Risk factors for lumbar disc herniation recurrence after percutaneous endoscopic lumbar discectomy: a meta-analysis of 58 cohort studies.

Recurrent lumbar disc herniation (rLDH) is one of the most serious complications and major causes of surgical failure and paralysis following percutaneous endoscopic lumbar discectomy (PELD). There are reports in the literature on the identification of risk factors associated with rLDH; however, the results are controversial. Therefore, we conducted a meta-analysis to identify risk factors for rLDH among patients following spinal surgery. PubMed, EMBASE, and the Cochrane Library were searched without language restrictions from inception to April 2018 for studies reporting risk factors for LDH recurrence after PELD. MOOSE guidelines were followed in this meta-analysis. We used a random effects model to aggregate odds ratios (ORs) with 95% confidence intervals (CIs). The evidence of observational studies was classified into high quality (class I), medium quality (class II/III), and low quality (class IV) based on the P value of the total sample size and heterogeneity between studies. Fifty-eight studies were identified with a mean follow-up of 38.8 months. Studies with high-quality (class I) evidence showed that postoperative LDH recurrence after PELD was significantly correlated with diabetes (OR, 1.64; 95% CI, 1.14 to 2.31), the protrusion type LDH (OR, 1.62; 95% CI, 1.02 to 2.61), and less experienced surgeons (OR, 1.54; 95% CI, 1.10 to 2.16). Studies with medium-quality (class II or III) evidence showed that postoperative LDH recurrence was significantly correlated with advanced age (OR, 1.11; 95% CI, 1.05 to 1.19), Modic changes (OR, 2.23; 95% CI, 1.53 to 2.29), smoking (OR, 1.31; 95% CI, 1.00 to 1.71), no college education (OR, 1.56; 95% CI, 1.05 to 2.31), obesity (BMI ≥ 25 kg/m2) (OR, 1.66; 95% CI, 1.11 to 2.47), and inappropriate manual labor (OR, 2.18; 95% CI, 1.33 to 3.59). Based on the current literature, eight patient-related and one surgery-related risk factor are predictors of postoperative LDH recurrence after PELD. These findings may help clinicians raise awareness of early intervention for patients at high risk of LDH recurrence after PELD.

Environmental polycyclic aromatic hydrocarbon exposure is associated with low back pain.

Several studies have assessed the influence of several often-ignored environmental factors on low back pain (LBP), but the effects of environmental polycyclic aromatic hydrocarbon (PAH) exposure on LBP are unclear. During the 2001-2004 cycle of the National Health and Nutrition Examination Survey (NHANES), our study was given to a representative sample of US participants older than 20 (N = 2743). Environmental PAH exposure was calculated using urinary PAH metabolite concentrations. Weighted logistic regression was performed to assess the connection between PAH levels and LBP, with mediation analysis utilised to explore the underlying mechanism. Levels of 1-hydroxynaphthalene (1-OHNa), 2-hydroxynaphthalene (2-OHNa) and total PAHs had a statistically significant positive association with LBP. The odds ratios per 1-unit increase for log-transformed levels of urinary 1-OHNa, 2-OHNa, and total PAHs with LBP were 1.01 (95% CI 1.02-1.19), 1.19 (95% CI 1.04-1.36) and 1.16 (95% CI 1.03-1.32), respectively. The results revealed a strong dose-response association between 1-OHNa, 2-OHNa, total PAHs, and LBP risk. Subgroup analysis indicated that 2&3-OHPh may increase the risk of LBP in the lower family income subgroup. Gamma-glutamyl transaminase (GGT), known as a biomarker of oxidative stress, was strongly related to PAHs. The relationship between total PAHs and LBP was mediated in part by GGT. Our study demonstrates associations between environmental PAH exposure and LBP that need more research to determine the precise effects of various PAH compounds on LBP.

FADngs: Federated Learning for Anomaly Detection.

