A new prediction model for acute kidney injury following liver transplantation using grafts from donors after cardiac death.

Acute kidney injury (AKI) is a major complication following liver transplantation (LT), which utilizes grafts from donors after cardiac death (DCD). We developed a machine-learning-based model to predict AKI, using data from 894 LT recipients (January 2015-March 2021), split into training and testing sets. Five machine learning algorithms were employed to construct the prediction models using 17 clinical variables. The performance of the models was assessed by the area under the receiver operating characteristic curve (AUC), accuracy, F1-score, sensitivity and specificity. The best-performing model was further validated in an independent cohort of 195 LT recipients who received DCD grafts between April 2021 and December 2021. The Shapley additive explanations method was utilized to elucidate the predictions and identify the most crucial features. The gradient boosting machine (GBM) model demonstrated the highest AUC (0.76, 95% CI: 0.70-0.82), F1-score (0.73, 95% CI: 0.66-0.79) and sensitivity (0.74, 95% CI: 0.66-0.80) in the testing set and a comparable AUC (0.75, 95% CI: 0.67-0.81) in the validation set. The GBM model identified high preoperative indirect bilirubin, low intraoperative urine output, prolonged anesthesia duration, low preoperative platelet count and graft steatosis graded NASH Clinical Research Network 1 and above as the top five important features for predicting AKI following LT using DCD grafts. The GBM model is a reliable and interpretable tool for predicting AKI in recipients of LT using DCD grafts. This model can assist clinicians in identifying patients at high risk and providing timely interventions to prevent or mitigate AKI.

Enhancing nursing students' clinical competency using a multi-patient simulation learning model: A randomized controlled study.

BACKGROUND: For nurses, clinical competency is paramount in ensuring that patients receive safe, high-quality care. Multi-patient simulation (MPS) in nursing education is gaining attention, and evidence shows its suitability for real-life situations. MPS can be an effective solution for nurses' continuing clinical education. OBJECTIVES: This project compares the effectiveness of MPS (involving both a standardized patient and a high fidelity simulator) and a single high-fidelity simulation (single HFS; only involving a high fidelity simulator) for enhancing the clinical competency of nursing students. DESIGN: A stratified, permuted, block randomized controlled study design was used. SETTINGS AND PARTICIPANTS: Sixty undergraduate nursing students in years 3, 4, and 5 were selected to participate. Subgroups with each comprising three undergraduate nursing students from different years were formed. METHODS: The participants were randomized to receive either an MPS (intervention group) or single HFS (control group) for 1 day; they later received the same intervention after a 30-day washout period. One objectively measured questionnaire and two self-reported questionnaires were used to measure clinical competency: the Creighton Competency Evaluation Instrument (CCEI), Clinical Competence Questionnaire (CCQ), and Simulation Effectiveness Tool - Modified Questionnaire (SET-M). RESULTS: The results revealed significant between-group differences. Specifically, the intervention group showed greater improvement than the control group in both the CCQ (linear contrast [d] = 71.4; 95 % confidence interval [CI] = 53.407, 89.393; P < 0.001) and CCEI total scores (d = 7.17; 95 % CI = 5.837, 8.503; P < 0.001). The SET-M results indicated that 85 % of the participants (n = 51) strongly agreed that they felt more confident about performing a patient handover to the healthcare team after the simulation. CONCLUSIONS: The study findings indicated that both the MPS and single HFS effectively enhanced students' clinical competency. However, MPSs have superior educational outcomes relative to traditional single HFSs.

Mask-Pyramid Network: A Novel Panoptic Segmentation Method.

In this paper, we introduce a novel panoptic segmentation method called the Mask-Pyramid Network. Existing Mask RCNN-based methods first generate a large number of box proposals and then filter them at each feature level, which requires a lot of computational resources, while most of the box proposals are suppressed and discarded in the Non-Maximum Suppression process. Additionally, for panoptic segmentation, it is a problem to properly fuse the semantic segmentation results with the Mask RCNN-produced instance segmentation results. To address these issues, we propose a new mask pyramid mechanism to distinguish objects and generate much fewer proposals by referring to existing segmented masks, so as to reduce computing resource consumption. The Mask-Pyramid Network generates object proposals and predicts masks from larger to smaller sizes. It records the pixel area occupied by the larger object masks, and then only generates proposals on the unoccupied areas. Each object mask is represented as a H × W × 1 logit, which fits well in format with the semantic segmentation logits. By applying SoftMax to the concatenated semantic and instance segmentation logits, it is easy and natural to fuse both segmentation results. We empirically demonstrate that the proposed Mask-Pyramid Network achieves comparable accuracy performance on the Cityscapes and COCO datasets. Furthermore, we demonstrate the computational efficiency of the proposed method and obtain competitive results.

