Multiple-Inputs Convolutional Neural Network for COVID-19 Classification and Critical Region Screening From Chest X-ray Radiographs: Model Development and Performance Evaluation.

Background: The COVID-19 pandemic is becoming one of the largest, unprecedented health crises, and chest X-ray radiography (CXR) plays a vital role in diagnosing COVID-19. However, extracting and finding useful image features from CXRs demand a heavy workload for radiologists. Objective: The aim of this study was to design a novel multiple-inputs (MI) convolutional neural network (CNN) for the classification of COVID-19 and extraction of critical regions from CXRs. We also investigated the effect of the number of inputs on the performance of our new MI-CNN model. Methods: A total of 6205 CXR images (including 3021 COVID-19 CXRs and 3184 normal CXRs) were used to test our MI-CNN models. CXRs could be evenly segmented into different numbers (2, 4, and 16) of individual regions. Each region could individually serve as one of the MI-CNN inputs. The CNN features of these MI-CNN inputs would then be fused for COVID-19 classification. More importantly, the contributions of each CXR region could be evaluated through assessing the number of images that were accurately classified by their corresponding regions in the testing data sets. Results: In both the whole-image and left- and right-lung region of interest (LR-ROI) data sets, MI-CNNs demonstrated good efficiency for COVID-19 classification. In particular, MI-CNNs with more inputs (2-, 4-, and 16-input MI-CNNs) had better efficiency in recognizing COVID-19 CXRs than the 1-input CNN. Compared to the whole-image data sets, the efficiency of LR-ROI data sets showed approximately 4% lower accuracy, sensitivity, specificity, and precision (over 91%). In considering the contributions of each region, one of the possible reasons for this reduced performance was that nonlung regions (eg, region 16) provided false-positive contributions to COVID-19 classification. The MI-CNN with the LR-ROI data set could provide a more accurate evaluation of the contribution of each region and COVID-19 classification. Additionally, the right-lung regions had higher contributions to the classification of COVID-19 CXRs, whereas the left-lung regions had higher contributions to identifying normal CXRs. Conclusions: Overall, MI-CNNs could achieve higher accuracy with an increasing number of inputs (eg, 16-input MI-CNN). This approach could assist radiologists in identifying COVID-19 CXRs and in screening the critical regions related to COVID-19 classifications.

Construction of a Prediction Model for College Students' Psychological Disorders Based on Decision Systems and Improved Neural Networks.

Modeling and prediction of psychological disorders is a hot topic in current research. Neural networks are very important factors in improving the accuracy and precision ratios of the models which are developed for the prediction of the psychological disorders. An upgraded neural network prediction model of psychological diseases was suggested in order to attain an optimum prediction effect of psychological disorders. First, it analyzes the current progress in predicting the psychological barrier, finds the current limitations of various psychological barrier forecast model, collects the historical data of psychological barriers, and introduces the chaos algorithm of mental disorder history data preprocessing, psychological barriers to better mining change characteristic, and then, after pretreatment using neural network to the psychological barriers to learning history data, introduce the grain subgroup algorithm to improve the problems existing in the neural network, establish a prediction model of the optimal psychological barriers, and finally, through the contrast test and other psychological obstacle prediction model, the results depict enhanced neural network psychological barrier prediction accuracy of more than 95%, compared with the contrast model. Precision is improved by more than 5%. At the same time, the psychological barrier modeling time is shorter, improving the psychological barriers to predict. The efficiency has a higher practical application value.

Overcoming Reproductive Compromise Under Heat Stress in Wheat: Physiological and Genetic Regulation, and Breeding Strategy.

The reproductive compromise under heat stress is a major obstacle to achieve high grain yield and quality in wheat worldwide. Securing reproductive success is the key solution to sustain wheat productivity by understanding the physiological mechanism and molecular basis in conferring heat tolerance and utilizing the candidate gene resources for breeding. In this study, we examined the performance on both carbon supply source (as leaf photosynthetic rate) and carbon sink intake (as grain yields and quality) in wheat under heat stress varying with timing, duration, and intensity, and we further surveyed physiological processes from source to sink and the associated genetic basis in regulating reproductive thermotolerance; in addition, we summarized the quantitative trait loci (QTLs) and genes identified for heat stress tolerance associated with reproductive stages. Discovery of novel genes for thermotolerance is made more efficient via the combination of transcriptomics, proteomics, metabolomics, and phenomics. Gene editing of specific genes for novel varieties governing heat tolerance is also discussed.

