient quantity of instances, namely distribution-aware pixel contrast, in which we approximate the true distribution of each semantic category from the statistics of labeled source data. Moreover, such an optimization objective can derive a closed-form upper bound by implicitly involving an infinite number of (dis)similar pairs, making it computationally efficient. Extensive experiments show that SePiCo not only helps stabilize training but also yields discriminative representations, making significant progress on both synthetic-to-real and daytime-to-nighttime adaptation scenarios. The code and models are available at https://github.com/BIT-DA/SePiCo.

Convolution-Enhanced Evolving Attention Networks

Attention-based neural networks, such as Transformers, have become ubiquitous in numerous applications, including computer vision, natural language processing, and time-series analysis. In all kinds of attention networks, the attention maps are crucial as they encode semantic dependencies between input tokens. However, most existing attention networks perform modeling or reasoning based on representations, wherein the attention maps of different layers are learned separately without explicit interactions. In this paper, we propose a novel and generic evolving attention mechanism, which directly models the evolution of inter-token relationships through a chain of residual convolutional modules. The major motivations are twofold. On the one hand, the attention maps in different layers share transferable knowledge, thus adding a residual connection can facilitate the information flow of inter-token relationships across layers. On the other hand, there is naturally an evolutionary trend among attention maps at different abstraction levels, so it is beneficial to exploit a dedicated convolution-based module to capture this process. Equipped with the proposed mechanism, the convolution-enhanced evolving attention networks achieve superior performance in various applications, including time-series representation, natural language understanding, machine translation, and image classification. Especially on time-series representation tasks, Evolving Attention-enhanced Dilated Convolutional (EA-DC-) Transformer outperforms state-of-the-art models significantly, achieving an average of 17% improvement compared to the best SOTA. To the best of our knowledge, this is the first work that explicitly models the layer-wise evolution of attention maps. Our implementation is available at https://github.com/pkuyym/EvolvingAttention.

Glance and Focus Networks for Dynamic Visual Recognition

Spatial redundancy widely exists in visual recognition tasks, i.e., discriminative features in an image or video frame usually correspond to only a subset of pixels, while the remaining regions are irrelevant to the task at hand. Therefore, static models which process all the pixels with an equal amount of computation result in considerable redundancy in terms of time and space consumption. In this paper, we formulate the image recognition problem as a sequential coarse-to-fine feature learning process, mimicking the human visual system. Specifically, the proposed Glance and Focus Network (GFNet) first extracts a quick global representation of the input image at a low resolution scale, and then strategically attends to a series of salient (small) regions to learn finer features. The sequential process naturally facilitates adaptive inference at test time, as it can be terminated once the model is sufficiently confident about its prediction, avoiding further redundant computation. It is worth noting that the problem of locating discriminant regions in our model is formulated as a reinforcement learning task, thus requiring no additional manual annotations other than classification labels. GFNet is general and flexible as it is compatible with any off-the-shelf backbone models (such as MobileNets, EfficientNets and TSM), which can be conveniently deployed as the feature extractor. Extensive experiments on a variety of image classification and video recognition tasks and with various backbone models demonstrate the remarkable efficiency of our method. For example, it reduces the average latency of the highly efficient MobileNet-V3 on an iPhone XS Max by 1.3x without sacrificing accuracy. Code and pre-trained models are available at https://github.com/blackfeather-wang/GFNet-Pytorch.

Joint optic disc and cup segmentation based on elliptical-like morphological feature and spatial geometry constraint

Glaucoma is a chronic degenerative disease that is the second leading cause of irreversible blindness worldwide. For a precise and automatic screening of glaucoma, detecting the optic disc and cup precisely is significant. In this paper, combining the elliptical-like morphological features of the disc and cup, we reformulate the segmentation task from a perspective of ellipse detection to explicitly segment and directly get the glaucoma screening indicator. We detect the minimum bounding boxes of ellipses firstly, and then learn the ellipse parameters of these regions to achieve optic disc and cup segmentation. Considering the spatial geometry prior knowledge that the cup should be within the disc region, Paired-Box RPN is introduced to simultaneously detect the disc and cup coupled. In addition, boundary attention module is introduced to use edges of the disc and cup as an important guide for context aggregation to improve the accuracy. Comprehensive experiments clearly show that our method outperforms the state-of-the-art methods for optic disc and cup segmentation. Simultaneously, the proposed method also obtains the good glaucoma screening performance with calculated vCDR value. Joint optic disc and cup segmentation, which utilizes the elliptical-like morphological features and spatial geometry constraint, could improve the performance of optic disc and cup segmentation.

Meta-Reinforcement Learning With Dynamic Adaptiveness Distillation

Deep reinforcement learning is confronted with problems of sampling inefficiency and poor task migration capability. Meta-reinforcement learning (meta-RL) enables meta-learners to utilize the task-solving skills trained on similar tasks and quickly adapt to new tasks. However, meta-RL methods lack enough queries toward the relationship between task-agnostic exploitation of data and task-related knowledge introduced by latent context, limiting their effectiveness and generalization ability. In this article, we develop an algorithm for off-policy meta-RL that can provide the meta-learners with self-oriented cognition toward how they adapt to the family of tasks. In our approach, we perform dynamic task-adaptiveness distillation to describe how the meta-learners adjust the exploration strategy in the meta-training process. Our approach also enables the meta-learners to balance the influence of task-agnostic self-oriented adaption and task-related information through latent context reorganization. In our experiments, our method achieves 10%-20% higher asymptotic reward than probabilistic embeddings for actor-critic RL (PEARL).