SVCNet: Scribble-Based Video Colorization Network With Temporal Aggregation.

In this paper, we propose a scribble-based video colorization network with temporal aggregation called SVCNet. It can colorize monochrome videos based on different user-given color scribbles. It addresses three common issues in the scribble-based video colorization area: colorization vividness, temporal consistency, and color bleeding. To improve the colorization quality and strengthen the temporal consistency, we adopt two sequential sub-networks in SVCNet for precise colorization and temporal smoothing, respectively. The first stage includes a pyramid feature encoder to incorporate color scribbles with a grayscale frame, and a semantic feature encoder to extract semantics. The second stage finetunes the output from the first stage by aggregating the information of neighboring colorized frames (as short-range connections) and the first colorized frame (as a long-range connection). To alleviate the color bleeding artifacts, we learn video colorization and segmentation simultaneously. Furthermore, we set the majority of operations on a fixed small image resolution and use a Super-resolution Module at the tail of SVCNet to recover original sizes. It allows the SVCNet to fit different image resolutions at the inference. Finally, we evaluate the proposed SVCNet on DAVIS and Videvo benchmarks. The experimental results demonstrate that SVCNet produces both higher-quality and more temporally consistent videos than other well-known video colorization approaches. The codes and models can be found at https://github.com/zhaoyuzhi/SVCNet.

HSGAN: Hyperspectral Reconstruction From RGB Images With Generative Adversarial Network.

Hyperspectral (HS) reconstruction from RGB images denotes the recovery of whole-scene HS information, which has attracted much attention recently. State-of-the-art approaches often adopt convolutional neural networks to learn the mapping for HS reconstruction from RGB images. However, they often do not achieve high HS reconstruction performance across different scenes consistently. In addition, their performance in recovering HS images from clean and real-world noisy RGB images is not consistent. To improve the HS reconstruction accuracy and robustness across different scenes and from different input images, we present an effective HSGAN framework with a two-stage adversarial training strategy. The generator is a four-level top-down architecture that extracts and combines features on multiple scales. To generalize well to real-world noisy images, we further propose a spatial-spectral attention block (SSAB) to learn both spatial-wise and channel-wise relations. We conduct the HS reconstruction experiments from both clean and real-world noisy RGB images on five well-known HS datasets. The results demonstrate that HSGAN achieves superior performance to existing methods. Please visit https://github.com/zhaoyuzhi/HSGAN to try our codes.

Impact of Perineural Invasion and Preexisting Type 2 Diabetes on Patients with Esophageal Squamous Cell Carcinoma Receiving Neoadjuvant Chemoradiotherapy.

Neoadjuvant chemoradiotherapy (neoCRT) followed by surgery is the cornerstone treatment strategy in locally advanced esophageal squamous cell carcinoma (ESCC). Despite this high- intensity multimodality therapy, most patients still experience recurrences and metastases, especially those who do not achieve a pathological complete response (pCR) after neoCRT. Here, we focused on identifying poor prognostic factors. In this retrospective cohort study; we enrolled 140 patients who completed neoCRT plus surgery treatment sequence with no interval metastasis. Overall, 45 of 140 patients (32.1%) achieved a pCR. The overall survival, disease-free survival (DFS), and metastasis-free survival was significantly better in patients with a pCR than in patients with a non-pCR. In the non-pCR subgroup, the presence of perineural invasion (PNI) and preexisting type 2 diabetes (T2DM) were two factors adversely affecting DFS. After adjusting for other factors, multivariate analysis showed that the hazard ratio (HR) was 2.354 (95% confidence interval [CI] 1.240-4.467, p = 0.009) for the presence of PNI and 2.368 (95% CI 1.351-4.150, p = 0.003) for preexisting T2DM. Patients with a combination of both factors had the worst survival. In conclusion, PNI and preexisting T2DM may adversely affect the prognosis of patients with ESCC receiving neoadjuvant chemoradiotherapy.

M6PR- and EphB4-Rich Exosomes Secreted by Serglycin-Overexpressing Esophageal Cancer Cells Promote Cancer Progression.