Machine-learning-assisted spontaneous Raman spectroscopy classification and feature extraction for the diagnosis of human laryngeal cancer.

The early detection of laryngeal cancer significantly increases the survival rates, permits more conservative larynx sparing treatments, and reduces healthcare costs. A non-invasive optical form of biopsy for laryngeal carcinoma can increase the early detection rate, allow for more accurate monitoring of its recurrence, and improve intraoperative margin control. In this study, we evaluated a Raman spectroscopy system for the rapid intraoperative detection of human laryngeal carcinoma. The spectral analysis methods included principal component analysis (PCA), random forest (RF), and one-dimensional (1D) convolutional neural network (CNN) methods. We measured the Raman spectra from 207 normal and 500 tumor sites collected from 10 human laryngeal cancer surgical specimens. Random Forest analysis yielded an overall accuracy of 90.5%, sensitivity of 88.2%, and specificity of 92.8% on average over 10 trials. The 1D CNN demonstrated the highest performance with an accuracy of 96.1%, sensitivity of 95.2%, and specificity of 96.9% on average over 50 trials. In predicting the first three principal components (PCs) of normal and tumor data, both RF and CNN demonstrated high performances, except for the tumor PC2. This is the first study in which CNN-assisted Raman spectroscopy was used to identify human laryngeal cancer tissue with extracted feature weights. The proposed Raman spectroscopy feature extraction approach has not been previously applied to human cancer diagnosis. Raman spectroscopy, as assisted by machine learning (ML) methods, has the potential to serve as an intraoperative, non-invasive tool for the rapid diagnosis of laryngeal cancer and margin detection.

Detection of pancreatic cancer by convolutional-neural-network-assisted spontaneous Raman spectroscopy with critical feature visualization.

Pancreatic cancer is the deadliest cancer type with a five-year survival rate of less than 9%. Detection of tumor margins plays an essential role in the success of surgical resection. However, histopathological assessment is time-consuming, expensive, and labor-intensive. We constructed a lab-designed, hand-held Raman spectroscopic system that could enable intraoperative tissue diagnosis using convolutional neural network (CNN) models to efficiently distinguish between cancerous and normal pancreatic tissue. To our best knowledge, this is the first reported effort to diagnose pancreatic cancer by CNN-aided spontaneous Raman scattering with a lab-developed system designed for intraoperative applications. Classification based on the original one-dimensional (1D) Raman, two-dimensional (2D) Raman images, and the first principal component (PC1) from the principal component analysis on the 2D image, could all achieve high performance: the testing sensitivity, specificity, and accuracy were over 95%, and the area under the curve approached 0.99. Although CNN models often show great success in classification, it has always been challenging to visualize the CNN features in these models, which has never been achieved in the Raman spectroscopy application in cancer diagnosis. By studying individual Raman regions and by extracting and visualizing CNN features from max-pooling layers, we identified critical Raman peaks that could aid in the classification of cancerous and noncancerous tissues. 2D Raman PC1 yielded more critical peaks for pancreatic cancer identification than that of 1D Raman, as the Raman intensity was amplified by 2D Raman PC1. To our best knowledge, the feature visualization was achieved for the first time in the field of CNN-aided spontaneous Raman spectroscopy for cancer diagnosis. Based on these CNN feature peaks and their frequency at specific wavenumbers, pancreatic cancerous tissue was found to contain more biochemical components related to the protein contents (particularly collagen), whereas normal pancreatic tissue was found to contain more lipids and nucleic acid (particularly deoxyribonucleic acid/ribonucleic acid). Overall, the CNN model in combination with Raman spectroscopy could serve as a useful tool for the extraction of key features that can help differentiate pancreatic cancer from a normal pancreas.

Apoptotic Caspases Suppress Type I Interferon Production via the Cleavage of cGAS, MAVS, and IRF3.

