Clinical domain knowledge-derived template improves post hoc AI explanations in pneumothorax classification.

OBJECTIVE: Pneumothorax is an acute thoracic disease caused by abnormal air collection between the lungs and chest wall. Recently, artificial intelligence (AI), especially deep learning (DL), has been increasingly employed for automating the diagnostic process of pneumothorax. To address the opaqueness often associated with DL models, explainable artificial intelligence (XAI) methods have been introduced to outline regions related to pneumothorax. However, these explanations sometimes diverge from actual lesion areas, highlighting the need for further improvement. METHOD: We propose a template-guided approach to incorporate the clinical knowledge of pneumothorax into model explanations generated by XAI methods, thereby enhancing the quality of the explanations. Utilizing one lesion delineation created by radiologists, our approach first generates a template that represents potential areas of pneumothorax occurrence. This template is then superimposed on model explanations to filter out extraneous explanations that fall outside the template's boundaries. To validate its efficacy, we carried out a comparative analysis of three XAI methods (Saliency Map, Grad-CAM, and Integrated Gradients) with and without our template guidance when explaining two DL models (VGG-19 and ResNet-50) in two real-world datasets (SIIM-ACR and ChestX-Det). RESULTS: The proposed approach consistently improved baseline XAI methods across twelve benchmark scenarios built on three XAI methods, two DL models, and two datasets. The average incremental percentages, calculated by the performance improvements over the baseline performance, were 97.8% in Intersection over Union (IoU) and 94.1% in Dice Similarity Coefficient (DSC) when comparing model explanations and ground-truth lesion areas. We further visualized baseline and template-guided model explanations on radiographs to showcase the performance of our approach. CONCLUSIONS: In the context of pneumothorax diagnoses, we proposed a template-guided approach for improving model explanations. Our approach not only aligns model explanations more closely with clinical insights but also exhibits extensibility to other thoracic diseases. We anticipate that our template guidance will forge a novel approach to elucidating AI models by integrating clinical domain expertise.

Be Careful When Using Peer-Influence on Nudging Solicitation: Evidence of Potential Negative Effect from a Sample of Chinese University Students.

Purpose: Peer information is now commonly used in solicitation. However, scholars have long focused on testing its effectiveness on increasing the donation amount without paying attention to its potential negative effects on donors. Thus, the current study employs high vs low peer donation amount (HPDA vs LPDA) information to explore its effect on "how-much-to-donate" decisions and the corresponding neural and psychological reactions at the same time. Participants and Methods: Student samples from a Chinese university and behavioral experiments with the event-related potential (ERP) method were used in this study. Results: The behavioral results are consistent with previous research in which HPDA was positively associated with higher donation levels. ERP results show the mechanisms behind decision-making can be summarized into a cognitive approach represented by cost-benefit analysis and an affective approach represented by reward perception. More surprisingly, in contrast to the behavioral results, LPDA elicits higher level of reward perception than HPDA. Conclusion: The results indicate that although HPDA leads to higher levels of donation, donors do not show higher levels of reward anticipation at the neurological level, indicating the increment of donation may come at the cost of donors. Theoretical and practical implications are also discussed.

Ratings or Sales? The Neural and Psychological Processes of Online Experience Product Purchase: Evidence from a Sample of Chinese University Students.

Extrinsic cues are ubiquitous in daily commodity consumption scenarios, not to mention online consumption scenarios. Among the many online cues, monthly sales and product ratings are two of the most representative. Some scholars have researched the impact of these cues on consumer decision making, but only search products have been investigated. Based on previous research, this article expanded the types of products to experience products and further explored consumer purchase behaviours and the underlying purchase processes influenced by these two extrinsic cues with the assistance of a neuroscience tool, event-related potentials (ERPs). The behavioural results indicated that the subjects decided mainly based on ratings, while the effect of sales was continuously inhibited. The ERP results further suggested that consumers recognised low ratings and low sales as more negative stimuli than high ratings and high sales, as larger P2 amplitudes were observed. Following the early processing of these cues, low ratings were considered unacceptable and evoked more significant emotional conflicts than high ratings, which was reflected by larger N400 amplitudes. Moreover, in the late stage, high ratings, which activated evaluation categorisation and produced more significant emotional arousal than low-rating conditions, guided the formation of purchase intention and reflected greater LPP amplitudes. Theoretical and managerial implications were discussed.