A coupled-lines system to determine the anteroposterior position of maxillary central incisors for smiling profile esthetics

OBJECTIVES

To develop a coupled-lines system to determine the anteroposterior position of maxillary central incisors (U1) for smiling profile esthetics.

MATERIALS AND METHODS

Thirty Asian females with ordinary and good facial harmony were selected as the study sample and the control sample, respectively. Three-dimensional facial images and 45°- and 90°-angled profiles were collected. The anteroposterior relationships between U1 and upper- and mid-facial soft tissue landmarks were measured. By morphing photos of the study sample, two artificial images were created to represent the well-balanced 45°- and 90°-angled profiles and were further processed with combined variations of soft tissue subnasale (SSn)-Glabella and the mid-point of facial axial points of the bilateral central incisor (mFA)-SSn distances. Esthetic assessments were performed on these images by layperson (n = 94) and orthodontist (n = 94) raters.

RESULTS

Both upper- and mid-facial soft tissue landmarks were indispensable in assessing anteroposterior positions of U1 for well-balanced smiling profiles. As assessed in 45°- and 90°-angled profiles, the most esthetically sensitive parameters were mFA-Glabella and mFA-SSn distances. A coupled-lines system was constructed, comprising the Glabella and SSn vertical lines. In smiling profiles with optimal esthetics, the mFA point was at 2 to 5 mm posterior to the Glabella vertical and concomitantly 4 to 7 mm posterior to the SSn vertical, as perceived by orthodontists. Laypersons gave a wider range for mFA-Glabella distances, at 2 to 6 mm.

CONCLUSIONS

The coupled-lines system could serve as a reliable reference for determining esthetically optimal anteroposterior positions of U1 for female facial profiles.

Ultrasound Fusion-Guided Core Needle Biopsy for Deep Head and Neck Space Lesions: Technical Feasibility, Histopathologic Yield, and Safety

BACKGROUND AND PURPOSE

Ultrasound is generally considered to have a minor role in guiding biopsies for deep head and neck space lesions. However, the ultrasound fusion technique may have the potential to change this opinion. This study evaluated the feasibility, histopathologic yield, and safety of ultrasound fusion-guided core needle biopsies for deep head and neck space lesions.

MATERIALS AND METHODS

From March 2021 to April 2022, patients with primary deep head and neck space lesions were prospectively included in this study. Ultrasound fusion was performed with contemporaneous CT, MR imaging, or PET/CT studies, and ultrasound fusion-guided core needle biopsy was performed by using a Micro-Convex probe via 4 different needle approaches. Feasibility, histopathologic results, and biopsy-related complications were observed. Descriptive statistics were applied.

RESULTS

Ultrasound-guided biopsy was feasible in all 16 patients (11 women and 5 men; mean age 46 [SD, 16] years; range, 16-76 years). The lesions were located in the parapharyngeal space, infratemporal fossa, and skull base, with a median diameter of 3.8 cm (range, 2.2-6.5 cm). An adequate and definite histopathologic yield was obtained in 15/16 (93.8%) patients; among them, 4/15 lesions (26.7%) were malignant, and 11/15 (73.6%) were benign. No major complications occurred. Minor complications were noted in 2 of the 16 (12.5%) patients (self-limiting inflammation in 1 and bleeding in 1).

CONCLUSIONS

This study demonstrates that ultrasound fusion-guided biopsy of deep head and neck space lesions is feasible and safe, with a high histopathologic yield.

Correlation study between apparent diffusion coefficients and the prognostic factors in breast cancer

AIM

To analyse the correlation between apparent diffusion coefficients (ADC) derived from intratumoural and peritumoural regions with prognostic factors and immune-inflammatory markers in breast cancer (BC).

MATERIALS AND METHODS

In this retrospective study, 89 patients (age range, 28-66 years; median, 45 years) with a diagnosis of invasive BC who underwent routine blood tests and multiparametric magnetic resonance imaging (MRI) were enrolled. The study cohort was stratified according to tumour maximum cross-section ≥20 mm, lymph node metastasis (LNM), time-signal intensity curve (TIC) type, and receptor status. Minimum, maximum, mean, and heterogeneity values of tumour ADC (ADC_tmin, ADC_tmax, ADC_tmean, and ADC_heter), maximum values of peritumoural ADC (ADC_pmax), and the ratio of peritumoural-tumour ADC (ADC_ratio) were obtained on the ADC maps. Linear regression analyses were performed to investigate the correlation between immune-inflammatory markers, prognostic factors and ADC values.

RESULTS

HER-2 was positively associated with ADC_tmax, ADC_tmean, and ADC_pmax values (β = 0.306, p=0.004; β = 0.283, p=0.007; β = 0.262, p=0.007, respectively), while platelet-to-lymphocyte ratio (PLR) was positively associated with ADC_pmax and ADC_ratio values (β = 0.227, p=0.020; β = 0.231, p=0.020, respectively). Among ADC parameters, ADC_pmax showed the highest predictive values for evaluating the presence of LNM (AUC, 0.751; sensitivity, 70.4%; specificity, 77.1%).