With the increasing demand for data privacy, federated learning (FL) has gained popularity for various applications. Most existing FL works focus on the classification task, overlooking those scenarios where anomaly detection may also require privacy-preserving. Traditional anomaly detection algorithms cannot be directly applied to the FL setting due to false and missing detection issues. Moreover, with common aggregation methods used in FL (e.g., averaging model parameters), the global model cannot keep the capacities of local models in discriminating anomalies deviating from local distributions, which further degrades the performance. For the aforementioned challenges, we propose Federated Anomaly Detection with Noisy Global Density Estimation, and Self-supervised Ensemble Distillation (FADngs). Specifically, FADngs aligns the knowledge of data distributions from each client by sharing processed density functions. Besides, FADngs trains local models in an improved contrastive learning way that learns more discriminative representations specific for anomaly detection based on the shared density functions. Furthermore, FADngs aggregates capacities by ensemble distillation, which distills the knowledge learned from different distributions to the global model. Our experiments demonstrate that the proposed method significantly outperforms state-of-the-art federated anomaly detection methods. We also empirically show that the shared density function is privacy-preserving. The code for the proposed method is provided for research purposes https://github.com/kanade00/Federated_Anomaly_detection.

Ask Questions with Double Hints: Visual Question Generation with Answer-awareness and Region-reference.

The visual question generation (VQG) task aims to generate human-like questions from an image and potentially other side information (e.g. answer type). Previous works on VQG fall in two aspects: i) They suffer from one image to many questions mapping problem, which leads to the failure of generating referential and meaningful questions from an image. ii) They fail to model complex implicit relations among the visual objects in an image and also overlook potential interactions between the side information and image. To address these limitations, we first propose a novel learning paradigm to generate visual questions with answer-awareness and region-reference. Concretely, we aim to ask the right visual questions with Double Hints - textual answers and visual regions of interests, which could effectively mitigate the existing one-to-many mapping issue. Particularly, we develop a simple methodology to self-learn the visual hints without introducing any additional human annotations. Furthermore, to capture these sophisticated relationships, we propose a new double-hints guided Graph-to-Sequence learning framework, which first models them as a dynamic graph and learns the implicit topology end-to-end, and then utilizes a graph-to-sequence model to generate the questions with double hints. Experimental results demonstrate the priority of our proposed method.

The impact of perioperative opioid use on postoperative outcomes following spinal surgery: a meta-analysis of 60 cohort studies with 13 million participants.

BACKGROUND CONTEXT: An important factor for the prognosis of spinal surgery is the perioperative use of opioids. However, the relationship is not clear. PURPOSE: The purpose of this study was to evaluate the effect of perioperative opioid use on the prognosis of patients following spinal surgery. STUDY DESIGN/SETTING: Systematic review and meta-analysis. OUTCOME MEASURES: A meta-analysis was conducted using the random-effects method to calculate pooled odds ratios (ORs) with 95% confidence intervals (CIs). METHODS: The PubMed, Embase, and Cochrane Library databases were systematically searched to find relevant articles that were published until September 2, 2022. The primary outcome was prolonged postoperative opioid use, and secondary outcomes included the length of stay (LOS), reoperation, the time to return to work (RTW), postoperative complications, gastrointestinal complications, new permanent disability, central nervous system events and infection. In addition, subgroup analysis of the primary outcome was conducted to explore the main sources of heterogeneity, and sensitivity analysis of all outcomes was performed to evaluate the stability of the results. RESULTS: A total of 60 cohort studies involving 13,219,228 individuals met the inclusion criteria. Meta-analysis showed that perioperative opioid use was specifically related to prolonged postoperative opioid use (OR 6.91, 95% CI 6.09 to 7.84, p<.01). Furthermore, the results also showed that perioperative opioid use was significantly associated with prolonged LOS (OR 1.74, 95% CI 1.39 to 2.18, p<.01), postoperative complications (OR 1.72, 95% CI 1.26 to 2.36, p<.01), reoperation (OR 2.38, 95% CI 1.85 to 3.07, p<.01), the time to RTW (OR 0.45, 95% CI 0.39 to 0.52, p<.01), gastrointestinal complications (OR 1.39, 95% CI 1.30 to 1.48, p<.01), central nervous system events (OR 1.99, 95% CI 1.21 to 3.27, p=.07) and infection (OR 1.22, 95% CI 1.09 to 1.36, p=.01). These results were corroborated by the trim-and-fill procedure and leave-one-out sensitivity analyses. CONCLUSIONS: Based on the current evidence, patients with perioperative opioid use, in comparison to controls, appear to have prolonged postoperative opioid use, which may increase the risk of poor outcomes including prolonged LOS, complications, reoperation, RTW and so on. However, these results must be carefully interpreted as the number of studies included was small and the studies were statistically heterogeneous. These findings may help clinicians to realize the harmfulness of perioperative use of opioids, reduce the use of prescription opioids, necessarily withdraw before operation or significantly wean to the lowest tolerable preoperative amount, and provide some inspiration for standardizing the use of opioids in the future.