Viral infection triggers host defenses through pattern-recognition receptor-mediated cytokine production, inflammasome activation, and apoptosis of the infected cells. Inflammasome-activated caspases are known to cleave cyclic GMP-AMP synthase (cGAS). Here, we found that apoptotic caspases are critically involved in regulating both DNA and RNA virus-triggered host defenses, in which activated caspase-3 cleaved cGAS, MAVS, and IRF3 to prevent cytokine overproduction. Caspase-3 was exclusively required in human cells, whereas caspase-7 was involved only in murine cells to inactivate cGAS, reflecting distinct regulatory mechanisms in different species. Caspase-mediated cGAS cleavage was enhanced in the presence of dsDNA. Alternative MAVS cleavage sites were used to ensure the inactivation of this critical protein. Elevated type I IFNs were detected in caspase-3-deficient cells without any infection. Casp3-/- mice consistently showed increased resistance to viral infection and experimental autoimmune encephalomyelitis. Our results demonstrate that apoptotic caspases control innate immunity and maintain immune homeostasis against viral infection.

MAVS activates TBK1 and IKKε through TRAFs in NEMO dependent and independent manner.

Mitochondrial antiviral-signaling protein (MAVS) transmits signals from RIG-I-like receptors after RNA virus infections. However, the mechanism by which MAVS activates downstream components, such as TBK1 and IKKα/ß, is unclear, although previous work suggests the involvement of NEMO or TBK1-binding proteins TANK, NAP1, and SINTBAD. Here, we report that MAVS-mediated innate immune activation is dependent on TRAFs, partially on NEMO, but not on TBK1-binding proteins. MAVS recruited TBK1/IKKε by TRAFs that were pre-associated with TBK1/IKKε via direct interaction between the coiled-coil domain of TRAFs and the SDD domain of TBK1/IKKε. TRAF2-/-3-/-5-/-6-/- cells completely lost RNA virus responses. TRAFs' E3 ligase activity was required for NEMO activation by synthesizing ubiquitin chains that bound to NEMO for NF-κB and TBK1/IKKε activation. NEMO-activated IKKα/ß were important for TBK1/IKKε activation through IKKα/ß-mediated TBK1/IKKε phosphorylation. Moreover, individual TRAFs differently mediated TBK1/IKKε activation and thus fine-tuned antiviral immunity under physiological conditions.

NEMO-IKKß Are Essential for IRF3 and NF-κB Activation in the cGAS-STING Pathway.

Cytosolic dsDNA activates the cyclic GMP-AMP synthase (cGAS)-stimulator of IFN genes (STING) pathway to produce cytokines, including type I IFNs. The roles of many critical proteins, including NEMO, IKKß, and TBK1, in this pathway are unclear because of the lack of an appropriate system to study. In this article, we report that lower FBS concentrations in culture medium conferred high sensitivities to dsDNA in otherwise unresponsive cells, whereas higher FBS levels abrogated this sensitivity. Based on this finding, we demonstrated genetically that NEMO was critically involved in the cGAS-STING pathway. Cytosolic DNA activated TRIM32 and TRIM56 to synthesize ubiquitin chains that bound NEMO and subsequently activated IKKß. Activated IKKß, but not IKKα, was required for TBK1 and NF-κB activation. In contrast, TBK1 was reciprocally required for NF-κB activation, probably by directly phosphorylating IKKß. Thus, our findings identified a unique innate immune activation cascade in which TBK1-IKKß formed a positive feedback loop to assure robust cytokine production during cGAS-STING activation.

Accelerated progression of Hodgkin's-like lymphomas in golli deficient SJL mice.

Spontaneously occurring lymphomas in SJL mice have many pathological features similar to Hodgkin's lymphoma in humans. The malignant growth of the tumor cells is dependent on the support of host FoxP3(+)CD4(+) regulatory T cells (Tregs). In this study, we report that the ablation of golli protein, a negative regulator of CRAC (calcium release activated calcium) channel, in SJL mice results in an accelerated progression of Hodgkin's-like lymphoma which is accompanied by a facilitated conversion of FoxP3(+) Treg cells. Our results suggest that golli protein might affect the progression of Hodgkin's-like lymphomas through regulating the induction of Treg cells.