The Neural and Psychological Processes of Peer-Influenced Online Donation Decision: An Event-Related Potential Study.

With the rapid development of information and communication technology (ICT), social media-based donation platforms emerged. These platforms innovatively demonstrate peer information (e.g., number of donated peers) on the donation page, which inevitably brings the peer influence into donors' donation decision process. However, how the peer influence will affect the psychological process of donation decisions are remained unknown. This study used the number of donated peers to examine the effects of peer influence on donors' donation decisions and extracted event-related potential (ERP) from electroencephalographic data to explore the underlying psychological process. The behavioral results indicated that the number of donated peers positively influenced donors' willingness to donate. The ERP results suggested that a larger number of donated peers might indicate a higher level of conformity and greater perceived emotional rewards, as a larger P2 amplitude was observed. Following the early processing of emotional stimuli, cognitive detection of decisional risk took place, and the donors reckoned a smaller number of donated peers as a high potential risk, which was reflected by a larger N2 amplitude. In the later stage, the larger number of donated peers, which represented a higher magnitude of prospective emotional rewards, led to a higher incentive to donate, and reflected in a larger amplitude of P3. Additionally, implications and future directions were discussed.

The Influence of Industry Leaders' Behavior on the Decisions of Common Enterprise Leaders in Enterprise Clustering: An Event-Related Potential Study.

PURPOSE: This study aims to explore the influence of industry leaders' behavior on common enterprise leaders' decisions in enterprise clustering by recognizing top executives' cognitive processes of brains. METHODS: Twenty-one real top executives from twelve textile enterprises were recruited in the lab experiment, and decisions about whether entering an industrial zone under two conditions of following an industry leader or a common enterprise were designed as the experiment task. Throughout the formal experimental task, participants' electroencephalograms were recorded. RESULTS: The behavioral results preliminarily proved the effect of industry leaders' behaviors on the real top executives' decisions in common enterprises: participants had a higher acceptance rate with a shorter reaction time in the condition of following an industry leader rather than that in the condition of following a common enterprise. Event-related potential results indicated that choices of following an industry leader led to a more positive perception of emotional valence (reflected by a smaller P2 amplitude) and better evaluation categorization and greater decision confidence (reflected by a larger late positive potential amplitude) than choices of following a common enterprise. CONCLUSION: Top executives from common enterprises tend to evaluate industry leaders' behaviors better than other common enterprises' behaviors, and they tend to make a similar business decision to keep their enterprises consistent with these industry leaders.

Online Attention Accumulation for Weakly Supervised Semantic Segmentation.

Object attention maps generated by image classifiers are usually used as priors for weakly supervised semantic segmentation. However, attention maps usually locate the most discriminative object parts. The lack of integral object localization maps heavily limits the performance of weakly supervised segmentation approaches. This paper attempts to investigate a novel way to identify entire object regions in a weakly supervised manner. We observe that image classifiers' attention maps at different training phases may focus on different parts of the target objects. Based on this observation, we propose an online attention accumulation (OAA) strategy that utilizes the attention maps at different training phases to obtain more integral object regions. Specifically, we maintain a cumulative attention map for each target category in each training image and utilize it to record the discovered object regions at different training phases. Albeit OAA can effectively mine more object regions for most images, for some training images, the range of the attention movement is not large, limiting the generation of integral object attention regions. To overcome this problem, we propose incorporating an attention drop layer into the online attention accumulation process to enlarge the range of attention movement during training explicitly. Our method (OAA) can be plugged into any classification network and progressively accumulate the discriminative regions into cumulative attention maps as the training process goes. Additionally, we also explore utilizing the final cumulative attention maps to serve as the pixel-level supervision, which can further assist the network in discovering more integral object regions. When applying the resulting attention maps to the weakly supervised semantic segmentation task, our approach improves the existing state-of-the-art methods on the PASCAL VOC 2012 segmentation benchmark, achieving a mIoU score of 67.2 percent on the test set.

LayerCAM: Exploring Hierarchical Class Activation Maps for Localization.