Risk factors affecting spinal fusion: A meta-analysis of 39 cohort studies.

PURPOSE: We performed a meta-analysis to identify risk factors affecting spinal fusion. METHODS: We systematically searched PubMed, Embase, and the Cochrane Library from inception to January 6, 2023, for articles that report risk factors affecting spinal fusion. The pooled odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using fixed-effects models for each factor for which the interstudy heterogeneity I2 was < 50%, while random-effects models were used when the interstudy heterogeneity I2 was ≥ 50%. Using sample size, Egger's P value, and heterogeneity across studies as criteria, we categorized the quality of evidence from observational studies as high-quality (Class I), moderate-quality (Class II or III), or low-quality (Class IV). Furthermore, the trim-and-fill procedure and leave-one-out protocol were conducted to investigate potential sources of heterogeneity and verify result stability. RESULTS: Of the 1,257 citations screened, 39 unique cohort studies comprising 7,145 patients were included in the data synthesis. High-quality (Class I) evidence showed that patients with a smoking habit (OR, 1.57; 95% CI, 1.11 to 2.21) and without the use of bone morphogenetic protein-2 (BMP-2) (OR, 4.42; 95% CI, 3.33 to 5.86) were at higher risk for fusion failure. Moderate-quality (Class II or III) evidence showed that fusion failure was significantly associated with vitamin D deficiency (OR, 2.46; 95% CI, 1.24 to 4.90), diabetes (OR, 3.42; 95% CI, 1.59 to 7.36), allograft (OR, 1.82; 95% CI, 1.11 to 2.96), conventional pedicle screw (CPS) fixation (OR, 4.77; 95% CI, 2.23 to 10.20) and posterolateral fusion (OR, 3.63; 95% CI, 1.25 to 10.49). CONCLUSIONS: Conspicuous risk factors affecting spinal fusion include three patient-related risk factors (smoking, vitamin D deficiency, and diabetes) and four surgery-related risk factors (without the use of BMP-2, allograft, CPS fixation, and posterolateral fusion). These findings may help clinicians strengthen awareness for early intervention in patients at high risk of developing fusion failure.

Risk factors of postoperative delirium following spine surgery: A meta-analysis of 50 cohort studies with 1.1 million participants.

