Search

Adversarial training (AT) has been demonstrated to be effective in improving model robustness by leveraging adversarial examples for training. However, most AT methods are in face of expensive time and computational cost for calculating gradients at multiple steps in generating adversarial examples. To boost training efficiency, fast gradient sign method (FGSM) is adopted in fast AT methods by calculating gradient only once. Unfortunately, the robustness is far from satisfactory. One reason may arise from the initialization fashion. Existing fast AT generally uses a random sample-agnostic initialization, which facilitates the efficiency yet hinders a further robustness improvement. Up to now, the initialization in fast AT is still not extensively explored. In this paper, focusing on image classification, we boost fast AT with a sample-dependent adversarial initialization, i.e., an output from a generative network conditioned on a benign image and its gradient information from the target network. As the generative network and the target network are optimized jointly in the training phase, the former can adaptively generate an effective initialization with respect to the latter, which motivates gradually improved robustness. Experimental evaluations on four benchmark databases demonstrate the superiority of our proposed method over state-of-the-art fast AT methods, as well as comparable robustness to advanced multi-step AT methods. The code is released at https://github.com//jiaxiaojunQAQ//FGSM-SDI.

In February 2020, an inpatient in Peking University People's Hospital (PKUPH), China, was confirmed positive for the novel coronavirus. In this case, 143 hemodialysis patients were labeled as close contacts and required to be placed under the hospital-based group medical quarantine (HB-GMQ) for 2 weeks by the authorities. After the case was reported, false or misleading information about the case flourished on social media platforms, which led to infodemic. Under this context, PKUPH adopted patient-centered humanistic care to implement the HB-GMQ, through the synergy of administrative, healthcare, logistical, and other measures under the model of patient-centered care of the Massachusetts Medical Society (MMS). As a result, all the patients tided over the HB-GMQ with no COVID-19 infection and no unanticipated adverse events, and all met the criteria for lifting the HB-GMQ. According to the questionnaires taken during the HB-GMQ, a high level of satisfaction was found among the quarantined and no symptomatic increase of anxiety and depression in the patients before and during the HB-GMQ, by comparing the Zung self-rating anxiety scale (SAS) and self-rating depression scale (SDS) conducted in December 2019 and on the 12th day of the HB-GMQ. This article is to brief on PKUPH's experience in implementing patient-centered humanistic care tailored to hemodialysis patients under the HB-GMQ, and to validate the hypothesis that patient-centered humanistic care is effective and helpful to help them tide over the HB-GMQ, so as to shed light on how to implement the HB-GMQ and cope with the HB-GMQ-induced problems in other hospitals., Competing Interests: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest., (Copyright © 2021 Zhang, Gan, Li, Shao, Zuo, Gao, Huang, Jia, Liang, Hou, Wang, Wang, Gao, Wang and Chen.)

Glioma is a prevalent primary brain tumor. Temozolomide (TMZ) has been used to treat glioma. However, the resistance of TMZ to glioma poses heavy burden to glioma treatment. In this study, the effects of glioma resistance to TMZ and underlying mechanism were revealed. The expression levels of circ-VPS18, microRNA-370 (miR-370) and runt-related transcription factor 1 (RUNX1) were detected by quantitative real-time polymerase chain reaction (qRT-PCR). The protein expression of RUNX1, multidrug resistance-associated protein 1 (MRP1), and multi-drug resistance gene-1 (MDR1) was determined by western blot. The functional effects of circ-VPS18 knockdown on TMZ sensitivity and glioma progression were revealed by cell counting kit-8 proliferation (CCK-8), flow cytometry, and transwell assays. The impacts of circ-VPS18 deletion on TMZ sensitivity in vivo were illustrated by in vivo tumor formation assay. The binding relationship between miR-370 and circ-VPS18 or RUNX1 was predicted by starBase v2.0 online database and identified by dual-luciferase reporter assay. Circ-VPS18 expression and the mRNA and protein levels of RUNX1 were dramatically upregulated, and miR-370 expression was significantly downregulated in glioma cells, TMZ-resistant glioma tissues, or tissue compared with control groups. Functionally, circ-VPS18 knockdown improved TMZ sensitivity, induced cell apoptosis, whereas repressed cell viability, migration and invasion in U251/TR and LN229/TR cells, which was reversed by miR-370 inhibitor. Additionally, RUNX1 overexpression hindered the effects of miR-370 on TMZ sensitivity and glioma progression. Circ-VPS18 knockdown enhanced TMZ sensitivity in vivo. Mechanistically, circ-VPS18 functioned as a sponge of miR-370 and miR-370 targeted RUNX1. Circ-VPS18 knockdown improved TMZ sensitivity and repressed glioma progression by sponging miR-370 to downregulate RUNX1 expression, which provided a new insight in further studying glioma resistance to TMZ.

In this meta-analysis, we evaluated overall strengths of relation between beliefs about pain, health, or illness and problems in functioning (i.e., functional impairment, affective distress, pain severity) in osteoarthritis and rheumatoid arthritis samples as well as moderators of these associations. In sum, 111 samples (N = 17,365 patients) met inclusion criteria. On average, highly significant, medium effect sizes were observed for associations between beliefs and problems in functioning but heterogeneity was also inflated. Effect sizes were not affected by arthritis subtype, gender, or age. However, pain belief content emerged as a significant moderator, with larger effect sizes for studies in which personal incapacity or ineffectiveness in controlling pain was a content theme of belief indices (i.e., pain catastrophizing, helplessness, self-efficacy) compared to those examining locus of control and fear/threat/harm beliefs. Furthermore, analyses of longitudinal study subsets supported the status of pain beliefs risk factors for later problems in functioning in these groups.