Learning to See Through Obstructions With Layered Decomposition.

We present a learning-based approach for removing unwanted obstructions, such as window reflections, fence occlusions, or adherent raindrops, from a short sequence of images captured by a moving camera. Our method leverages motion differences between the background and obstructing elements to recover both layers. Specifically, we alternate between estimating dense optical flow fields of the two layers and reconstructing each layer from the flow-warped images via a deep convolutional neural network. This learning-based layer reconstruction module facilitates accommodating potential errors in the flow estimation and brittle assumptions, such as brightness consistency. We show that the proposed approach learned from synthetically generated data performs well to real images. Experimental results on numerous challenging scenarios of reflection and fence removal demonstrate the effectiveness of the proposed method.

Clinical Characteristics and Prognosis of HER2 Gene Phenotype in Patients with Non-Small Cell Lung Cancer.

INTRODUCTION: We aim to investigate the relationship between HER2 gene phenotype and clinical characteristics, distribution and prognosis of non-small cell lung cancer (NSCLC) patients. METHODS: A total of 249 NSCLC patients admitted to the oncology department of our hospital from January 2015 to January 2018 were retrospectively analyzed. The clinicopathological information, CT signs, clinical efficacy and long-term prognosis were collected and compared. RESULTS: A total of 249 NSCLC patients underwent HER2 gene testing, 21 of them (8.43%) complied with HER2 alterations [HER2 (+)], and there were significant differences in tumor stages among patients with different HER2 phenotypes (P<0.05). Among 21 NSCLC patients with HER2 (+), HER2 gene mutation was found in 17 patients (81%), and HER2 gene amplification in 4 patients (19%). Among the HER2 mutations, 12 cases (57%) were 20 exon mutations, and 5 cases (19%) were other mutations. Analysis of CT signs showed that border lobulation/burr, necrosis sign and pleural depression were correlated with HER2 gene mutation (P<0.05). The incidence of EGRF mutation in HER (+) patients was significantly lower than that in HER (-) patients (P<0.05), but there was no significant difference in the incidence of ALK gene mutation among different HER phenotypes (P>0.05). The disease control rate of HER2 (+) patients was significantly lower than that of HER2 (-) patients, and the 12-month progression-free survival rate and survival rate of HER2 (+) patients were significantly higher than those of HER2 (-) patients (P<0.05). There was no significant difference in the incidence of ADR among HER2 patients with different phenotypes, but the incidence of ADR (adverse drug reaction) in HER2 (+) patients with Grade 3 or 4 was significantly higher than that in the control group (P<0.05). DISCUSSION: The incidence of HER2 gene mutations in NSCLC patients is relatively low, but it is far commoner in patients with stage IIIB~IV, among which exon 20 mutations are the most prevalent. In CT signs, the lesion lobulated sign/spiculated sign, necrosis signs, and pleural depression signs are related to HER2 gene mutations. In addition, HER2 gene mutations play a crucial role in the clinical prognosis and treatment safety of patients.

Automated Movement Assessment in Stroke Rehabilitation.

We are developing a system for long term Semi-Automated Rehabilitation At the Home (SARAH) that relies on low-cost and unobtrusive video-based sensing. We present a cyber-human methodology used by the SARAH system for automated assessment of upper extremity stroke rehabilitation at the home. We propose a hierarchical model for automatically segmenting stroke survivor's movements and generating training task performance assessment scores during rehabilitation. The hierarchical model fuses expert therapist knowledge-based approaches with data-driven techniques. The expert knowledge is more observable in the higher layers of the hierarchy (task and segment) and therefore more accessible to algorithms incorporating high level constraints relating to activity structure (i.e., type and order of segments per task). We utilize an HMM and a Decision Tree model to connect these high level priors to data driven analysis. The lower layers (RGB images and raw kinematics) need to be addressed primarily through data driven techniques. We use a transformer based architecture operating on low-level action features (tracking of individual body joints and objects) and a Multi-Stage Temporal Convolutional Network(MS-TCN) operating on raw RGB images. We develop a sequence combining these complimentary algorithms effectively, thus encoding the information from different layers of the movement hierarchy. Through this combination, we produce a robust segmentation and task assessment results on noisy, variable and limited data, which is characteristic of low cost video capture of rehabilitation at the home. Our proposed approach achieves 85% accuracy in per-frame labeling, 99% accuracy in segment classification and 93% accuracy in task completion assessment. Although the methodology proposed in this paper applies to upper extremity rehabilitation using the SARAH system, it can potentially be used, with minor alterations, to assist automation in many other movement rehabilitation contexts (i.e., lower extremity training for neurological accidents).