The class activation maps are generated from the final convolutional layer of CNN. They can highlight discriminative object regions for the class of interest. These discovered object regions have been widely used for weakly-supervised tasks. However, due to the small spatial resolution of the final convolutional layer, such class activation maps often locate coarse regions of the target objects, limiting the performance of weakly-supervised tasks that need pixel-accurate object locations. Thus, we aim to generate more fine-grained object localization information from the class activation maps to locate the target objects more accurately. In this paper, by rethinking the relationships between the feature maps and their corresponding gradients, we propose a simple yet effective method, called LayerCAM. It can produce reliable class activation maps for different layers of CNN. This property enables us to collect object localization information from coarse (rough spatial localization) to fine (precise fine-grained details) levels. We further integrate them into a high-quality class activation map, where the object-related pixels can be better highlighted. To evaluate the quality of the class activation maps produced by LayerCAM, we apply them to weakly-supervised object localization and semantic segmentation. Experiments demonstrate that the class activation maps generated by our method are more effective and reliable than those by the existing attention methods. The code will be made publicly available.

Delving Deep Into Label Smoothing.

Label smoothing is an effective regularization tool for deep neural networks (DNNs), which generates soft labels by applying a weighted average between the uniform distribution and the hard label. It is often used to reduce the overfitting problem of training DNNs and further improve classification performance. In this paper, we aim to investigate how to generate more reliable soft labels. We present an Online Label Smoothing (OLS) strategy, which generates soft labels based on the statistics of the model prediction for the target category. The proposed OLS constructs a more reasonable probability distribution between the target categories and non-target categories to supervise DNNs. Experiments demonstrate that based on the same classification models, the proposed approach can effectively improve the classification performance on CIFAR-100, ImageNet, and fine-grained datasets. Additionally, the proposed method can significantly improve the robustness of DNN models to noisy labels compared to current label smoothing approaches. The source code is available at our project page: https://mmcheng.net/ols/.

Immunosuppressive receptor LILRB1 acts as a potential regulator in hepatocellular carcinoma by integrating with SHP1.

BACKGROUND: Immunosuppressive receptor LILRB1 regulates tumors progression by transducing immune inhibitory signals via intracellular immunoreceptor tyrosine-based inhibitory motifs. However, its role in Hepatocellular Carcinoma (HCC) remains vague. OBJECTIVE: This study is aimed to disclose the association between LILRB1 and HCC. METHODS: Immunoblotting and qRT-PCR were employed to evaluate the level of LILRB1 in hepatocarcinoma cells. LILRB1-positive cells in tissue array were measured using immunohistochemistry staining. The relation among LILRB1, SHP1 and SHP2 and survival rates were analyzed using Gene Expression Profiling Interactive Analysis (GEPIA) and Oncomine database. RESULTS: LILRB1 was robustly reduced in hepatocarcinoma cells compared to normal cells. Clinically, LILRB1 was significantly higher in 49 of 75 (65%) paired paracarcinoma tissues than that in paired HCC samples. 48 of 75 (64%) HCC subjects in tissue microarray showed low level of LILRB1, compared to 25 of 75 (33%) in paired-adjacent tissues. Oncomine database and GEPIA analysis confirmed that LILRB1 was lower in HCC than normal tissues. Additionally, lowLILRB1 had a significant association with clinicopathological characteristics and Disease Free Survival, but no association with Overall Survival in HCC patients. Mechanismly, positive correlation between LILRB1 and SHP1, but not SHP2 was observed in HCC. CONCLUSIONS: LILRB1 possibly plays an antitumor effect in hepatocarcinoma cells by integrating SHP1, providing evidence that LILRB1 might be involved in the pathologic progression of HCC.

The Emerging Role of Myeloid-Derived Suppressor Cells in the Glioma Immune Suppressive Microenvironment.

Myeloid-derived suppressor cells (MDSCs) are a heterogeneous group of myeloid progenitor and precursor cells at different stages of differentiation, which play an important role in tumor immunosuppression. Glioma is the most common and deadliest primary malignant tumor of the brain, and ample evidence supports key contributions of MDSCs to the immunosuppressive tumor microenvironment, which is a key factor stimulating glioma progression. In this review, we summarize the source and characterization of MDSCs, discuss their immunosuppressive functions, and current approaches that target MDSCs for tumor control. Overall, the review provides insights into the roles of MDSC immunosuppression in the glioma microenvironment and suggests that MDSC control is a powerful cellular therapeutic target for currently incurable glioma tumors.