Design, synthesis, and biological activities of aromatic gossypol Schiff base derivatives.

A series of aromatic gossypol Schiff bases have been successfully synthesized via a feasible chemical modification. The antiviral activity against tobacco mosaic virus (TMV) of these gossypol Schiff bases has been tested for the first time. The bioassay studies indicated most of these derivatives exhibited excellent anti-TMV activity, in which o-trifluoromethylaniline Schiff base (19) displayed the best antiviral activities. Furthermore, compound 19 exhibited an eminent anti-TMV effect in the field and low toxicity to mice. These results suggest it is a promising candidate for the inhibitor of plant virus.

Transient receptor potential melastatin 4 channel controls calcium signals and dental follicle stem cell differentiation.

Elevations in the intracellular Ca(2+) concentration are a phenomena commonly observed during stem cell differentiation but cease after the process is complete. The transient receptor potential melastatin 4 (TRPM4) is an ion channel that controls Ca(2+) signals in excitable and nonexcitable cells. However, its role in stem cells remains unknown. The aim of this study was to characterize TRPM4 in rat dental follicle stem cells (DFSCs) and to determine its impact on Ca(2+) signaling and the differentiation process. We identified TRPM4 gene expression in DFSCs, but not TRPM5, a closely related channel with similar function. Perfusion of cells with increasing buffered Ca(2+) resulted in a concentration-dependent activation of currents typical for TRPM4, which were also voltage-dependent and had Na(+) conductivity. Molecular suppression with shRNA decreased channel activity and cell proliferation during osteogenesis but not adipogenesis. As a result, enhanced mineralization and phosphatase enzyme activity were observed during osteoblast formation, although DFSCs failed to differentiate into adipocytes. Furthermore, the normal agonist-induced first and secondary phases of Ca(2+) signals were transformed into a gradual and sustained increase which confirmed the channels' ability to control Ca(2+) signaling. Using whole genome microarray analysis, we identified several genes impacted by TRPM4 during DFSC differentiation. These findings suggest an inhibitory role for TRPM4 on osteogenesis while it appears to be required for adipogenesis. The data also provide a potential link between the Ca(2+) signaling pattern and gene expression during stem cell differentiation.

Vasoactive intestinal peptide loss leads to impaired CNS parenchymal T-cell infiltration and resistance to experimental autoimmune encephalomyelitis.

The neuropeptide vasoactive intestinal peptide (VIP) has been shown to inhibit macrophage proinflammatory actions, promote a positive Th2/Th1 balance, and stimulate regulatory T-cell production. The fact that this peptide is highly efficacious in animal models of inflammatory diseases such as collagen-induced arthritis and experimental autoimmune encephalomyelitis (EAE) suggests that the endogenous peptide might normally provide protection against such pathologies. We thus studied the response of VIP-deficient (i.e., VIP KO) mice to myelin oligodendrocyte protein-induced EAE. Surprisingly, VIP KO mice were almost completely resistant to EAE, with delayed onset and mild or absent clinical profile. Despite this, flow cytometric analyses and antigen-rechallenge experiments indicated that myelin oligodendrocyte protein-treated VIP KO mice exhibited robust Th1/Th17 cell inductions and antigen-specific proliferation and cytokine responses. Moreover, adoptive transfer of lymphocytes from immunized VIP KO mice to WT recipients resulted in full-blown EAE, supporting their encephalitogenic potential. In contrast, transfer of encephalitogenic WT cells to VIP KO hosts did not produce EAE, suggesting that loss of VIP specifically affected the effector phase of the disease. Histological analyses indicated that CD4 T cells entered the meningeal and perivascular areas of VIP-deficient mice, but that parenchymal infiltration was strongly impaired. Finally, VIP pretreatment of VIP KO mice before immunization was able to restore their sensitivity to EAE. These results indicate that VIP plays an unanticipated permissive and/or proinflammatory role in the propagation of the inflammatory response in the CNS, a finding with potential therapeutic relevance in autoimmune neuroinflammatory diseases such as multiple sclerosis.

Interleukin-15 receptor is essential to facilitate GABA transmission and hippocampal-dependent memory.