Show, Match and Segment: Joint Weakly Supervised Learning of Semantic Matching and Object Co-Segmentation.

We present an approach for jointly matching and segmenting object instances of the same category within a collection of images. In contrast to existing algorithms that tackle the tasks of semantic matching and object co-segmentation in isolation, our method exploits the complementary nature of the two tasks. The key insights of our method are two-fold. First, the estimated dense correspondence fields from semantic matching provide supervision for object co-segmentation by enforcing consistency between the predicted masks from a pair of images. Second, the predicted object masks from object co-segmentation in turn allow us to reduce the adverse effects due to background clutters for improving semantic matching. Our model is end-to-end trainable and does not require supervision from manually annotated correspondences and object masks. We validate the efficacy of our approach on five benchmark datasets: TSS, Internet, PF-PASCAL, PF-WILLOW, and SPair-71k, and show that our algorithm performs favorably against the state-of-the-art methods on both semantic matching and object co-segmentation tasks.

The Interaction Between N6-Methyladenosine Modification and Non-Coding RNAs in Gastrointestinal Tract Cancers.

N6-methyladenosine (m6A) is the most common epigenetic modification of eukaryotic RNA, which can participate in the growth and development of the body and a variety of physiological and disease processes by affecting the splicing, processing, localization, transport, translation, and degradation of RNA. Increasing evidence shows that non-coding RNAs, particularly microRNA, long non-coding RNA, and circular RNA, can also regulate the RNA m6A modification process by affecting the expression of m6A-related enzymes. The interaction between m6A modification and non-coding RNAs provides a new perspective for the exploration of the potential mechanism of tumor genesis and development. In this review, we summarize the potential mechanisms and effects of m6A and non-coding RNAs in gastrointestinal tract cancers.

Analysis of N6-Methyladenosine Methylome in Adenocarcinoma of Esophagogastric Junction.

Background: From previous studies, we found that there are more than 100 types of RNA modifications in RNA molecules. m6A methylation is the most common. The incidence rate of adenocarcinoma of the esophagogastric junction (AEG) at home and abroad has increased faster than that of stomach cancer at other sites in recent years. Here, we systematically analyze the modification pattern of m6A mRNA in adenocarcinoma at the esophagogastric junction. Methods: m6A sequencing, RNA sequencing, and bioinformatics analysis were used to describe the m6A modification pattern in adenocarcinoma and normal tissues at the esophagogastric junction. Results: In AEG samples, a total of 4,775 new m6A peaks appeared, and 3,054 peaks disappeared. The unique m6A-related genes in AEG are related to cancer-related pathways. There are hypermethylated or hypomethylated m6A peaks in AEG in differentially expressed mRNA transcripts. Conclusion: This study preliminarily constructed the first m6A full transcriptome map of human AEG. This has a guiding role in revealing the mechanism of m6A-mediated gene expression regulation.

Progressive Representation Adaptation for Weakly Supervised Object Localization.

We address the problem of weakly supervised object localization where only image-level annotations are available for training object detectors. Numerous methods have been proposed to tackle this problem through mining object proposals. However, a substantial amount of noise in object proposals causes ambiguities for learning discriminative object models. Such approaches are sensitive to model initialization and often converge to undesirable local minimum solutions. In this paper, we propose to overcome these drawbacks by progressive representation adaptation with two main steps: 1) classification adaptation and 2) detection adaptation. In classification adaptation, we transfer a pre-trained network to a multi-label classification task for recognizing the presence of a certain object in an image. Through the classification adaptation step, the network learns discriminative representations that are specific to object categories of interest. In detection adaptation, we mine class-specific object proposals by exploiting two scoring strategies based on the adapted classification network. Class-specific proposal mining helps remove substantial noise from the background clutter and potential confusion from similar objects. We further refine these proposals using multiple instance learning and segmentation cues. Using these refined object bounding boxes, we fine-tune all the layer of the classification network and obtain a fully adapted detection network. We present detailed experimental validation on the PASCAL VOC and ILSVRC datasets. Experimental results demonstrate that our progressive representation adaptation algorithm performs favorably against the state-of-the-art methods.

Joint Image Filtering with Deep Convolutional Networks.

