Centralized Space Learning for open-set computer-aided diagnosis.

In computer-aided diagnosis (CAD), diagnosing untrained diseases as known categories will cause serious medical accidents, which makes it crucial to distinguish the new class (open set) meanwhile preserving the known classes (closed set) performance so as to enhance the robustness. However, how to accurately define the decision boundary between known and unknown classes is still an open problem, as unknown classes are never seen during the training process, especially in medical area. Moreover, manipulating the latent distribution of known classes further influences the unknown's and makes it even harder. In this paper, we propose the Centralized Space Learning (CSL) method to address the open-set recognition problem in CADs by learning a centralized space to separate the known and unknown classes with the assistance of proxy images generated by a generative adversarial network (GAN). With three steps, including known space initialization, unknown anchor generation and centralized space refinement, CSL learns the optimized space distribution with unknown samples cluster around the center while the known spread away from the center, achieving a significant identification between the known and the unknown. Extensive experiments on multiple datasets and tasks illustrate the proposed CSL's practicability in CAD and the state-of-the-art open-set recognition performance.

Identification of pediatric respiratory diseases using a fine-grained diagnosis system.

Respiratory diseases, including asthma, bronchitis, pneumonia, and upper respiratory tract infection (RTI), are among the most common diseases in clinics. The similarities among the symptoms of these diseases precludes prompt diagnosis upon the patients' arrival. In pediatrics, the patients' limited ability in expressing their situation makes precise diagnosis even harder. This becomes worse in primary hospitals, where the lack of medical imaging devices and the doctors' limited experience further increase the difficulty of distinguishing among similar diseases. In this paper, a pediatric fine-grained diagnosis-assistant system is proposed to provide prompt and precise diagnosis using solely clinical notes upon admission, which would assist clinicians without changing the diagnostic process. The proposed system consists of two stages: a test result structuralization stage and a disease identification stage. The first stage structuralizes test results by extracting relevant numerical values from clinical notes, and the disease identification stage provides a diagnosis based on text-form clinical notes and the structured data obtained from the first stage. A novel deep learning algorithm was developed for the disease identification stage, where techniques including adaptive feature infusion and multi-modal attentive fusion were introduced to fuse structured and text data together. Clinical notes from over 12000 patients with respiratory diseases were used to train a deep learning model, and clinical notes from a non-overlapping set of about 1800 patients were used to evaluate the performance of the trained model. The average precisions (AP) for pneumonia, RTI, bronchitis and asthma are 0.878, 0.857, 0.714, and 0.825, respectively, achieving a mean AP (mAP) of 0.819. These results demonstrate that our proposed fine-grained diagnosis-assistant system provides precise identification of the diseases.

P-ODN: Prototype-based Open Deep Network for Open Set Recognition.

Most of the existing recognition algorithms are proposed for closed set scenarios, where all categories are known beforehand. However, in practice, recognition is essentially an open set problem. There are categories we know called "knowns", and there are more we do not know called "unknowns". Enumerating all categories beforehand is never possible, consequently, it is infeasible to prepare sufficient training samples for those unknowns. Applying closed set recognition methods will naturally lead to unseen-category errors. To address this problem, we propose the prototype-based Open Deep Network (P-ODN) for open set recognition tasks. Specifically, we introduce prototype learning into open set recognition. Prototypes and prototype radiuses are trained jointly to guide a CNN network to derive more discriminative features. Then P-ODN detects the unknowns by applying a multi-class triplet thresholding method based on the distance metric between features and prototypes. Manual labeling the unknowns which are detected in the previous process as new categories. Predictors for new categories are added to the classification layer to "open" the deep neural networks to incorporate new categories dynamically. The weights of new predictors are initialized exquisitely by applying a distances based algorithm to transfer the learned knowledge. Consequently, this initialization method speeds up the fine-tuning process and reduce the samples needed to train new predictors. Extensive experiments show that P-ODN can effectively detect unknowns and needs only few samples with human intervention to recognize a new category. In the real world scenarios, our method achieves state-of-the-art performance on the UCF11, UCF50, UCF101 and HMDB51 datasets.