CONCLUSION

The ADC_pmax value could provide additional assistance in predicting prognostic factors of BC.

ncRNA-mediated upregulation of FAM83A is associated with poor prognosis and immune infiltration in pancreatic cancer

INTRODUCTION

Malignant pancreatic cancer has poor long-term survival. Increasing evidence shows that FAM83A (family with sequence similarity 83 member A) plays a vital role in tumorigenesis and malignant progression in some human cancer types. The present study explored the potential mechanism of FAM83A in improving the prognosis of pancreatic cancer patients.

METHODS

Transcriptomic and clinical data from patients were obtained from The Cancer Genome Atlas while FAM83A expression was measured in tumorous pancreatic tissue compared with normal controls by quantitative real-time PCR and immunohistochemistry.

RESULTS

FAM83A is a vital prognostic indicator and potential oncogene in pancreatic cancer via pan-cancer analysis. In silico analysis revealed that AL049555.1/hsa-miR-129-5p axis was the pivotal upstream ncRNA- mediated pathway of FAM83A in pancreatic cancer. Furthermore, FAM83A expression was related to immune cell infiltration through vital immune-related genes including programmed cell death 1 (PDCD1), and tumorigenesis through common mutation genes including KRAS protooncogene GTPase (KRAS), and SMAD family member 4 (SMAD4). In summary, ncRNA-mediated upregulation of FAM83A is associated with poor long-term survival and immune cell infiltration in pancreatic cancer.

DISCUSSION

FAM83A may be used as a novel survival-related and immune-related biomarker. This information suggests that FAM83A may be a novel therapeutic target for combined or individual treatment for patients with pancreatic cancer.

Genetic influences of the effect of circulating inflammatory cytokines on osteoarthritis in humans

OBJECTIVE

The causal relationship between inflammatory cytokines and Osteoarthritis (OA) has not been well investigated. This study investigated the causal role of inflammatory cytokines in the risk of OA and total joint arthroplasty using the Mendelian randomization (MR) method.

METHOD

Single nucleotide polymorphisms (SNPs) robustly associated with inflammatory cytokines were used as instrumental variables. The inverse-variance weighted (IVW) method with false discovery rate (FDR) adjusted P-value (q-value) for multiple comparisons were used as the main MR method to estimate causal effects based on the summary-level data for OA (knee and hip OA, respectively) and total joint arthroplasty (TJA). Sensitivity analyses validated the robustness of the results and ensured the absence of heterogeneity and horizontal pleiotropy.

RESULTS

After FDR adjustment, macrophage colony-stimulating factor (MCSF) and vascular endothelial growth factor (VEGF) were identified as causally associated with knee OA (MCSF, odds ratio [OR]: 1.16, 95% confidence interval [CI]: 1.09-1.23, q = 5.05 × 10^-5; VEGF, OR: 1.09, 95% CI: 1.04-1.15, q = 0.011). We also observed that genetically predicted MCSF and VEGF were positively associated with the risk of TJA, and MCP3 was negatively associated with for the risk of TJA, although the effects seem fairly modest. Sensitivity analysis further excluded the influence of heterogeneity and horizontal pleiotropy.

CONCLUSIONS

Inflammatory cytokines, namely MCSF and VEGF, were causally associated with knee OA, which could enhance our understanding of inflammation in OA pathology.

Convolutional Networks with Dense Connectivity

Recent work has shown that convolutional networks can be substantially deeper, more accurate, and efficient to train if they contain shorter connections between layers close to the input and those close to the output. In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion. Whereas traditional convolutional networks with L layers have L connections-one between each layer and its subsequent layer-our network has [Formula: see text] direct connections. For each layer, the feature-maps of all preceding layers are used as inputs, and its own feature-maps are used as inputs into all subsequent layers. DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, encourage feature reuse and substantially improve parameter efficiency. We evaluate our proposed architecture on four highly competitive object recognition benchmark tasks (CIFAR-10, CIFAR-100, SVHN, and ImageNet). DenseNets obtain significant improvements over the state-of-the-art on most of them, whilst requiring less parameters and computation to achieve high performance.

Temporal-Spatial Causal Interpretations for Vision-Based Reinforcement Learning

Deep reinforcement learning (RL) agents are becoming increasingly proficient in a range of complex control tasks. However, the agent's behavior is usually difficult to interpret due to the introduction of black-box function, making it difficult to acquire the trust of users. Although there have been some interesting interpretation methods for vision-based RL, most of them cannot uncover temporal causal information, raising questions about their reliability. To address this problem, we present a temporal-spatial causal interpretation (TSCI) model to understand the agent's long-term behavior, which is essential for sequential decision-making. TSCI model builds on the formulation of temporal causality, which reflects the temporal causal relations between sequential observations and decisions of RL agent. Then a separate causal discovery network is employed to identify temporal-spatial causal features, which are constrained to satisfy the temporal causality. TSCI model is applicable to recurrent agents and can be used to discover causal features with high efficiency once trained. The empirical results show that TSCI model can produce high-resolution and sharp attention masks to highlight task-relevant temporal-spatial information that constitutes most evidence about how vision-based RL agents make sequential decisions. In addition, we further demonstrate that our method is able to provide valuable causal interpretations for vision-based RL agents from the temporal perspective.