Joint image filters leverage the guidance image as a prior and transfer the structural details from the guidance image to the target image for suppressing noise or enhancing spatial resolution. Existing methods either rely on various explicit filter constructions or hand-designed objective functions, thereby making it difficult to understand, improve, and accelerate these filters in a coherent framework. In this paper, we propose a learning-based approach for constructing joint filters based on Convolutional Neural Networks. In contrast to existing methods that consider only the guidance image, the proposed algorithm can selectively transfer salient structures that are consistent with both guidance and target images. We show that the model trained on a certain type of data, e.g., RGB and depth images, generalizes well to other modalities, e.g., flash/non-Flash and RGB/NIR images. We validate the effectiveness of the proposed joint filter through extensive experimental evaluations with state-of-the-art methods.

Robust Visual Tracking via Hierarchical Convolutional Features.

Visual tracking is challenging as target objects often undergo significant appearance changes caused by deformation, abrupt motion, background clutter and occlusion. In this paper, we propose to exploit the rich hierarchical features of deep convolutional neural networks to improve the accuracy and robustness of visual tracking. Deep neural networks trained on object recognition datasets consist of multiple convolutional layers. These layers encode target appearance with different levels of abstraction. For example, the outputs of the last convolutional layers encode the semantic information of targets and such representations are invariant to significant appearance variations. However, their spatial resolutions are too coarse to precisely localize the target. In contrast, features from earlier convolutional layers provide more precise localization but are less invariant to appearance changes. We interpret the hierarchical features of convolutional layers as a nonlinear counterpart of an image pyramid representation and explicitly exploit these multiple levels of abstraction to represent target objects. Specifically, we learn adaptive correlation filters on the outputs from each convolutional layer to encode the target appearance. We infer the maximum response of each layer to locate targets in a coarse-to-fine manner. To further handle the issues with scale estimation and re-detecting target objects from tracking failures caused by heavy occlusion or out-of-the-view movement, we conservatively learn another correlation filter, that maintains a long-term memory of target appearance, as a discriminative classifier. We apply the classifier to two types of object proposals: (1) proposals with a small step size and tightly around the estimated location for scale estimation; and (2) proposals with large step size and across the whole image for target re-detection. Extensive experimental results on large-scale benchmark datasets show that the proposed algorithm performs favorably against the state-of-the-art tracking methods.

Fast and Accurate Image Super-Resolution with Deep Laplacian Pyramid Networks.

Convolutional neural networks have recently demonstrated high-quality reconstruction for single image super-resolution. However, existing methods often require a large number of network parameters and entail heavy computational loads at runtime for generating high-accuracy super-resolution results. In this paper, we propose the deep Laplacian Pyramid Super-Resolution Network for fast and accurate image super-resolution. The proposed network progressively reconstructs the sub-band residuals of high-resolution images at multiple pyramid levels. In contrast to existing methods that involve the bicubic interpolation for pre-processing (which results in large feature maps), the proposed method directly extracts features from the low-resolution input space and thereby entails low computational loads. We train the proposed network with deep supervision using the robust Charbonnier loss functions and achieve high-quality image reconstruction. Furthermore, we utilize the recursive layers to share parameters across as well as within pyramid levels, and thus drastically reduce the number of parameters. Extensive quantitative and qualitative evaluations on benchmark datasets show that the proposed algorithm performs favorably against the state-of-the-art methods in terms of run-time and image quality.

The association between metabolic syndrome and bladder cancer susceptibility and prognosis: an updated comprehensive evidence synthesis of 95 observational studies involving 97,795,299 subjects.

BACKGROUND: The association between metabolic syndrome (MS) and bladder cancer (BC) was not fully investigated, and most primary studies and pooled analyses were only focused on certain specific components. OBJECTIVE: To further investigate this issue and obtain more precise findings, we conducted this updated evidence synthesis of published studies, which involved not only MS components but also the MS in its entirety. MATERIALS AND METHODS: We searched the PubMed, EMBASE, and Web of Science databases for observational studies on the association between BC susceptibility and/or mortality, and MS and its components. We extracted data from included studies, evaluated heterogeneity, and performed meta-analytic quantitative syntheses. RESULTS: A total of 95 studies with 97,795,299 subjects were included in the present study. According to the results, MS significantly increased the risk of BC (risk ratio [RR]=1.11, 95% CI=1.00-1.23); diabetes significantly increased the risk of BC (RR=1.29, 95% CI=1.19-1.39) and associated with poor survival (RR=1.24, 95% CI=1.08-1.43). Excessive body weight was associated with increased susceptibility (RR=1.07, 95% CI=1.02-1.12), recurrence (RR=1.46, 95% CI=1.18-1.81), and mortality (RR=1.17, 95% CI=1.00-1.37). As indicated by cumulative meta-analysis, sample size was inadequate for the association between BC susceptibility and MS, the association between BC recurrence and excessive body weight, and the association between BC survival and diabetes. The sample size of the meta-analysis was enough to reach a stable pooled effect for other associations. CONCLUSION: Diabetes and excessive body weight as components of MS are associated with increased susceptibility and poor prognosis of BC. Uncertainty remains concerning the impact of overall MS, hypertension, and dyslipidemia on BC susceptibility and prognosis, for which further investigations are needed.

