Search

The purpose of medical image registration is to find geometric transformations that align two medical images so that the corresponding voxels on two images are spatially consistent. Nonrigid medical image registration is a key step in medical image processing, such as image comparison, data fusion, target recognition, and pathological change analysis. Existing registration methods only consider registration accuracy but largely neglect the uncertainty of registration results. In this work, a method based on the Bayesian fully convolutional neural network is proposed for nonrigid medical image registration. The proposed method can generate a geometric uncertainty map to calculate the uncertainty of registration results. This uncertainty can be interpreted as a confidence interval, which is essential for judging whether the source data are abnormal. Moreover, the proposed method introduces group normalization, which is conducive to the network convergence of the Bayesian neural network. Some representative learning-based image registration methods are compared with the proposed method on different image datasets. Experimental results show that the registration accuracy of the proposed method is better than that of the methods, and its antifolding performance is comparable to that of fast image registration and VoxelMorph. Furthermore, the proposed method can evaluate the uncertainty of registration results.

We investigate the influence of local Confucian culture on firms’ commitment to environment, social, and governance (ESG). We hypothesize that local Confucian culture will encourage firms to increase their commitment and effort to ESG practices, as a basic idea of Confucianism is akin to the golden rule, which dictates that one must treat people fairly and well if one wishes to be treated the same manner. Using geographic distances to local Confucian temples and centers to create geographic proximity-based measures of local Confucian culture, we confirm that local Confucian culture in the vicinity of corporate headquarters is positively associated with firms’ ESG performance. Further, we hypothesize and find that the role of local Confucian culture is more prominent in state-owned firms, larger firms, firms with greater media coverage and analyst following, and in more recent years, when environmental, social, and governance practices are more relevant and receive greater attention. Our research shed light on studies evaluating the influence of local cultural norms on investment decisions within the framework of ESG.

Hybrid Automatic Repeat reQuest (HARQ) technology which is widely used in modern communication systems is a combination of ARQ technology and Forward Error Control (FEC) technology. Finding a high-performance FEC coding scheme has a great influence on the performance of HARQ system. Therefore, a new HARQ scheme based on systematic polar codes is designed in this paper. Besides, convolutional coded HARQ scheme, LDPC coded HARQ scheme and Turbo coded HARQ scheme are also simulated in our work as a comparison. In the simulation results, we mainly discuss the following three performance metrics of HARQ system: throughput efficiency, average number of transmission and frame error rate (FER). The simulation results show that, in the low Signal-to-Noise Ratio (SNR) regime, the throughput efficiency gap between the systematic polar coded HARQ scheme and the Turbo coded HARQ scheme is very small. More specifically, the proposed systematic polar coded HARQ scheme is only about 0.5dB away from the Turbo coded HARQ scheme in terms of the throughput efficiency. By the simulation of the average number of transmission, it shows that there is a slight gap between the systematic polar coded HARQ scheme and the Turbo coded HARQ scheme. Besides, as the SNR goes higher, the gap is getting smaller and smaller. The performance of FER of systematic polar coded HARQ scheme is slightly worse than that of Turbo coded HARQ scheme, but better than that of convolutional coded HARQ scheme and LDPC coded HARQ scheme.

The evaluation of bank efficiency is the premise and guidance of high-quality development of the banking industry. On the basis of the existing research on bank efficiency, the operation process of banks is divided into two stages of saving collection and making profits. Based on the conditions of free disposal and the assumption of intermediate connection, the two-stage FDH Production Possibility Set is constructed, and then we build the two-stage FDH model. This model can not only avoid the inconsistency of the estimator of DEA method due to strict assumption but also improve the resolution of the traditional FDH .The efficiency of 57 commercial banks in China from 2013 to 2017 was calculated by using the traditional FDH and two-stage FDH method. The results show that compared with the traditional FDH method, the two-stage FDH method has a great improvement on the resolution to the bank, and can get the benchmark based on the real subsystem of the evaluated bank, and excavate the information of further improving the bank efficiency. The results also show that state-owned banks are more efficient than joint-stock banks, and the efficiency of these two is significantly higher than that of city commercial banks and rural commercial banks.

In many classification problems, the class distribution is imbalanced. Learning from the imbalance data is a remarkable challenge in the knowledge discovery and data mining field. In this paper, we propose a scaling kernel-based support vector machine (SVM) approach to deal with the multi-class imbalanced data classification problem. We first use standard SVM algorithm to gain an approximate hyperplane. Then, we present a scaling kernel function and calculate its parameters using the chi-square test and weighting factors. Experimental results on KEEL data sets show the proposed algorithm can resolve the classifier performance degradation problem due to data skewed distribution and has a good generalization.

A three-dimensional model in rectangular microchannels with constant heat flux is developed to predict steady annular condensation. The condensate flow field on the side wall, which is dominated by surface tension, is divided into two regions: the thin–film region and the meniscus region. The momentum and mass equations, in both the vapor and meniscus regions, along with the film thickness equation in thin–film region are solved numerically. The distribution of the meniscus curvature radius, thickness of the condensate film, heat transfer coefficient, and wall temperature are all determined. The results indicate that with the development of condensation, the condensate in the thin–film assumes a convex profile shape at the side wall, with the crest located at the midpoint of the side wall. The film thickness in the thin-film region increases at upstream locations and decreases as the flow moves downstream. The average heat transfer coefficient in the thin-film region is much larger than that occurring in ...