Accumulating evidence shows that exosomes participate in cancer progression. However, the functions of cancer cell exosome-transmitted proteins are rarely studied. Previously, we reported that serglycin (SRGN) overexpression promotes invasion and metastasis of esophageal squamous cell carcinoma (ESCC) cells. Here, we investigated the paracrine effects of exosomes from SRGN-overexpressing ESCC cells (SRGN Exo) on ESCC cell invasion and tumor angiogenesis, and used mass spectrometry to identify exosomal proteins involved. Cation-dependent mannose-6-phosphate receptor (M6PR) and ephrin type-B receptor 4 (EphB4) were pronouncedly upregulated in SRGN Exo. Upregulated exosomal M6PR mediated the pro-angiogenic effects of SRGN Exo both in vitro and in vivo, while augmented exosomal EphB4 mediated the pro-invasive effect of SRGN Exo on ESCC cells in vitro. In addition, in vitro studies showed that manipulation of M6PR expression affected the viability and migration of ESCC cells. Both M6PR and EphB4 expression levels were positively correlated with that of SRGN in the serum of patients with ESCC. High level of serum M6PR was associated with poor overall survival rates. Taken together, this study presents the first proof that exosomal M6PR and EphB4 play essential roles in tumor angiogenesis and malignancy, and that serum M6PR is a novel prognostic marker for ESCC patients.

Single-cell multiomics identifies clinically relevant mesenchymal stem-like cells and key regulators for MPNST malignancy.

Malignant peripheral nerve sheath tumor (MPNST), a highly aggressive Schwann cell (SC)-derived soft tissue sarcoma, arises from benign neurofibroma (NF); however, the identity, heterogeneity and origins of tumor populations remain elusive. Nestin+ cells have been implicated as tumor stem cells in MPNST; unexpectedly, single-cell profiling of human NF and MPNST and their animal models reveal a broad range of nestin-expressing SC lineage cells and dynamic acquisition of discrete cancer states during malignant transformation. We uncover a nestin-negative mesenchymal neural crest-like subpopulation as a previously unknown malignant stem-like state common to murine and human MPNSTs, which correlates with clinical severity. Integrative multiomics profiling further identifies unique regulatory networks and druggable targets against the malignant subpopulations in MPNST. Targeting key epithelial-mesenchymal transition and stemness regulators including ZEB1 and ALDH1A1 impedes MPNST growth. Together, our studies reveal the underlying principles of tumor cell-state evolution and their regulatory circuitries during NF-to-MPNST transformation, highlighting a hitherto unrecognized mesenchymal stem-like subpopulation in MPNST disease progression.

Human fetal cerebellar cell atlas informs medulloblastoma origin and oncogenesis.

Medulloblastoma (MB) is the most common malignant childhood brain tumour1,2, yet the origin of the most aggressive subgroup-3 form remains elusive, impeding development of effective targeted treatments. Previous analyses of mouse cerebella3-5 have not fully defined the compositional heterogeneity of MBs. Here we undertook single-cell profiling of freshly isolated human fetal cerebella to establish a reference map delineating hierarchical cellular states in MBs. We identified a unique transitional cerebellar progenitor connecting neural stem cells to neuronal lineages in developing fetal cerebella. Intersectional analysis revealed that the transitional progenitors were enriched in aggressive MB subgroups, including group 3 and metastatic tumours. Single-cell multi-omics revealed underlying regulatory networks in the transitional progenitor populations, including transcriptional determinants HNRNPH1 and SOX11, which are correlated with clinical prognosis in group 3 MBs. Genomic and Hi-C profiling identified de novo long-range chromatin loops juxtaposing HNRNPH1/SOX11-targeted super-enhancers to cis-regulatory elements of MYC, an oncogenic driver for group 3 MBs. Targeting the transitional progenitor regulators inhibited MYC expression and MYC-driven group 3 MB growth. Our integrated single-cell atlases of human fetal cerebella and MBs show potential cell populations predisposed to transformation and regulatory circuitries underlying tumour cell states and oncogenesis, highlighting hitherto unrecognized transitional progenitor intermediates predictive of disease prognosis and potential therapeutic vulnerabilities.

Nondestructive circadian profiling of starch content in fresh intact Arabidopsis leaf with two-photon fluorescence and second-harmonic generation imaging.