Objectives: Postoperative delirium (POD) is considered to be a common complication of spine surgery. Although many studies have reported the risk factors associated with POD, the results remain unclear. Therefore, we performed a meta-analysis to identify risk factors for POD among patients following spinal surgery. Methods: We systematically searched the PubMed, Embase and the Cochrane Library for relevant articles published from 2006 to February 1, 2023 that reported risk factors associated with the incidence of POD among patients undergoing spinal surgery. The Meta-Analysis of Observational Studies in Epidemiology (MOOSE) guidelines were followed, and random effects models were used to estimate pooled odds ratio (OR) estimates with 95 % confidence intervals (CIs) for each factor. The evidence from observational studies was classified according to Egger's P value, total sample size, and heterogeneity between studies. Results: Of 11,329 citations screened, 50 cohort studies involving 1,182,719 participants met the inclusion criteria. High-quality evidence indicated that POD was associated with hypertension, diabetes mellitus, cardiovascular disease, pulmonary disease, older age (>65 years), patients experiencing substance use disorder (take drug ≥1 month), cerebrovascular disease, kidney disease, neurological disorder, parkinsonism, cervical surgery, surgical site infection, postoperative fever, postoperative urinary tract infection, and admission to the intensive care unit (ICU). Moderate-quality evidence indicated that POD was associated with depression, American Society of Anesthesiologists (ASA) fitness grade (>II), blood transfusion, abnormal potassium, electrolyte disorder, length of stay, inability to ambulate and intravenous fluid volume. Conclusions: Conspicuous risk factors for POD were mainly patient- and surgery-related. These findings help clinicians identify high-risk patients with POD following spinal surgery and recognize the importance of early intervention.

Attribute-driven streaming edge partitioning with reconciliations for distributed graph neural network training.

Current distributed graph training frameworks evenly partition a large graph into small chunks to suit distributed storage, leverage a uniform interface to access neighbors, and train graph neural networks in a cluster of machines to update weights. Nevertheless, they consider a separate design of storage and training, resulting in huge communication costs for retrieving neighborhoods. During the storage phase, traditional heuristic graph partitioning not only suffers from memory overhead because of loading the full graph into the memory but also damages semantically related structures because of its neglecting meaningful node attributes. What is more, in the weight-update phase, directly averaging synchronization is difficult to tackle with heterogeneous local models where each machine's data are loaded from different subgraphs, resulting in slow convergence. To solve these problems, we propose a novel distributed graph training approach, attribute-driven streaming edge partitioning with reconciliations (ASEPR), where the local model loads only the subgraph stored on its own machine to make fewer communications. ASEPR firstly clusters nodes with similar attributes in the same partition to maintain semantic structure and keep multihop neighbor locality. Then streaming partitioning combined with attribute clustering is applied to subgraph assignment to alleviate memory overhead. After local graph neural network training on distributed machines, we deploy cross-layer reconciliation strategies for heterogeneous local models to improve the averaged global model by knowledge distillation and contrastive learning. Extensive experiments conducted on four large graph datasets on node classification and link prediction tasks show that our model outperforms DistDGL, with fewer resource requirements and up to quadruple the convergence speed.

Variational Cross-Graph Reasoning and Adaptive Structured Semantics Learning for Compositional Temporal Grounding.

Temporal grounding is the task of locating a specific segment from an untrimmed video according to a query sentence. This task has achieved significant momentum in the computer vision community as it enables activity grounding beyond pre-defined activity classes by utilizing the semantic diversity of natural language descriptions. The semantic diversity is rooted in the principle of compositionality in linguistics, where novel semantics can be systematically described by combining known words in novel ways (compositional generalization). However, existing temporal grounding datasets are not carefully designed to evaluate the compositional generalizability. To systematically benchmark the compositional generalizability of temporal grounding models, we introduce a new Compositional Temporal Grounding task and construct two new dataset splits, i.e., Charades-CG and ActivityNet-CG. We empirically find that they fail to generalize to queries with novel combinations of seen words. We argue that the inherent compositional structure (i.e., composition constituents and their relationships) inside the videos and language is the crucial factor to achieve compositional generalization. Based on this insight, we propose a variational cross-graph reasoning framework that explicitly decomposes video and language into hierarchical semantic graphs, respectively, and learns fine-grained semantic correspondence between the two graphs. Meanwhile, we introduce a novel adaptive structured semantics learning approach to derive the structure-informed and domain-generalizable graph representations, which facilitate the fine-grained semantic correspondence reasoning between the two graphs. To further evaluate the understanding of the compositional structure, we also introduce a more challenging setting, where one of the components in the novel composition is unseen. This requires more sophisticated understanding of the compositional structure to infer the potential semantics of the unseen word based on the other learned composition constituents appearing in both the video and language context, and their relationships. Extensive experiments validate the superior compositional generalizability of our approach, demonstrating its ability to handle queries with novel combinations of seen words as well as novel words in the testing composition.