Glioma in Schizophrenia: Is the Risk Higher or Lower?

Whether persons with schizophrenia have a higher or lower incidence of cancer has been discussed for a long time. Due to the complex mechanisms and characteristics of different types of cancer, it is difficult to evaluate the exact relationship between cancers and schizophrenia without considering the type of tumor. Schizophrenia, a disabling mental illness that is now recognized as a neurodevelopmental disorder, is more correlated with brain tumors, such as glioma, than other types of tumors. Thus, we mainly focused on the relationship between schizophrenia and glioma morbidity. Glioma tumorigenesis and schizophrenia may share similar mechanisms; gene/pathway disruption would affect neurodevelopment and reduce the risk of glioma. The molecular defects of disrupted-in-schizophrenia-1 (DISC1), P53, brain-derived neurotrophic factor (BDNF) and C-X-C chemokine receptors type 4 (CXCR4) involved in schizophrenia pathogenesis might play opposite roles in glioma development. Many microRNAs (miRNAs) such as miR-183, miR-9, miR-137 and miR-126 expression change may be involved in the cross talk between glioma prevalence and schizophrenia. Finally, antipsychotic drugs may have antitumor effects. All these factors show that persons with schizophrenia have a decreased incidence of glioma; therefore, epidemiological investigation and studies comparing genetic and epigenetic aberrations involved in both of these complex diseases should be performed. These studies can provide more insightful knowledge about glioma and schizophrenia pathophysiology and help to determine the target/strategies for the prevention and treatment of the two diseases.

The Protein Encoded by the CCDC170 Breast Cancer Gene Functions to Organize the Golgi-Microtubule Network.

Genome-Wide Association Studies (GWAS) and subsequent fine-mapping studies (>50) have implicated single nucleotide polymorphisms (SNPs) located at the CCDC170/C6ORF97-ESR1 locus (6q25.1) as being associated with the risk of breast cancer. Surprisingly, our analysis using genome-wide differential allele-specific expression (DASE), an indicator for breast cancer susceptibility, suggested that the genetic alterations of CCDC170, but not ESR1, account for GWAS-associated breast cancer risk at this locus. Breast cancer-associated CCDC170 nonsense mutations and rearrangements have also been detected, with the latter being specifically implicated in driving breast cancer. Here we report that the wild type CCDC170 protein localizes to the region of the Golgi apparatus and binds Golgi-associated microtubules (MTs), and that breast cancer-linked truncations of CCDC170 result in loss of Golgi localization. Overexpression of wild type CCDC170 triggers Golgi reorganization, and enhances Golgi-associated MT stabilization and acetyltransferase ATAT1-dependent α-tubulin acetylation. Golgi-derived MTs regulate cellular polarity and motility, and we provide evidence that dysregulation of CCDC170 affects polarized cell migration. Taken together, our findings demonstrate that CCDC170 plays an essential role in Golgi-associated MT organization and stabilization, and implicate a mechanism for how perturbations in the CCDC170 gene may contribute to the hallmark changes in cell polarity and motility seen in breast cancer.

MYOSLID Is a Novel Serum Response Factor-Dependent Long Noncoding RNA That Amplifies the Vascular Smooth Muscle Differentiation Program.

OBJECTIVE: Long noncoding RNAs (lncRNA) represent a growing class of noncoding genes with diverse cellular functions. We previously reported on SENCR, an lncRNA that seems to support the vascular smooth muscle cell (VSMC) contractile phenotype. However, information about the VSMC-specific lncRNAs regulated by myocardin (MYOCD)/serum response factor, the master switch for VSMC differentiation, is unknown. APPROACH AND RESULTS: To define novel lncRNAs with functions related to VSMC differentiation, we performed RNA sequencing in human coronary artery SMCs that overexpress MYOCD. Several novel lncRNAs showed altered expression with MYOCD overexpression and one, named MYOcardin-induced Smooth muscle LncRNA, Inducer of Differentiation (MYOSLID), was activated by MYOCD and selectively expressed in VSMCs. MYOSLID was a direct transcriptional target of both MYOCD/serum response factor and transforming growth factor-ß/SMAD pathways. Functional studies revealed that MYOSLID promotes VSMC differentiation and inhibits VSMC proliferation. MYOSLID showed reduced expression in failed human arteriovenous fistula samples compared with healthy veins. Although MYOSLID did not affect gene expression of transcription factors, such as serum response factor and MYOCD, its depletion in VSMCs disrupted actin stress fiber formation and blocked nuclear translocation of MYOCD-related transcription factor A (MKL1). Finally, loss of MYOSLID abrogated transforming growth factor-ß1-induced SMAD2 phosphorylation. CONCLUSIONS: We have demonstrated that MYOSLID, the first human VSMC-selective and serum response factor/CArG-dependent lncRNA, is a novel modulator in amplifying the VSMC differentiation program, likely through feed-forward actions of both MKL1 and transforming growth factor-ß/SMAD pathways.