Plant chloroplasts conduct photosynthesis to convert solar energy into sugars for the carbon source essential for cell living and growth during the day. One fraction of photosynthetic products is stored in chloroplasts by forming starch granules to continue the provision of carbon energy during the night. Currently, profiling the starch temporal pattern requires either: (i) sacrificing the leaves, or (ii) generating transgenic plants at the risk of changing the metabolisms by incorporating a genetically modified granule-bound starch synthase (GBSS). In this paper, we demonstrated a nondestructive method using two-photon fluorescence (TPF) and second-harmonic generation (SHG) imaging to quantify starch granules within chloroplasts of fresh intact leaves across a day-night cycle. We did so using two Arabidopsis lines having normal and excess starch contents: wild-type (Columbia-0) and starch excess 1 (sex1). The starch granules were visualized by SHG imaging, while the chloroplasts in mesophyll cells were visualized by TPF imaging. Our results provided micron scale spatial resolution of starch distribution within leaves and showed starch circadian patterns consistent with those profiled by enzymatic assays in previous studies. We demonstrated that TPF-SHG imaging is a potential tool for revealing the real-time heterogeneity of starch circadian rhythm in leaf cells, without the need for destructive sample preparation.

ShaTure: Shape and Texture Deformation for Human Pose and Attribute Transfer.

In this paper, we present a novel end-to-end pose transfer framework to transform a source person image to an arbitrary pose with controllable attributes. Due to the spatial misalignment caused by occlusions and multi-viewpoints, maintaining high-quality shape and texture appearance is still a challenging problem for pose-guided person image synthesis. Without considering the deformation of shape and texture, existing solutions on controllable pose transfer still cannot generate high-fidelity texture for the target image. To solve this problem, we design a new image reconstruction decoder - ShaTure which formulates shape and texture in a braiding manner. It can interchange discriminative features in both feature-level space and pixel-level space so that the shape and texture can be mutually fine-tuned. In addition, we develop a new bottleneck module - Adaptive Style Selector (AdaSS) Module which can enhance the multi-scale feature extraction capability by self-recalibration of the feature map through channel-wise attention. Both quantitative and qualitative results show that the proposed framework has superiority compared with the state-of-the-art human pose and attribute transfer methods. Detailed ablation studies report the effectiveness of each contribution, which proves the robustness and efficacy of the proposed framework.

Using Exciton/Trion Dynamics to Spatially Monitor the Catalytic Activities of MoS2 during the Hydrogen Evolution Reaction.

The adsorption and desorption of electrolyte ions strongly modulates the carrier density or carrier type on the surface of monolayer-MoS2 catalyst during the hydrogen evolution reaction (HER). The buildup of electrolyte ions onto the surface of monolayer MoS2 during the HER may also result in the formation of excitons and trions, similar to those observed in gate-controlled field-effect transistor devices. Using the distinct carrier relaxation dynamics of excitons and trions of monolayer MoS2 as sensitive descriptors, an in situ microcell-based scanning time-resolved liquid cell microscope is set up to simultaneously measure the bias-dependent exciton/trion dynamics and spatially map the catalytic activity of monolayer MoS2 during the HER. This operando probing technique used to monitor the interplay between exciton/trion dynamics and electrocatalytic activity for two-dimensional transition metal dichalcogenides provides an excellent platform to investigate the local carrier behaviors at the atomic layer/liquid electrolyte interfaces during electrocatalytic reaction.

Correlation of vacuole morphology with stomatal lineage development by whole-cell electron tomography.

Stomatal movement is essential for plants to optimize transpiration and therefore photosynthesis. Rapid changes in the stomatal aperture are accompanied by adjustment of vacuole volume and morphology in guard cells (GCs). In Arabidopsis (Arabidopsis thaliana) leaf epidermis, stomatal development undergoes a cell-fate transition including four stomatal lineage cells: meristemoid, guard mother cell, young GC, and GC. Little is known about the mechanism underlying vacuole dynamics and vacuole formation during stomatal development. Here, we utilized whole-cell electron tomography (ET) analysis to elucidate vacuole morphology, formation, and development in different stages of stomatal lineage cells at nanometer resolution. The whole-cell ET models demonstrated that large vacuoles were generated from small vacuole stepwise fusion/maturation along stomatal development stages. Further ET analyses verified the existence of swollen intraluminal vesicles inside distinct vacuoles at certain developmental stages of stomatal lineage cells, implying a role of multivesicular body fusion in stomatal vacuole formation. Collectively, our findings demonstrate a mechanism mediating vacuole formation in Arabidopsis stomatal development and may shed light on the role of vacuoles in stomatal movement.

Angular Deep Supervised Vector Quantization for Image Retrieval.

