Low-Shot Unsupervised Visual Anomaly Detection via Sparse Feature Representation.

Visual anomaly detection is an essential component in modern industrial manufacturing. Existing studies using notions of pairwise similarity distance between a test feature and nominal features have achieved great breakthroughs. However, the absolute similarity distance lacks certain generalizations, making it challenging to extend the comparison beyond the available samples. This limitation could potentially hamper anomaly detection performance in scenarios with limited samples. This article presents a novel sparse feature representation anomaly detection (SFRAD) framework, which formulates the anomaly detection as a sparse feature representation problem; and notably proposes an anomaly score by orthogonal matching pursuit (ASOMP) as a novel detection metric. Specifically, SFRAD calculates the Gaussian kernel distance between the test feature and its sparse representation in the nominal feature space for anomaly detection. Here, the orthogonal matching pursuit (OMP) algorithm is adopted to achieve the sparse feature representation. Moreover, to construct a low-redundancy memory bank storing the basis features for sparse representation, a novel basis feature sampling (BFS) algorithm is proposed by considering both the maximum coverage and the optimum feature representation simultaneously. As a result, SFRAD incorporates both the advantages of absolute similarity and linear representation; and this enhances the generalization in low-shot scenarios. Extensive experiments on the MVTec anomaly detection (MVTec AD), Kolektor surface-defect dataset (KolektorSDD), Kolektor surface-defect dataset 2 (KolektorSDD2), MVTec logical constraints anomaly detection (MVTec LOCO AD), Visual anomaly (VISA), Modified national institute of standards and technology (MNIST), and CIFAR-10 datasets demonstrate that our proposed SFRAD outperforms the previous methods and achieves state-of-the-art unsupervised anomaly detection performance. Notably, significantly improved outcomes and results have also been achieved on low-shot anomaly detection. Code is available at https://github.com/fanghuisky/SFRAD.

KIF2C is a critical regulator for malignant progression of head and neck squamous cell carcinoma.

Head and neck squamous cell carcinoma (HNSCC) is a significant cause of mortality, while the underlying mechanism remains unclear. Our studies have revealed that KIF2C plays a crucial role in tumor proliferation and metastasis in HNSCC. The results demonstrate that KIF2C is highly expressed at both the mRNA and protein levels and is closely associated with lymph node metastasis. The gene ontology and Kyoto Encyclopedia of Genes and Genomes pathway analyses indicate that the differentially expressed genes are enriched in processes or pathways related to cell adhesion and cell mitosis in HNSCC. Moreover, the established protein-protein interaction network identifies KIF2C as a potential hub gene in HNSCC. Knockdown of KIF2C has been demonstrated to significantly reduce cell migration and invasion ability, leading to cell cycle arrest, a high proportion of abnormal cell apoptosis, and cell chromosome division mismatches in the HNSCC cell line. Downstream genes such as PDGFA, EGFR, TP63, SNAI2, KRT5, and KRT14 were found to be down-regulated, and multiple critical pathways, including mTOR, ERK, and PI3K-AKT pathways, were inactivated as a result of KIF2C knockdown. These findings provide strong evidence for the crucial role of KIF2C in HNSCC and suggest that targeting KIF2C may be a promising therapeutic strategy for this disease. Knockdown of KIF2C has been shown to significantly inhibit tumor proliferation in nude mice, demonstrating the potential therapeutic role of KIF2C in HNSCC treatment.

Expandable-RCNN: toward high-efficiency incremental few-shot object detection.

This study aims at addressing the challenging incremental few-shot object detection (iFSOD) problem toward online adaptive detection. iFSOD targets to learn novel categories in a sequential manner, and eventually, the detection is performed on all learned categories. Moreover, only a few training samples are available for all sequential novel classes in these situations. In this study, we propose an efficient yet suitably simple framework, Expandable-RCNN, as a solution for the iFSOD problem, which allows online sequentially adding new classes with zero retraining of the base network. We achieve this by adapting the Faster R-CNN to the few-shot learning scenario with two elegant components to effectively address the overfitting and category bias. First, an IOU-aware weight imprinting strategy is proposed to directly determine the classifier weights for incremental novel classes and the background class, which is with zero training to avoid the notorious overfitting issue in few-shot learning. Second, since the above zero-retraining imprinting approach may lead to undesired category bias in the classifier, we develop a bias correction module for iFSOD, named the group soft-max layer (GSL), that efficiently calibrates the biased prediction of the imprinted classifier to organically improve classification performance for the few-shot classes, preventing catastrophic forgetting. Extensive experiments on MS-COCO show that our method can significantly outperform the state-of-the-art method ONCE by 5.9 points in commonly encountered few-shot classes.

Systemic inflammation markers and the prevalence of hypertension: A NHANES cross-sectional study.