Upper gastrointestinal hemorrhage in acute-on-chronic liver failure: prevalence, characteristics, and impact on prognosis.

BACKGROUND: Upper gastrointestinal hemorrhage (UGH) is a life-threatening complication in patients with cirrhosis; however, data regarding the role of UGH in acute-on-chronic liver failure (ACLF) are limited. METHODS: A prospective, observational cohort study was performed from February 2014 to Mach 2015. RESULTS: UGH was identified in 170 of 492 cirrhotic patients with acute decompensation (AD) at the time of admission. Logistic regression analysis showed that fecal occult blood test positivity was an independent risk factor for UGH in patients with or without ACLF [OR(95%CI): 8.31(4.89-14.10), p < 0.001; and 6.29 (1.48-26.76), p = 0.031]. Other independent risk factors were a history of gastrointestinal bleeding [OR(95% CI): 13.43 (7.17-25.15), p < 0.001], older age [OR(95% CI): 0.98(0.96-0.99), p = 0.003], greater INR level [OR(95% CI): 0.48(0.28-0.81), p = 0.007] in patients without ACLF. Multivariate Cox proportional hazard model analysis indicated that UGH did not increase mortality at different times in cirrhotic patients with acute decompensation. CONCLUSIONS: UGH is a frequent complication in cirrhotic patients with AD, even those with ACLF. Positive fecal occult blood tests and previous GI bleeding were shown to be associated with the risk of UGH. UGH did not significantly increase the risk of mortality in cirrhotic patients with AD or ACLF.

The discovery of a novel eight-mRNA-lncRNA signature predicting survival of hepatocellular carcinoma patients.

Increasing evidence indicates that the expressions of messenger RNAs (mRNAs) and long non-coding RNAs (lncRNAs) undergo a frequent and aberrant change in carcinogenesis and cancer development. But some research was carried out on mRNA-lncRNA signatures for prediction of hepatocellular carcinoma (HCC) prognosis. We aimed to establish an mRNA-lncRNA signature to improve the ability to predict HCC patients' survival. The subjects from the cancer genome atlas (TCGA) data set were randomly divided into two parts: training data set (n = 246) and testing data set (n = 124). Using computational methods, we selected eight gene signatures (five mRNAs and three lncRNAs) to generate the risk score model, which were significantly correlated with overall survival of patients with HCC in both training and testing data set. The signature had the ability to classify the patients in training data set into a high-risk group and low-risk group with significantly different overall survival (hazard ratio = 4.157, 95% confidence interval = 2.648-6.526, P < 0.001). The prognostic value was further validated in testing data set and the entire data set. Further analysis revealed that this signature was independent of tumor stage. In addition, Gene Set Enrichment Analysis suggested that high risk score group was associated with cell proliferation and division related pathways. Finally, we developed a well-performed nomogram integrating the prognostic signature and other clinical information to predict 3- and 5-year overall survival. In conclusion, the prognostic mRNAs and lncRNAs identified in our study indicate their potential role in HCC biogenesis. The risk score model based on the mRNA-lncRNA may be an efficient classification tool to evaluate the prognosis of patients' with HCC.

Prognostic utility of novel biomarkers in acute-on-chronic liver failure (ACLF) associated with hepatitis B: A multicenter prospective study.

AIM: Flare-ups of chronic hepatitis B can sometimes be severe and even progress to acute-on-chronic liver failure (ACLF), with high short-term mortality. A timely estimation of the risk of death should be initiated early. The aim of the present study was to determine whether novel biomarkers add prognostic information beyond current clinical scoring systems. METHODS: Patients with hepatitis B-associated ACLF were prospectively enrolled from five hospitals in China between August 2017 and March 2018. Their plasma was screened for soluble CD163 (sCD163), neutrophil gelatinase-associated lipocalin (NGAL), and copeptin. The association between these biomarkers and mortality was analyzed. The performance of the Model for End-stage Liver Disease, Asian-Pacific Association for the Study of the Liver-ACLF Research Consortium score, and the Chronic Liver Failure Consortium ACLF score, with or without biomarkers, were compared. RESULTS: One hundred fifty one patients were enrolled. Advanced ACLF patients had significantly higher levels than early ACLF individuals of plasma biomarkers sCD163 (P = 0.001), NGAL (P = 0.006), and copeptin (P = 0.049). Thirty-four deaths occurred during the 28-day follow-up period (22.5%). Both sCD163 and NGAL showed a strong independent association with 28-day mortality, whereas copeptin did not. Scoring systems incorporating sCD163 and NGAL had better discrimination and calibration, as measured by area under the receiver operating characteristic curves, the Akaike information criteria, integrated discrimination improvement, and net reclassification improvement. CONCLUSIONS: Soluble CD163 and NGAL are independently associated with short-term mortality in hepatitis B-associated ACLF. Use of a combination of sCD163 and NGAL improves prognostication.