Most of the deep quantization methods adopt unsupervised approaches, and the quantization process usually occurs in the Euclidean space on top of the deep feature and its approximate value. When this approach is applied to the retrieval tasks, since the internal product space of the retrieval process is different from the Euclidean space of quantization, minimizing the quantization error (QE) does not necessarily lead to a good performance on the maximum inner product search (MIPS). To solve these problems, we treat Softmax classification as vector quantization (VQ) with angular decision boundaries and propose angular deep supervised VQ (ADSVQ) for image retrieval. Our approach can simultaneously learn the discriminative feature representation and the updatable codebook, both lying on a hypersphere. To reduce the QE between centroids and deep features, two regularization terms are proposed as supervision signals to encourage the intra-class compactness and inter-class balance, respectively. ADSVQ explicitly reformulates the asymmetric distance computation in MIPS to transform the image retrieval process into a two-stage classification process. Moreover, we discuss the extension of multiple-label cases from the perspective of quantization with binary classification. Extensive experiments demonstrate that the proposed ADSVQ has excellent performance on four well-known image data sets when compared with the state-of-the-art hashing methods.

Nurse-Led Consultation for Patients With Rheumatoid Arthritis at Low Disease Activity: A Randomized Noninferiority Trial.

OBJECTIVE: To determine the effectiveness of nurse-led consultations in patients with stable rheumatoid arthritis (RA) in Hong Kong. METHODS: The present work was a single-center, randomized, open-label, noninferiority trial. Patients who had rheumatoid arthritis (RA) with low disease activity (LDA) were randomized at a 1:1 ratio to attend a nurse-led consultation or rheumatologist follow-up visit for 2 years. The primary end point was the proportion of patients whose RA remained at LDA. Secondary end points included the proportion of patients with RA in disease remission and the scores recorded on the Leeds Satisfaction Questionnaire at 2 years, changes from baseline on the Disease Activity Score in 28 joints using the C-reactive protein level (DAS28-CRP), modified Sharp/van der Heijde score (SHS), Health Assessment Questionnaire disability index (HAQ DI), Short Form 36 (SF-36) physical component score, and 19-item Compliance Questionnaire for Rheumatology (CQR-19) score. RESULTS: Among 280 patients who were randomized equally to either attend nurse-led consultations or rheumatologist follow-up visits, 267 patients completed the study. In the nurse-led consultation and rheumatologist follow-up groups, 92.1% and 91.4% patients, respectively, remained at LDA at 2 years. The 95% confidence intervals (95% CIs) of the adjusted treatment difference were within the predefined noninferiority margin in both the intention-to-treat analysis (95% CI 5.75, 7.15) and the per-protocol analysis (95% CI 1.67, 7.47). Although the changes in DAS28-CRP score over 2 years were significantly different between the 2 treatment groups (P < 0.001), there were no significant changes from baseline in SHS, HAQ DI, SF-36 physical component scores, and CQR-19 scores. At the end of the study, more patients expressed satisfaction with nurse-led consultations. CONCLUSION: Nurse-led consultations were not inferior to rheumatologist follow-up visits in patients with stable RA.

Glioblastoma genetic drivers dictate the function of tumor-associated macrophages/microglia and responses to CSF1R inhibition.

BACKGROUND: Tumor-associated macrophages/microglia (TAMs) are prominent microenvironment components in human glioblastoma (GBM) that are potential targets for anti-tumor therapy. However, TAM depletion by CSF1R inhibition showed mixed results in clinical trials. We hypothesized that GBM subtype-specific tumor microenvironment (TME) conveys distinct sensitivities to TAM targeting. METHODS: We generated syngeneic PDGFB- and RAS-driven GBM models that resemble proneural-like and mesenchymal-like gliomas, and determined the effect of TAM targeting by CSF1R inhibitor PLX3397 on glioma growth. We also investigated the co-targeting of TAMs and angiogenesis on PLX3397-resistant RAS-driven GBM. Using single-cell transcriptomic profiling, we further explored differences in TME cellular compositions and functions in PDGFB- and RAS-driven gliomas. RESULTS: We found that growth of PDGFB-driven tumors was markedly inhibited by PLX3397. In contrast, depletion of TAMs at the early phase accelerated RAS-driven tumor growth and had no effects on other proneural and mesenchymal GBM models. In addition, PLX3397-resistant RAS-driven tumors did not respond to PI3K signaling inhibition. Single-cell transcriptomic profiling revealed that PDGFB-driven gliomas induced expansion and activation of pro-tumor microglia, whereas TAMs in mesenchymal RAS-driven GBM were enriched in pro-inflammatory and angiogenic signaling. Co-targeting of TAMs and angiogenesis decreased cell proliferation and changed the morphology of RAS-driven gliomas. CONCLUSIONS: Our work identifies functionally distinct TAM subpopulations in the growth of different glioma subtypes. Notably, we uncover a potential responsiveness of resistant mesenchymal-like gliomas to combined anti-angiogenic therapy and CSF1R inhibition. These data highlight the importance of characterization of the microenvironment landscape in order to optimally stratify patients for TAM-targeted therapy.