Interleukin-15 (IL15) is a cytokine produced by normal brain, but the functions of the IL15 system in normal adults are not yet clear. The hypothesis that the hippocampal IL15 system is essential for memory consolidation was tested by use of IL15Ralpha knock-out mice in behavioral, biochemical, immunohistological, and electron microscopic analyses. The knock-out mice showed deficits in memory, determined by the Stone T-maze and fear conditioning. In their hippocampi, the concentration of GABA was significantly lower. There were region-specific changes of the GABA-synthesizing enzyme glutamic acid decarboxylase (GAD), with increased GAD-67-immunopositive interneurons in the stratum oriens of the CA1 region of the hippocampus, accompanied by nonsignificant reduction of GAD-67 synapses in the CA3 region. Western blotting showed an increase of GAD-65, but not GAD-67, in the hippocampal homogenate. The ultrastructure of the hippocampus remained intact in the knock-out mice. To further test the hypothesis that IL15 directly modulates GABA turnover by reuptake mechanisms, the dose-response relationship of IL15 on (3)H-GABA uptake was determined in two neuronal cell lines. The effective and nontoxic dose was further used in the synaptosomal uptake studies. IL15 decreased the uptake of (3)H-GABA in synaptosomes from the forebrain of wild-type mice. Consistent with this, IL15Ralpha knock-out mice had increased synaptosomal uptake of (3)H-GABA. Overall, the results show novel functions of a unique cytokine in normal hippocampal activity by regulating GABA transmission.

Transient receptor potential melastatin type 7 channel is critical for the survival of bone marrow derived mesenchymal stem cells.

The transient receptor potential melastatin type 7 channel (TRPM7) is a member of the TRP family of ion channels that is essential for cell proliferation and viability. Mesenchymal stem cells (MSCs) from bone marrow are a potential source for tissue repair due to their ability to differentiate into specialized cells. However, the role of TRPM7 in stem cells is unknown. In this study, we characterized TRPM7 in mouse MSCs using molecular biology, immunocytochemistry, and patch clamp. We also investigated TRPM7 function using a lentiviral vector and specific shRNA to knockdown gene expression. By RT-PCR and immunocytochemistry, we identified TRPM7, but not TRPM6, a close family member with similar function. Electrophysiological recordings during depletion of intracellular Mg(2+) or Mg(2+)-ATP resulted in the development of currents typical for the channel. Furthermore, 2-aminoethoxydiphenyl borate (1 pM-100 microM) inhibited TRPM7 in a concentration-dependent manner. The molecular suppression of TRPM7 significantly decreased MSC proliferation and viability as determined by MTT assay. In addition, TRPM7 gene expression was up-regulated during osteogenesis. These findings demonstrate that TRPM7 is required for MSC survival and perhaps involved in the differentiation process.

Age-associated parallel increase of Foxp3(+)CD4(+) regulatory and CD44(+)CD4(+) memory T cells in SJL/J mice.

Effector/memory T cells (Tem) are required to maintain successful immunity, while regulatory T cells (Treg) are required to prevent excessive/uncontrolled inflammation and/or autoimmunity. Although both Tem and Treg cells are increased during aging, the relationship between the increased proportion of Foxp3(+) Treg cells and CD44(+) Tem cells with aging is not clearly understood. We found in this report that Foxp3(+) Treg cells are increased in parallel with CD44(+) Tem cells in SJL/J mice with aging, and that all Foxp3(+) Treg cells are of CD44(+) Tem phenotype, suggesting that the increased Foxp3(+) Treg cells originated from the expanded pool of CD44(+) Tem cells with aging. Our in vitro kinetic studies further suggested that Foxp3(+) Treg cells are converted through the CD44(+) stage. Furthermore, we observed that although the balance between Foxp3(+) Treg and CD44(+)Foxp3(-) Tem cells remained with aging, the aged mice have higher ratios of both Tem and Treg cells vs. naïve T cells resulting in the "shrunken" naïve T cell pools. Our results suggest that an age-associated imbalance of T cell repertoire is a mechanism that contributes to spontaneous occurrence of Hodgkin's-like lymphoma in aged SJL/J mice.