Dynamic Neural Networks: A Survey

Dynamic neural network is an emerging research topic in deep learning. Compared to static models which have fixed computational graphs and parameters at the inference stage, dynamic networks can adapt their structures or parameters to different inputs, leading to notable advantages in terms of accuracy, computational efficiency, adaptiveness, etc. In this survey, we comprehensively review this rapidly developing area by dividing dynamic networks into three main categories: 1) sample-wise dynamic models that process each sample with data-dependent architectures or parameters; 2) spatial-wise dynamic networks that conduct adaptive computation with respect to different spatial locations of image data; and 3) temporal-wise dynamic models that perform adaptive inference along the temporal dimension for sequential data such as videos and texts. The important research problems of dynamic networks, e.g., architecture design, decision making scheme, optimization technique and applications, are reviewed systematically. Finally, we discuss the open problems in this field together with interesting future research directions.

Attention-Based Meta-Reinforcement Learning for Tracking Control of AUV With Time-Varying Dynamics

Reinforcement learning (RL) is a promising technique for designing a model-free controller by interacting with the environment. Several researchers have applied RL to autonomous underwater vehicles (AUVs) for motion control, such as trajectory tracking. However, the existing RL-based controller usually assumes that the unknown AUV dynamics keep invariant during the operation period, limiting its further application in the complex underwater environment. In this article, a novel meta-RL-based control scheme is proposed for trajectory tracking control of AUV in the presence of unknown and time-varying dynamics. To this end, we divide the tracking task for AUV with time-varying dynamics into multiple specific tasks with fixed time-varying dynamics, to which we apply meta-RL for training to distill the general control policy. The obtained control policy can transfer to the testing phase with high adaptability. Inspired by the line-of-sight (LOS) tracking rule, we formulate each specific task as a Markov decision process (MDP) with a well-designed state and reward function. Furthermore, a novel policy network with an attention module is proposed to extract the hidden information of AUV dynamics. The simulation environment with time-varying dynamics is established, and the simulation results reveal the effectiveness of our proposed method.

Increased frequency of angiotensin converting enzyme D allele in Chinese Han patients with idiopathic pulmonary fibrosis: A systematic review and meta-analysis

BACKGROUND

To explore the association between Angiotensin Converting Enzyme (ACE) insert(I)/defect(D) gene polymorphism and the susceptibility to idiopathic pulmonary fibrosis (IPF).

METHODS

Searching PubMed, EMbase, CENTRAL, MEDLINE, CBM, China National Knowledge Infrastructure, WanFang Database and VIP Chinese Science database through a computer and collect the literature from China and foreign countries published before January 22, 2022. Screen the literatures and extract data such as first author, year of publication, diagnostic criteria and gene frequency, and draw a funnel chart and perform Begg's Test and Egger's test to evaluate publication bias. The influence analysis was performed for heterogeneous results and at the same time, the trial sequential analysis (TSA) was also conducted to confirm the robustness of the meta-analysis results. Registration number: CRD42021259341.

RESULTS

There were a total of 4 literatures (4 studies conducted in the Chinese Han population), and a total of 292 IPF patients and 351 healthy controls were included in this study. The results showed that in the Chinese Han population, the ACE I/D gene polymorphism was associated with the susceptibility of IPF (D vs I: [odds ratio, OR] = 0.53, 95% confidence interval [95%CI] [0.42, 0.67], P < .00001; DD vs II: [OR] = 0.37, 95%CI [0.24, 0.57], P < .00001; DD vs II + ID:[OR] = 0.30, 95%CI [0.21, 0.43], P < .00001), and the angiotensin II (Ang Ⅱ) level of IPF patients was higher than that of the control group (mean difference [MD] = 14.29, 95%CI [11.20,17.37], P < .00001).The TSA also confirmed that D allele was closely related to the susceptibility of IPF.

CONCLUSION

In the Chinese Han population, the D allele of the ACE I/D gene polymorphism is associated with the susceptibility of IPF.

Circ_0136666 aggravates osteosarcoma development through mediating miR-1244/CEP55 axis

Background

Accumulating articles demonstrate that circular RNAs play pivotal functions in tumorigenesis. However, the working mechanism of circ_0136666 in osteosarcoma (OS) progression remains to be further clarified.

Methods

Real time-quantitative polymerase chain reaction and western blot assay were applied to determine RNA and protein expression, respectively. Cell proliferation was assessed by 5-Ethynyl-2′-deoxyuridine assay and colony formation assay. Transwell assays were carried out to assess cell migration and invasion abilities. Flow cytometry was performed to analyze cell apoptosis. Cell glycolysis was evaluated by analyzing the uptake of glucose and the production of lactate using the corresponding kits. Dual-luciferase reporter assay and biotinylated RNA-pull down assay were performed to confirm the target interaction between microRNA-1244 (miR-1244) and circ_0136666 or centrosomal protein 55 (CEP55). Xenograft tumor model was utilized to explore the role of circ_0136666 in tumor growth in vivo.