Proteoglycans and their functions in esophageal squamous cell carcinoma.

Esophageal squamous cell carcinoma (ESCC) is a highly malignant disease that has a poor prognosis. Its high lethality is mainly due to the lack of symptoms at early stages, which culminates in diagnosis at a late stage when the tumor has already metastasized. Unfortunately, the common cancer biomarkers have low sensitivity and specificity in esophageal cancer. Therefore, a better understanding of the molecular mechanisms underlying ESCC progression is needed to identify novel diagnostic markers and therapeutic targets for intervention. The invasion of cancer cells into the surrounding tissue is a crucial step for metastasis. During metastasis, tumor cells can interact with extracellular components and secrete proteolytic enzymes to remodel the surrounding tumor microenvironment. Proteoglycans are one of the major components of extracellular matrix. They are involved in multiple processes of cancer cell invasion and metastasis by interacting with soluble bioactive molecules, surrounding matrix, cell surface receptors, and enzymes. Apart from having diverse functions in tumor cells and their surrounding microenvironment, proteoglycans also have diagnostic and prognostic significance in cancer patients. However, the functional significance and underlying mechanisms of proteoglycans in ESCC are not well understood. This review summarizes the proteoglycans that have been studied in ESCC in order to provide a comprehensive view of the role of proteoglycans in the progression of this cancer type. A long term goal would be to exploit these molecules to provide new strategies for therapeutic intervention.

Glutathione S-transferase A2 promotes hepatocellular carcinoma recurrence after liver transplantation through modulating reactive oxygen species metabolism.

Hepatocellular carcinoma (HCC) recurrence after liver transplantation remains a significant clinical problem. Ischemia-reperfusion injury (IRI) occurred inevitably at the early phase after liver transplantation (LT) spawns a significant risk of HCC recurrence. However, their linkage and IRI-derived risk factors for HCC recurrence remain exclusive. Understanding the mechanism of post-transplantation hepatic injury could provide new strategies to prevent the later event of HCC recurrence. We demonstrated that glutathione S-transferase A2 (GSTA2) expression was significantly associated with early phase hepatic and systemic injury and ROS level after liver transplantation. Early phase circulating GSTA2 (EPCGSTA2) protein was a significant predictor of HCC recurrence and survival. Heterogeneous single nucleotide polymorphism at G335C of GSTA2 was significantly associated with poor survival of HCC recipients. Enhancement of GSTA2 could protect HCC cells against H2O2-induced cell death by compensating for the elevated ROS stress. We also demonstrated that GSTA2 played crucial roles in regulating the ROS-associated JNK and AKT signaling pathways and ROS metabolism in HCCs in responding to a dynamic ROS environment. Functionally, endogenous or exogenous upregulation of GSTA2 could promote HCC growth and invasion through activating the epithelial-mesenchymal-transition process. Targeted inhibition of GSTA2 could suppress HCC growth and metastasis. In conclusion, GSTA2 could be a novel prognostic and therapeutic target to combat HCC recurrence after liver transplantation.

ASNet: Auto-Augmented Siamese Neural Network for Action Recognition.

Human action recognition methods in videos based on deep convolutional neural networks usually use random cropping or its variants for data augmentation. However, this traditional data augmentation approach may generate many non-informative samples (video patches covering only a small part of the foreground or only the background) that are not related to a specific action. These samples can be regarded as noisy samples with incorrect labels, which reduces the overall action recognition performance. In this paper, we attempt to mitigate the impact of noisy samples by proposing an Auto-augmented Siamese Neural Network (ASNet). In this framework, we propose backpropagating salient patches and randomly cropped samples in the same iteration to perform gradient compensation to alleviate the adverse gradient effects of non-informative samples. Salient patches refer to the samples containing critical information for human action recognition. The generation of salient patches is formulated as a Markov decision process, and a reinforcement learning agent called SPA (Salient Patch Agent) is introduced to extract patches in a weakly supervised manner without extra labels. Extensive experiments were conducted on two well-known datasets UCF-101 and HMDB-51 to verify the effectiveness of the proposed SPA and ASNet.