Risk factors of epidural hematoma in patients undergoing spinal surgery: a meta-analysis of 29 cohort studies.

OBJECTIVE: The authors conducted this meta-analysis to identify risk factors for spinal epidural haematoma (SEH) among patients following spinal surgery. METHODS: The authors systematically searched Pub: Med, Embase, and the Cochrane Library for articles that reported risk factors associated with the development of SEH in patients undergoing spinal surgery from inception to 2 July 2022. The pooled odds ratio (OR) was estimated using a random-effects model for each investigated factor. The evidence of observational studies was classified as high quality (Class I), moderate quality (Class II or III) and low quality (Class IV) based on sample size, Egger's P value and between-study heterogeneity. In addition, subgroup analyses stratified by study baseline characteristics and leave-one-out sensitivity analyses were performed to explore the potential sources of heterogeneity and the stability of the results. RESULTS: Of 21 791 articles screened, 29 unique cohort studies comprising 150 252 patients were included in the data synthesis. Studies with high-quality evidence showed that older patients (≥60 years) (OR, 1.35; 95% CI, 1.03-1.77) were at higher risk for SEH. Studies with moderate-quality evidence suggested that patients with a BMI greater than or equal to 25 kg/m² (OR, 1.39; 95% CI, 1.10-1.76), hypertension (OR, 1.67; 95% CI, 1.28-2.17), and diabetes (OR, 1.25; 95% CI, 1.01-1.55) and those undergoing revision surgery (OR, 1.92; 95% CI, 1.15-3.25) and multilevel procedures (OR, 5.20; 95% CI, 2.89-9.37) were at higher risk for SEH. Meta-analysis revealed no association between tobacco use, operative time, anticoagulant use or American Society of Anesthesiologists (ASA) classification and SEH. CONCLUSIONS: Obvious risk factors for SEH include four patient-related risk factors, including older age, obesity, hypertension and diabetes, and two surgery-related risk factors, including revision surgery and multilevel procedures. These findings, however, must be interpreted with caution because most of these risk factors had small effect sizes. Nonetheless, they may help clinicians identify high-risk patients to improve prognosis.

EyeMap: a software system for visualizing and analyzing eye movement data in reading.

We have developed EyeMap, a freely available software system for visualizing and analyzing eye movement data specifically in the area of reading research. As compared with similar systems, including commercial ones, EyeMap has more advanced features for text stimulus presentation, interest area extraction, eye movement data visualization, and experimental variable calculation. It is unique in supporting binocular data analysis for unicode, proportional, and nonproportional fonts and spaced and unspaced scripts. Consequently, it is well suited for research on a wide range of writing systems. To date, it has been used with English, German, Thai, Korean, and Chinese. EyeMap is platform independent and can also work on mobile devices. An important contribution of the EyeMap project is a device-independent XML data format for describing data from a wide range of reading experiments. An online version of EyeMap allows researchers to analyze and visualize reading data through a standard Web browser. This facility could, for example, serve as a front-end for online eye movement data corpora.

Boosting RGB-D Saliency Detection by Leveraging Unlabeled RGB Images.