Results

Circ_0136666 expression was prominently elevated in OS tissues and cell lines. Circ_0136666 absence restrained the proliferation, migration, invasion and glycolytic metabolism and promoted the apoptosis of OS cells. Circ_0136666 negatively regulated miR-1244 expression by binding to it in OS cells. MiR-1244 overexpression suppressed the malignant behaviors of OS cells. CEP55 was a target of miR-1244 in OS cells. Circ_0136666 positively regulated CEP55 expression partly by sequestering miR-1244 in OS cells. CEP55 overexpression largely reversed circ_0136666 silencing-mediated influences in OS cells. Circ_0136666 silencing significantly suppressed tumor growth in vivo.

Conclusion

Circ_0136666 silencing inhibited OS progression partly by targeting miR-1244/CEP55 signaling. Silencing circ_0136666 and CEP55 or restoring miR-1244 level might be a potential therapeutic strategy for OS.

Upright and Crawling Locomotion and Its Transition for a Wheel-Legged Robot

To face the challenge of adapting to complex terrains and environments, we develop a novel wheel-legged robot that can switch motion modes to adapt to different environments. The robot can perform efficient and stable upright balanced locomotion on flat roads and flexible crawling in low and narrow passages. For passing through low and narrow passages, we propose a crawling motion control strategy and methods for transitioning between locomotion modes of wheel-legged robots. In practical applications, the smooth transition between the two motion modes is challenging. By optimizing the gravity work of the body, the optimal trajectory of the center of mass (CoM) for the transition from standing to crawling is obtained. By constructing and solving an optimization problem regarding the posture and motion trajectories of the underactuated model, the robot achieves a smooth transition from crawling to standing. In experiments, the wheel-legged robot successfully transitioned between the crawling mode and the upright balanced moving mode and flexibly passed a low and narrow passage. Consequently, the effectiveness of the control strategies and algorithms proposed in this paper are verified by experiments.

Generalized Domain Conditioned Adaptation Network

Domain adaptation (DA) attempts to transfer knowledge learned in the labeled source domain to the unlabeled but related target domain without requiring large amounts of target supervision. Recent advances in DA mainly proceed by aligning the source and target distributions. Despite the significant success, the adaptation performance still degrades accordingly when the source and target domains encounter a large distribution discrepancy. We consider this limitation may attribute to the insufficient exploration of domain-specialized features because most studies merely concentrate on domain-general feature learning in task-specific layers and integrate totally-shared convolutional networks (convnets) to generate common features for both domains. In this paper, we relax the completely-shared convnets assumption adopted by previous DA methods and propose Domain Conditioned Adaptation Network (DCAN), which introduces domain conditioned channel attention module with a multi-path structure to separately excite channel activation for each domain. Such a partially-shared convnets module allows domain-specialized features in low-level to be explored appropriately. Further, given the knowledge transferability varying along with convolutional layers, we develop Generalized Domain Conditioned Adaptation Network (GDCAN) to automatically determine whether domain channel activations should be separately modeled in each attention module. Afterward, the critical domain-specialized knowledge could be adaptively extracted according to the domain statistic gaps. As far as we know, this is the first work to explore the domain-wise convolutional channel activations separately for deep DA networks. Additionally, to effectively match high-level feature distributions across domains, we consider deploying feature adaptation blocks after task-specific layers, which can explicitly mitigate the domain discrepancy. Extensive experiments on four cross-domain benchmarks, including DomainNet, Office-Home, Office-31, and ImageCLEF, demonstrate the proposed approaches outperform the existing methods by a large margin, especially on the large-scale challenging dataset. The code and models are available at https://github.com/BIT-DA/GDCAN.

Regularizing Deep Networks With Semantic Data Augmentation

Data augmentation is widely known as a simple yet surprisingly effective technique for regularizing deep networks. Conventional data augmentation schemes, e.g., flipping, translation or rotation, are low-level, data-independent and class-agnostic operations, leading to limited diversity for augmented samples. To this end, we propose a novel semantic data augmentation algorithm to complement traditional approaches. The proposed method is inspired by the intriguing property that deep networks are effective in learning linearized features, i.e., certain directions in the deep feature space correspond to meaningful semantic transformations, e.g., changing the background or view angle of an object. Based on this observation, translating training samples along many such directions in the feature space can effectively augment the dataset for more diversity. To implement this idea, we first introduce a sampling based method to obtain semantically meaningful directions efficiently. Then, an upper bound of the expected cross-entropy (CE) loss on the augmented training set is derived by assuming the number of augmented samples goes to infinity, yielding a highly efficient algorithm. In fact, we show that the proposed implicit semantic data augmentation (ISDA) algorithm amounts to minimizing a novel robust CE loss, which adds minimal extra computational cost to a normal training procedure. In addition to supervised learning, ISDA can be applied to semi-supervised learning tasks under the consistency regularization framework, where ISDA amounts to minimizing the upper bound of the expected KL-divergence between the augmented features and the original features. Although being simple, ISDA consistently improves the generalization performance of popular deep models (e.g., ResNets and DenseNets) on a variety of datasets, i.e., CIFAR-10, CIFAR-100, SVHN, ImageNet, and Cityscapes. Code for reproducing our results is available at https://github.com/blackfeather-wang/ISDA-for-Deep-Networks.