Mechanism of subunit coordination of an AAA+ hexameric molecular nanomachine.

Simian virus 40 large tumor antigen (LT) is both a potent oncogenic protein and an efficient hexameric nanomachine that harnesses the energy from ATP binding/hydrolysis to melt origin DNA and unwind replication forks. However, how the six subunits of the helicase motor coordinate during ATP hydrolysis and DNA unwinding/translocation is unresolved. Here we investigated the subunit coordination mechanisms "binomial distribution mutant doping" experiments in the presence of various DNA substrates. For ATP hydrolysis, we observed multiple coordination modes, ranging from random and semi-random, and semi-coordinated modes, depending on which type of DNA is present. For DNA unwinding, however, the results indicated a fully-coordinated mode for the natural origin-containing duplex DNA, but a semi-coordinated mode for a pre-existing fork-DNA, providing direct evidence for LT to use potentially different mechanisms to unwind the two types of substrates. The results of this study provide insights into DNA translocation and unwinding mechanisms for LT hexameric biomotor. From the clinical editor: The study describes the subunit coordination of simian virus 40 large tumor antigen (LT) showing that multiple mechanisms exist that handle the specific needs of different stages of DNA replication.

A computational analysis of ATP binding of SV40 large tumor antigen helicase motor.

Simian virus 40 large tumor antigen (LTag) is an efficient helicase motor that unwinds and translocates DNA. The DNA unwinding and translocation of LTag is powered by ATP binding and hydrolysis at the nucleotide pocket between two adjacent subunits of an LTag hexamer. Based on the set of high-resolution hexameric structures of LTag helicase in different nucleotide binding states, we simulated a conformational transition pathway of the ATP binding process using the targeted molecular dynamics method and calculated the corresponding energy profile using the linear response approximation (LRA) version of the semi-macroscopic Protein Dipoles Langevin Dipoles method (PDLD/S). The simulation results suggest a three-step process for the ATP binding from the initial interaction to the final tight binding at the nucleotide pocket, in which ATP is eventually "locked" by three pairs of charge-charge interactions across the pocket. Such a "cross-locking" ATP binding process is similar to the binding zipper model reported for the F1-ATPase hexameric motor. The simulation also shows a transition mechanism of Mg2+ coordination to form the Mg-ATP complex during ATP binding, which is accompanied by the large conformational changes of LTag. This simulation study of the ATP binding process to an LTag and the accompanying conformational changes in the context of a hexamer leads to a refined cooperative iris model that has been proposed previously.

Simulating the electrostatic guidance of the vectorial translocations in hexameric helicases and translocases.

The molecular origin of the action of helicases is explored, starting with a model built based on the different X-ray structures of the large tumor antigen (LTag) hexameric helicase and a simplified model containing the ionized phosphate backbones of a single-strand DNA. The coupling between the protein structural changes and the translocation process is quantified using an effective electrostatic free-energy surface for the protein/DNA complex. This surface is then used in Langevin dynamics simulations of the time dependence of the translocation process. Remarkably, the simulated motion along the free-energy surface results in a vectorial translocation of the DNA, consistent with the biological process. The electrostatic energy of the system appears to reproduce the directionality of this process. Thus, we are able to provide a consistent structure-based molecular description of the energetic and dynamics of the translocation process. This analysis may have general implications for relating structural models to translocation directionality in helicases and other DNA translocases.