Increased expression of golli myelin basic proteins enhances calcium influx into oligodendroglial cells.

The myelin basic protein (MBP) gene encodes two families of proteins: the classic MBP constituents of myelin and the golli-MBPs, the function of which is less well understood. Previous work suggests that golli proteins may play a role in Ca2+ homeostasis in oligodendrocytes (OLs) and in T-cells. Overexpression of golli in OL cell lines induces elaboration of sheets and processes. Live imaging of these cells revealed a rapid retraction of the processes and sheets after depolarization with high K+. This phenomenon was associated with a significant increase in [Ca2+]int without changes in cell viability. The results indicated that golli produced its effect through Ca2+ influx, rather than Ca2+ release from intracellular stores. Furthermore, a specific [Ca2+]int chelator (BAPTA) or Cd2+, a specific blocker of voltage-operated Ca2+ channels, abolished the ability of golli to promote process extension in a dose-dependent manner. Analysis of the golli protein identified a myristoylation site at the C terminus of the golli domain, which was essential for the action of golli on Ca2+ influx, suggesting that binding of golli to the plasma membrane is important for modulating Ca2+ homeostasis. High-resolution spatiotemporal analysis along N19 processes revealed higher-amplitude local Ca2+ influx in regions with elevated levels of golli. These findings suggest a key role for golli proteins in regulating voltage-gated Ca2+ channels in OLs during process remodeling. Our observations are consistent with the hypothesis that golli proteins, as a part of a protein complex, modulate Ca2+ influx at the plasma membrane and along OL processes.

Minireview: expression and function of golli protein in immune system.

In this minireview, the author briefly reviews the development of our understanding on the immunological function of golli proteins. In the immune system, in addition to serving as autoantigens, golli proteins have been recently found to regulate T-cell activation directly, thus modulating EAE induction. The evidence that golli proteins function as signal molecules is summarized.

Golli protein negatively regulates store depletion-induced calcium influx in T cells.

Calcium influx is crucial for T cell activation and differentiation. The detailed regulation of this process remains unclear. We report here that golli protein, an alternatively spliced product of the myelin basic protein gene, plays a critical role in regulating calcium influx in T cells. Golli-deficient T cells were hyperproliferative and showed enhanced calcium entry upon T cell receptor stimulation. We further found that golli regulates calcium influx in T cells through the inhibition of the store depletion-induced calcium influx. Mutation of the myristoylation site on golli disrupted its association with the plasma membrane and reversed its inhibitory action on Ca2+ influx, indicating that membrane association of golli was essential for its inhibitory action. These results indicate that golli functions in a unique way to regulate T cell activation through a mechanism involving the modulation of the calcium homeostasis.

Region-specific myelin pathology in mice lacking the golli products of the myelin basic protein gene.

The myelin basic protein (MBP) gene encodes two families of proteins, the classic MBP constituents of myelin and the golli-MBPs, the function of which is less well understood. In this study, targeted ablation of the golli-MBPs, but not the classic MBPs, resulted in a distinct phenotype unlike that of knock-outs (KOs) of the classic MBPs or other myelin proteins. Although the golli KO animals did not display an overt dysmyelinating phenotype, they did exhibit delayed and/or hypomyelination in selected areas of the brain, such as the visual cortex and the optic nerve, as determined by Northern and Western blots and immunohistochemical analysis with myelin protein markers. Hypomyelination in some areas, such as the visual cortex, persisted into adulthood. Ultrastructural analysis of the KOs confirmed both the delay and hypomyelination and revealed abnormalities in myelin structure and in some oligodendrocytes. Abnormal visual-evoked potentials indicated that the hypomyelination in the visual cortex had functional consequences in the golli KO brain. Evidence that the abnormal myelination in these animals was a consequence of intrinsic problems with the oligodendrocyte was indicated by an impaired ability of oligodendrocytes to form myelin sheets in culture and by the presence of abnormal Ca2+ transients in purified cortical oligodendrocytes studied in vitro. The Ca2+ results reported in this study complement previous results implicating golli proteins in modulating intracellular signaling in T-cells. Together, all these findings suggest a role for golli proteins in oligodendrocyte differentiation, migration, and/or myelin elaboration in the brain.