P-154 growth hormone supplementation decline quality of maternally aged oocytes via reducing aneuploidy

Study question

Does growth hormone (GH) supplementation ameliorate the quality of oocytes in aged mice via reducing aneuploidy?

Summary answer

Daily injections of growth hormone for 2 months improve the age-related decline in ovarian reserve, and promote oocyte quality by reducing the aneuploidy rate.

What is known already

Women of advanced maternal age, particular over 35 years, could consequently bear poor quality oocytes, which are mainly caused by chromosomal aneuploidy. Age-related decline in GH levels may be related to the insufficient reproductive potential in women of advanced maternal age. To date, the role of GH in in vitro fertilization remains controversial. GH has been reported to affect oocyte quality, and studies in vivo have shown that intraperitoneal administration of GH reduces the occurrence rate of spindle/chromosome defects.

Study design, size, duration

The young (6-week-old) and aged (8-month-old) C57BL/6J mice were used in the study. In the in vivo experiment, aged mice were intraperitoneally injected daily with GH (1.6 mg/kg body weight, n = 85) or the equivalent volume of NS (n = 81) for consecutive for two months. In the in vitro experiment, GV oocytes from aged mice were treated with GH (Gene Science Pharmaceuticals, Changchun, China) (200ng/ml) in M16 medium for 14∼16 hours.

Participants/materials, setting, methods

We analyzed the number of oocytes and hormone levels to assess ovarian reserve, by immunohistochemistry and ELISA methods. The expression of GHR was measured by real-time quantitative RT-PCR and immunofluorescence. Time–lapse incubator was utilized to record the developmental potential of oocytes and embryos. Immunofluorescence was performed to assess parameters of oocyte quality (mitochondrial functions, ROS level, spindle/chromosome defects, and DNA damage). Chromosome spread and single-cell DNA sequencing were used to analyze chromosome aneuploidy rate.

Main results and the role of chance

The expression of GHR decreased in aged oocyte (21.68±1.08) and increased after GH treatment(30.64±0.70, P &lt; 0.001). GH ameliorated decline in ovarian reserve of aged mice, with increased ovarian index (0.040%±0.0031 vs 0.023%±0.0019, P &lt; 0.01), number of preantral follicles (7.33±0.89 vs 4.12±0.58, P &lt; 0.05) and antral follicles (8.67±1.2 vs 3.33±0.88, P &lt; 0.05) compared to the aged controls, but no other hormonal differences were detected. After GH supplementation, aged oocyte showed better quality and developmental potential, including restored mitochondrial functions (P &lt; 0.05) and decreased ROS level (19.83±1.50 vs 23.6±1.72, P &lt; 0.05). As expected, GH effectively promoted the fertilization rate (39.9%±1.10 vs 14.4%±0.62, P &lt; 0.001) and early embryo development (P &lt; 0.01). In addition, the spindle/chromosome defects (31.6%±2.12 vs 48.9%±1.83, P &lt; 0.01), DNA damage (16.37±1.76 vs 49.35±2.52, P &lt; 0.001) and aneuploidy rate (21.5%±1.61 vs 34.6%±1.84, P &lt; 0.01) were lower in GH group than that in control group. Overall, GH treatment restored age-related decline in ovarian reserve, and decreased the occurrence rate of aneuploidy caused by spindle/chromosome defects, inducing a better oocyte quality in aged mice.

Limitations, reasons for caution

The molecular and mechanism of GH to regulate spindle/chromosome function remains to be determined. Besides, future work should be extended to human oocyte to determine whether this mechanism is conserved between mice and humans.

Wider implications of the findings

Our work expounds a theoretical basis for application of GH to improve the fertility of aged women. Besides, the results also feed new ideas for the prevention and treatment of oocyte quality decline in assisted reproductive technology.

Trial registration number

not applicable

P-759 Comparison of DNA methylation profiles of human embryo cultured in either uninterrupted or interrupted incubators

Study question

Are there are differences in the DNA methylation profiles of human embryos cultured in Time-lapse imaging (TLI) and standard incubators (SI)?

Summary answer

The genome-wide DNA methylation landscape was globally similar between the SI and TLI groups.

What is known already

Early embryonic development is a special biological process, along with dynamic changes of DNA methylation. In vitro culturing ensured the successful development of human embryo, and is the most important step of ART cycle. TLI is one type of newly developed ART embryo culture systems, which allows for continuous assessment of embryonic development. Our previous study performed transcriptome analysis to compare SI and TLI and found that the global transcriptomic profiles were similar between the two groups. However, whether there are differences in the DNA methylation profiles of human embryos cultured in TLI and SI still unknown.

Study design, size, duration

This study was designed to explore the influence of SI and TLI incubator culturing on genome-wide DNA methylation of human eight-cell embryos. A total of 9 women who received IVF treatment, ≤ 30 years old (range: 20–30 years), without a history of genetic diseases or smoking were included in this study.

Participants/materials, setting, methods

Transvaginal oocyte retrieval was performed 36 h after HCG injection. Cumulus-enclosed oocytes were collected in 2.5 mL IVF medium and incubated in 5% O2, 6% CO2, and 37 °C incubators for insemination. The fertilized oocytes were transferred into pre-equilibrated Embryoslides. Then the embryoslides were cultured in either TLI or SI at 37 °C with 5% O2 and 6% CO2 until embryo transfer on Day 3. Samples were sequenced by the Illumina HiSeq 4000 with a 150-bp paired-end.