Small side branch compromise related to main vessel stenting: A retrospective cohort study comparing different treatment strategies.

Treatment strategies for small side branch compromise related to main vessel stenting are not well investigated and not established.This study is to compare the clinical prognosis of different strategies for bifurcations with or without percutaneous coronary intervention (PCI) of small side branch after it compromised.A total of 119 consecutive bifurcation subjects from January 2013 to March 2015 were enrolled, all bifurcations were characterized by small side branch (1.5 mm ≤side branch diameter ≤2.5 mm). Subjects were assigned into side branch treatment (SBT) group and nonside branch treatment group (NSBT) according to whether advanced treatment of side branch was taken or not after it compromised. Major adverse cardiovascular event (MACE) was evaluated, so were the CCS angina and NYHA heart function classification.SBT subjects were associated with longer procedure time (46.7 vs 19.6 min, P < .001) and more complications (18.9% vs 0.0%, P < .001). 12 MACEs were followed including 4 in SBT group and 8 in NSBT group (10.8% vs 9.8%, P = 1.00). There were no significant difference between 2 groups regarding the CCS and NYHA classification, neither were the calculated classification improvement rate, respectively. In subgroup analysis for true and nontrue bifurcations, no statistical difference was found in terms of the MACE rate, the CCS, and NYHA classification improvement rate.Nontreatment of side branch will not increase the risk of MACE and will not worsen the CCS and NYHA classification when small side branch compromises during the bifurcation PCI.

Structural alterations in the caudate nucleus and precuneus in un-medicated anorexia nervosa patients.

Anorexia nervosa (AN) is a mental disorder characterized by a preoccupation with thinness and may be associated with brain structural alteration. The aim of the study was to examine the brain structural alteration in AN patients, including subcortical structure volume and cortical thickness. Thirty-five un-medicated AN patients and 20 matched healthy controls underwent structural magnetic resonance imaging brain scans. High resolution structural images were acquired on a SIEMENS 3T scanner and preprocessed using FreeSurfer software. Subcortical structure volume and cortical thickness were compared between the two groups. We found larger percentage of caudate volume relative to total grey matter (GM) volume in the AN group. Reduced cortical thickness at the left precuneus was also observed in AN patients. Moreover, an interaction between group and hemisphere was found, suggesting that cortical thinning was more prominent in the left hemisphere in AN patients. These findings provide further evidence for structural brain abnormalities in patients with AN.

Prediction of nonrecurrent laryngeal nerve before thyroid surgery--experience with 1825 cases.

BACKGROUND: Nonrecurrent laryngeal nerve (NRLN) is a rare anatomic anomaly, which often co-occurs with aberrant right subclavian artery (ARSA). With this large case series, we present our experience of predicting the presence of NRLN by the means of chest X-ray film, thoracic computed tomography (CT), and ultrasonography. MATERIALS AND METHODS: A prospective, nonrandomized study has been carried out. A total of 1825 patients with various thyroid disorders scheduled for surgery were recruited between January 2006 and July 2012. All patients underwent preoperative chest X-ray examination. Those suspected with ARSA further underwent thoracic CT scan. Unsuspected patients who had NRLN revealed by surgery were analyzed with ultrasonography postoperatively. RESULTS: A total of 41 patients (2.25%) were suspected to have ARSA by X-ray, of those 19 (46.3%) were confirmed by thoracic CT and proven to have NRLN upon subsequent surgery. No NRLN injury was inflicted. For the remaining 22 cases, CT scan suggested a normal right subclavian artery and none had NRLN upon surgery. For the 1784 unsuspected patients, 4 (0.22%) were discovered to have NRLN upon surgery, of those one was injured. For the 19 predicted NRLN, the time used for identifying the nerve was significantly shorter than the four cases with unsuspected NRLN (t = -15.978; P = 0.000). After the operation, all these unsuspected NRLN were confirmed to have ARSA by ultrasonography. CONCLUSIONS: Patients scheduled for thyroid surgery should be screened for ARSA upon routine chest X-ray and thyroid ultrasonography before surgery. Detection of ARSA can accurately predict the existence of NRLN; hence prevent NRLN injury during subsequent surgery.