Systemic inflammation markers have been highlighted recently as related to cardiac and non-cardiac disorders. However, few studies have estimated pre-diagnostic associations between these markers and hypertension. In the National Health and Nutritional Examination Survey from 1999 to 2010, 22,290 adult participants were included for analysis. We assessed associations between four systemic inflammation markers based on blood cell counts: systemic immune-inflammation index (SII), neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), lymphocyte-to-monocyte ratio (LMR), and hypertension prevalence in multivariate logistic regression analysis with odds ratio (OR) and 95% confidence interval (CI). To further explore their associations, subgroup and sensitivity analyses were performed. In continuous analyses, the ORs for hypertension prevalence per ln-transformed increment in SII and NLR were estimated at 1.115 and 1.087 (95% CI: 1.045-1.188; 1.008-1.173; respectively). Compared to those in the lowest tertiles, the hypertension risks for subjects in the highest SII and NLR tertiles were 1.20 and 1.11 times, respectively. Conversely, we found that PLR and LMR were negatively associated with hypertension prevalence in continuous analyses (1.060, 0.972-1.157; 0.926, 0.845-1.014; respectively), and the highest PLR and LMR tertiles (1.041, 0.959-1.129; 0.943, 0.866-1.028; respectively). Also, subgroup and sensitivity analyses indicated that SII had a greater correlation to hypertension. In conclusion, we find positive associations between SII and NLR and the prevalence of hypertension in this cross-sectional study. Our findings highlight that SII may be a superior systemic inflammation warning marker for hypertension.

[Correlation between Mandarin acceptable noise level and cortical auditory evoked potential in young normal-hearing listeners].

Objective:To investigate the correlation between Mandarin acceptable noise level （M-ANL） and cortical auditory evoked potential （CAEP）, and to explore the possible mechanism leading to individual differences in M-ANL values. Methods:Thirty listeners aged 22-33 years with normal hearing were selected as the study subjects, and the M-ANL test and CAEP test were performed respectively. The most comfortable level （MCL）, maximum background noise level （BNL）, M-ANL and CAEP values of each subject were recorded. The latency of each wave of P1, N1, P2, N2, P300 and the amplitude of P1-N1, P2-N2, P300 in CAEP were recorded for each subject. SPSS 25.0 was used for statistical analysis to explore the correlation between the MCL value, BNL value and M-ANL values and the latency of P1, N1, P2, N2, P300 and P1-N1, P2-N2, P300 amplitudes of CAEP. Results:①The MCL value and M-ANL value were positively correlated with the P2 latency of CAEP, and the correlation coefficients were 0.404 and 0.400, respectively, and the differences were statistically significant （P<0.05）. There was no correlation with P1, N1, N2, and P300 latencies of CAEP （P>0.05）. ②The MCL value, BNL value and M-ANL value had no significant difference with the CAEP wave amplitudes of P1-N1, P2-N2, and P300 （P>0.05）. Conclusion:There was a certain correlation between M-ANL and CAEP in young adults with normal hearing, suggesting that the central auditory cortex might play a potential regulatory role in the background noise tolerance. Individuals with a greater background noise acceptance might have stronger central efferent mechanisms and/or less active central afferent mechanisms.

Passivity-based coupling control for underactuated three-dimensional overhead cranes.

This paper develops a novel Lyapunov function candidate for control of the three-dimensional (3-D) overhead crane, which yields a nonlinear controller to inject active damping. Different from the existing passivity-based controls that employ either the angular displacement or its integral as passive elements, the proposed controller incorporates both of them in a new coupled-dissipation signal, thus significantly enhancing the closed-loop passivity. Owing to the improved passivity, the proposed controller ensures the effective suppression of payload oscillations and robustness. Moreover, the control design is extended with the hyperbolic tangent function to prevent overdriving the trolley. The asymptotic stability is guaranteed by LaSalle's invariance principle. The transit performance of the closed-loop system, including robustness, is validated by numerical simulations.

Dual-axis control of flexure-based motion system for optical fibre transceiver assembly using fixed-order controller.

High-frequency resonant modes appearing in the flexure-based motion systems are the key factors that limit the motion/positioning performance. This paper presents an approach to design the fixed-order (low-order) controller for the flexure-based motion system, which is used for industrial optical fibre transceiver alignment and assembly. The uniqueness of the proposed algorithm is that one single controller can simultaneously stabilize the uncertain high-frequency resonant modes for both x and y-axis motion. Especially, the resulted controller can significantly enhance the tracking and disturbance rejection ability in low frequency. The novelty of the proposed method is mainly twofold. First, the finite-frequency specifications are introduced to robust control of the polytopic systems. As the entire-frequency specifications have the trend to place the dominant poles towards the central polynomial, the finite-frequency specifications largely relax the conservatism for the performance optimization. Second, a new and practical method is proposed to choose the central polynomial by designing a nominal controller, where two different central polynomials are utilized for x and y-axis separately for the practical applications, so that the stability and performance are guaranteed but the conservatism is reduced. Experiments are evaluated on the 2DOF flexure-based motion system, and the results show that the tracking error using designed fixed-order controller has a more than 50% reduction compared with the same-order notch filter based nominal controller, with respect to both the sinusoidal and triangular references.