Main results and the role of chance

The genome-wide methylation patterns and CpG methylation levels in transposable elements and imprinting control regions of TLI-cultured embryos were similar to those of the SI-cultured embryos. However, a small number of differentially methylated regions (DMRs) were detected, and these DMRs mainly occurred in exons other than promoters. Functional annotation revealed that the genes in DMRs tended to execute functions such as cell cycle, DNA damage stimulus, histone modification, mitochondrial, glucose import, and MAPK signaling pathway.

Limitations, reasons for caution

Small sample size.

Wider implications of the findings

Evaluated the safety of TLI culture system from the perspective of DNA methylation at single-cell level, and provided an important reference for understanding the association between embryo culture condition and epigenetic regulation.

Trial registration number

not applicable

A multimodality intervention to improve musculoskeletal health, function, metabolism, and well-being in spinal cord injury: study protocol for the FIT-SCI randomized controlled trial

BACKGROUND

A spinal cord injury (SCI) is a devastating, life-changing event that has profoundly deleterious effects on an individual's health and well-being. Dysregulation of neuromuscular, cardiometabolic, and endocrine organ systems following an SCI contribute to excess morbidity, mortality and a poor quality of life. As no effective treatments currently exist for SCI, the development of novel strategies to improve the functional and health status of individuals living with SCI are much needed. To address this knowledge gap, the current study will determine whether a Home-Based Multimodality Functional Recovery and Metabolic Health Enhancement Program that consists of functional electrical stimulation of the lower extremity during leg cycling (FES-LC) plus arm ergometry (AE) administered using behavioral motivational strategies, and testosterone therapy, is more efficacious than FES-LC plus AE and placebo in improving aerobic capacity, musculoskeletal health, function, metabolism, and wellbeing in SCI.

METHODS

This single-site, randomized, placebo-controlled, parallel group trial will enroll 88 community-dwelling men and women, 19 to 70 years of age, with cervical and thoracic level of SCI, ASIA Impairment Scale grade: A, B, C, or D, 6 months or later after an SCI. Participants randomized to the multimodality intervention will undergo 16 weeks of home-based FES-LC and AE training plus testosterone undecanoate. Testosterone undecanoate injections will be administered by study staff in clinic or by a visiting nurse in the participant's home. The control group will receive 16 weeks of home-based FES-LC and AE exercise plus placebo injections. The primary outcome of this trial is peak aerobic capacity, measured during an incremental exercise testing protocol. Secondary outcomes include whole body and regional lean and adipose tissue mass; muscle strength and power; insulin sensitivity, lipids, and inflammatory markers; SCI functional index and wellbeing (mood, anxiety, pain, life satisfaction and depressive symptoms); and safety.

DISCUSSION

We anticipate that a multimodality intervention that simultaneously addresses multiple physiological impairments in SCI will result in increased aerobic capacity and greater improvements in other musculoskeletal, metabolic, functional and patient-reported outcomes compared to the control intervention. The findings of this study will have important implications for improving the care of people living with an SCI.

TRIAL REGISTRATION

ClinicalTrials.gov :  ( NCT03576001 ). Prospectively registered: July 3, 2018.

[Interference of P2X4 receptor expression in tumor-associated macrophages suppresses migration and invasion of glioma cells]

OBJECTIVE

To investigate the effect of interference of P2X4 receptor expression in tumor-associated macrophages (TAMs) on invasion and migration of glioma cells.

METHODS

C57BL/6 mouse models bearing gliomas in the caudate nucleus were examined for glioma pathology with HE staining and expressions of Iba-1 and P2X4 receptor with immunofluorescence assay. RAW264.7 cells were induced into TAMs using conditioned medium from GL261 cells, and the changes in mRNA expressions of macrophage polarization-related markers and the mRNA and protein expressions of P2X4 receptor were detected with RT-qPCR and Western blotting. The effect of siRNA-mediated P2X4 interference on IL-1β and IL-18 mRNA and protein expressions in the TAMs was detected with RT-qPCR and Western blotting. GL261 cells were cultured in the conditioned medium from the transfected TAMs, and the invasion and migration abilities of the cells were assessed with Transwell invasion and migration experiment.

RESULTS

The glioma tissues from the tumor-bearing mice showed a significantly greater number of Iba-1-positive cells, where an obviously increased P2X4 receptor expression was detected (P=0.001), than the brain tissues of the control mice (P < 0.001). The M2 macrophage markers (Arg-1 and IL-10) and M1 macrophage markers (iNOS and TNF-α) were both significantly up-regulated in the TAMs derived from RAW264.7 cells (all P < 0.01), but the up-regulation of the M2 macrophage markers was more prominent; the expression levels of P2X4 receptor protein and mRNA were both increased in the TAMs (P < 0.05). Interference of P2X4 receptor expression significantly lowered the mRNA(P < 0.01)and protein (P < 0.01, P < 0.05)expression levels of IL-1β and IL-18 in the TAMs and obviously inhibited the ability of the TAMs to promote invasion and migration of the glioma cells (P < 0.05).