Hydrogen peroxide impairs autophagic flux in a cell model of nonalcoholic fatty liver disease.

Nonalcoholic fatty liver disease (NAFLD) has become the leading cause of chronic liver disease, but the pathogenesis of NAFLD is not fully clear. The aim of this study was to determine whether autophagy plays a role in the pathogenesis of NAFLD. We found that the levels of autophagy were elevated in hepatoma cells upon exposure to free fatty acids, as confirmed by the increase in the number of autophagosomes. However, exposure of hepatoma cells to H2O2 and TNF-α, two typical "second hit" factors, increased the initiation of autophagy but inhibited the autophagic flux. The inhibition of autophagy sensitized cells to pro-apoptotic stimuli. Taken together, our results suggest that autophagy acts as a protective mechanism in the pathogenesis of NAFLD and that impairment of autophagy might induce more severe lesions of the liver. These findings will be a benefit to the understanding of the pathogenesis of NAFLD and might suggest a strategy for the prevention and cure of NAFLD.

Serum IL-10 from systemic lupus erythematosus patients suppresses the differentiation and function of monocyte-derived dendritic cells.

The role played by cytokines, other than interferon (IFN)-α, in the differentiation and function of dendritic cells (DCs) in systemic lupus erythematosus (SLE), remains unclear. Serum interleukin-10 (IL-10) levels are generally elevated in SLE patients, which might modulate the differentiation of DCs. In this study, DCs were induced from monocytes either by transendothelial trafficking or by culture with granulocyte-macrophage colony-stimulating factor (GM-CSF) + IL-4 + tumor necrosis factor (TNF)-α. Both systems were used to investigate the effects of elevated serum IL-10 level on DC differentiation in SLE patients. The results showed that monocyte-derived DCs induced by either SLE serum or exogenous IL-10 reduced the expression of human leukocyte antigen (HLA)-DR and CD80, decreased IL-12p40 level, and increased IL-10 level, and exhibited an impaired capacity to stimulate allogenic T-cell proliferation. These results indicate that serum IL-10 may be involved in the pathogenesis of SLE by modulating the differentiation and function of DCs.

Interferon-alpha and interleukin-6 in SLE serum induce the differentiation and maturation of dendritic cells derived from CD34+ hematopoietic precursor cells.

The objective of this study was to investigate the effects of the cytokines IFN-alpha, IL-6 and IL-10 present in the serum of systemic lupus erythematosus (SLE) patients on the differentiation and maturation of DCs derived from CD34(+) hematopoietic precursor cells (HPCs). CD34(+) HPCs cultured in SLE serum containing elevated levels of IFN-alpha or IL-6 differentiated into DCs showing an increased expression of HLA-DR, CD80 and CD86, decreased IL-12 secretion along with increased IL-10 production, and an increased ability to stimulate allogenic T-cell proliferation, compared with DCs cultured in normal serum. DCs cultured with high levels of IFN-alpha increased the proportion of both CD3(+)CD8(-)IFN-gamma(+) and CD3(+)CD8(+)IFN-gamma(+) T-cell subsets, and increased the production of IFN-gamma in the allogenic MLR. DCs cultured with high levels of IL-6 decreased the proportion of both IFN-gamma(+) T-cell subsets, and decreased the secretion of IFN-gamma, but increased the production of IL-10. The IL-10 present in SLE serum did not significantly alter the phenotype or function of the DCs. IFN-alpha and IL-6 present in SLE serum induce CD34(+) HPCs to differentiate into DCs with different regulatory effects on T-cell differentiation, which might be involved in the initiation and maintenance of SLE.