Training deep models for RGB-D salient object detection (SOD) often requires a large number of labeled RGB-D images. However, RGB-D data is not easily acquired, which limits the development of RGB-D SOD techniques. To alleviate this issue, we present a Dual-Semi RGB-D Salient Object Detection Network (DS-Net) to leverage unlabeled RGB images for boosting RGB-D saliency detection. We first devise a depth decoupling convolutional neural network (DDCNN), which contains a depth estimation branch and a saliency detection branch. The depth estimation branch is trained with RGB-D images and then used to estimate the pseudo depth maps for all unlabeled RGB images to form the paired data. The saliency detection branch is used to fuse the RGB feature and depth feature to predict the RGB-D saliency. Then, the whole DDCNN is assigned as the backbone in a teacher-student framework for semi-supervised learning. Moreover, we also introduce a consistency loss on the intermediate attention and saliency maps for the unlabeled data, as well as a supervised depth and saliency loss for labeled data. Experimental results on seven widely-used benchmark datasets demonstrate that our DDCNN outperforms state-of-the-art methods both quantitatively and qualitatively. We also demonstrate that our semi-supervised DS-Net can further improve the performance, even when using an RGB image with the pseudo depth map.

Local-Global Memory Neural Network for Medication Prediction.

Electronic medical records (EMRs) play an important role in medical data mining and sequential data learning. In this article, we propose to use a sequential neural network with dynamic content-based memories to predict future medications, given EMRs. The local-global memory neural network contains two layers of memories: the local memory and the global memory. Particularly, our method learns the hidden knowledge within EMRs by locally remembering individual patterns of a patient (via local memory) and globally remembering group evidence of disease (via global memory). In addition, we show how our model can be modified to classify the hidden states of EMRs from different patients at each time step into different phases that indicate the progressions of medications in terms of a specific disease, in an unsupervised manner. Experimental results on real EMRs data sets show that, by learning EMRs with external local and global memories, with regard to a given disease, our model improves the prediction performance compared with several alternative methods.

Cross-modal learning to rank via latent joint representation.

Cross-modal ranking is a research topic that is imperative to many applications involving multimodal data. Discovering a joint representation for multimodal data and learning a ranking function are essential in order to boost the cross-media retrieval (i.e., image-query-text or text-query-image). In this paper, we propose an approach to discover the latent joint representation of pairs of multimodal data (e.g., pairs of an image query and a text document) via a conditional random field and structural learning in a listwise ranking manner. We call this approach cross-modal learning to rank via latent joint representation (CML²R). In CML²R, the correlations between multimodal data are captured in terms of their sharing hidden variables (e.g., topics), and a hidden-topic-driven discriminative ranking function is learned in a listwise ranking manner. The experiments show that the proposed approach achieves a good performance in cross-media retrieval and meanwhile has the capability to learn the discriminative representation of multimodal data.

Media and public reactions toward vaccination during the 'hepatitis B vaccine crisis' in China.

BACKGROUND: Public disputations affected vaccine confidence and vaccine rates particularly when adverse events occur. The vigorous development of Internet in China provides an opportunity to observe public reaction and sentiment toward vaccination when Kangtai Hepatitis B vaccine crisis happened and evolved to a widespread debate on the internet from December 12, 2013 to January 3, 2014. METHODS: This study conducted Internet surveillance by examining three daily indicators including the daily number of relevant online news article, Sina Weibo posts and Baidu search index during the crisis. We also analyzed the sentiments of relevant original microblog posts collected from Sina Weibo platform in the crisis. RESULTS: A total of 17 infant deaths were reported to associated with Hepatitis B vaccination. Three major waves of high media and public attention were detected. The daily indicators reached their peaks in the second wave after the relevant vaccine was suspended by the authority (from December 20 to December 29, 2013) with 23,200 daily online news reports, 34,018 Sina Weibo posts and 17,832 Baidu search indices. There were significant correlations between the daily amount of online news, Weibo posts, and Baidu searches (p<.001). The contents analysis suggested 1343 out of 1608 (83.5%) original Weibo posts expressed negative sentiment with almost 90% in the second wave. CONCLUSION: This study found the Kangtai vaccine crisis raised great public attention and negative sentiment toward vaccinations on the internet in China. Policy change such as suspension of the suspected vaccine might trigger even greater reaction and more negative sentiment. The government should provide ways to address emerging public concerns after policy change to avoid misinformation and misunderstanding during such a vaccine crisis.