CONCLUSION

Interference of P2X4 receptor in the TAMs suppresses the migration and invasion of glioma cells possibly by lowering the expressions of IL-1β and IL-18.

Diagnostic panel of serum miR-125b-5p, miR-182-5p, and miR-200c-3p as non-invasive biomarkers for urothelial bladder cancer

PURPOSE

This study aimed to identify a diagnostic panel of serum microRNAs (miRNAs) for the early detection of bladder cancer (BC).

METHODS

Serum samples were collected from 112 BC patients and 112 normal controls (NCs). A three-stage selection was conducted to identify differentially expressed miRNAs as candidates to construct the diagnostic panel. Further, to explore their potential roles in urothelial BC, bioinformatics analyses, including target genes prediction and functional annotation, were used.

RESULTS

Six downregulated miRNAs (miR-1-3p, miR-30a-5p, miR-100-5p, miR-125b-5p, miR-143-3p, and miR-200c-3p) and one upregulated, miR-182-5p, in BC patients' serum were detected compared to NCs and were selected to establish the diagnostic panel. Based on a backward stepwise logistic regression analysis, miR-125b-5p, miR-182-5p, and miR-200c-3p comprehended the diagnostic panel [area under the curve (AUC) = 0.959, sensitivity = 91.67%, specificity = 92.5%].

CONCLUSION

The panel of three miRNAs had an excellent diagnostic capability, representing a potential non-invasive method for early BC detection.

Self-Supervised Discovering of Interpretable Features for Reinforcement Learning

Deep reinforcement learning (RL) has recently led to many breakthroughs on a range of complex control tasks. However, the agent's decision-making process is generally not transparent. The lack of interpretability hinders the applicability of RL in safety-critical scenarios. While several methods have attempted to interpret vision-based RL, most come without detailed explanation for the agent's behavior. In this paper, we propose a self-supervised interpretable framework, which can discover interpretable features to enable easy understanding of RL agents even for non-experts. Specifically, a self-supervised interpretable network (SSINet) is employed to produce fine-grained attention masks for highlighting task-relevant information, which constitutes most evidence for the agent's decisions. We verify and evaluate our method on several Atari 2600 games as well as Duckietown, which is a challenging self-driving car simulator environment. The results show that our method renders empirical evidences about how the agent makes decisions and why the agent performs well or badly, especially when transferred to novel scenes. Overall, our method provides valuable insight into the internal decision-making process of vision-based RL. In addition, our method does not use any external labelled data, and thus demonstrates the possibility to learn high-quality mask through a self-supervised manner, which may shed light on new paradigms for label-free vision learning such as self-supervised segmentation and detection.

[Identification of rare variants in exons of NLRC4 gene in patients with type 1 diabetes and their impact on gene function]

Objective: To identify rare variants in exon and exon-intron boundary of containing NLR family CARD domain protein 4 (NLRC4) in type 1 diabetes (T1DM) patients, and to explore their effects on gene function. Methods: A total of 508 T1DM patients and 527 healthy controls in the Department of Metabolic Endocrinology, Second Xiangya Hospital of Central South University from August 2017 to September 2020 were selected. The case group included 264 males and 244 females, and the age [M (Q₁, Q₃)] was [27 (11, 43)] years. The control group included 290 males and 237 females, and their age［M（Q₁，Q₃）］was [47 (36, 60)] years old. Identification of rare variants in exons of NLRC4 gene in T1DM patients and healthy controls was performed and verified by next-generation sequencing and sanger sequencing. The NLRC4 gene wild-type and mutant plasmids were constructed and transfected into 293T cells. Western blot (WB) was used to detect the expression of NLRC4 protein and cleavage products of pro-cysteinyl aspartate specific proteinase(procaspase-1). Cycloheximide (CHX) was added to 293T cells transfected with wild-type or mutant NLRC4 plasmid to detect the degradation of NLRC4 protein. The localization of NLRC4 protein was detected by immunofluorescence, and the concentration of IL-1β in the cell supernatant was detected by enzyme-linked immunosorbent assay (ELISA). Results: The sequencing results showed that 4 patients and 2 healthy controls had a heterozygous variant c.208C>T in exon 3 of the NLRC4 gene. Two patient had a heterozygous variant c.1564T>C in exon 4, and 1 patients had c.1219G>C in exon 4. These three variants might be pathogenic variants in T1DM. In 293T cells transfected with NLRC4 wild-type and c.208C>T、c.1564T>Cc.1219G>C mutant plasmids, the expression level, degradation rate, localization of NLRC4 protein and the content of cleavage products of procaspase-1 did not change significantly. However, the concentration of IL-1β secreted by 293T cells transfected with c.1219G>C and c.208C>T plasmid [M(Q₁, Q₃)] was 15.25 (12.98, 17.52) and 15.44 (13.81, 17.07) ng/L, respectively, which was lower than 18.70 (16.59, 20.81) ng/L of 293T cells transfected wild-type plasmid (P=0.020, 0.010). Conclusions: NLRC4 gene rare variants c.208C>T, c.1564T>C and c.1219G>C may not change the protein expression, degradation and localization, but c.208C>T and c.1219G>C may inhibit the secretion of IL-1β. This result suggests that NLRC4 rare variants may have an impact on gene function.