Performance Improvement Research Articles

Single/Multi-Source Black-Box Domain Adaption for Sensor Time Series Data.

Unsupervised domain adaption (UDA), which transfers knowledge from a labeled source domain to an unlabeled target domain, has attracted tremendous attention in many machine learning applications. Recently, there have been attempts to apply domain adaption for sensor time series data, such as human activity recognition and gesture recognition. However, existing methods suffer from some drawbacks that hinder further performance improvement. They often require access to source data or source models during training, which is unavailable in some fields because of privacy protection and storage limit. Typically, the source domains may only provide an application programming interface (API) for the target domain to call. On the other hand, current UDA methods have not considered the temporal consistency and low-signal-to-noise ratio (SNR) of sensor time series. To address the challenges, this article presents a black-box domain adaption framework for sensor time series data (B2TSDA). First, we propose a single/multi-source teacher-student learning framework to distill the knowledge from the source domains to a customized target model. Then we design a new temporal consistency loss by combining an adaptive mask method and dynamic threshold method to maintain consistent temporal information and balance the learning difficulties of different classes. For the multisource black-box domain adaption, we further propose a Shapley-enhanced method to determine the contribution of each source domain. Experimental results on both single-source and multisource domain adaption show that our framework has superior performance compared to other black-box UDA methods.

Learning Multi-Agent Cooperation via Considering Actions of Teammates.

Recently value-based centralized training with decentralized execution (CTDE) multi-agent reinforcement learning (MARL) methods have achieved excellent performance in cooperative tasks. However, the most representative method among these methods, Q-network MIXing (QMIX), restricts the joint action Q values to be a monotonic mixing of each agent's utilities. Furthermore, current methods cannot generalize to unseen environments or different agent configurations, which is known as ad hoc team play situation. In this work, we propose a novel Q values decomposition that considers both the return of an agent acting on its own and cooperating with other observable agents to address the nonmonotonic problem. Based on the decomposition, we propose a greedy action searching method that can improve exploration and is not affected by changes in observable agents or changes in the order of agents' actions. In this way, our method can adapt to ad hoc team play situation. Furthermore, we utilize an auxiliary loss related to environmental cognition consistency and a modified prioritized experience replay (PER) buffer to assist training. Our extensive experimental results show that our method achieves significant performance improvements in both challenging monotonic and nonmonotonic domains, and can handle the ad hoc team play situation perfectly.

Cross-Image Pixel Contrasting for Semantic Segmentation.

This work studies the problem of image semantic segmentation. Current approaches focus mainly on mining "local" context, i.e., dependencies between pixels within individual images, by specifically-designed, context aggregation modules (e.g., dilated convolution, neural attention) or structure-aware optimization objectives (e.g., IoU-like loss). However, they ignore "global" context of the training data, i.e., rich semantic relations between pixels across different images. Inspired by recent advance in unsupervised contrastive representation learning, we propose a pixel-wise contrastive algorithm, dubbed as PiCo, for semantic segmentation in the fully supervised learning setting. The core idea is to enforce pixel embeddings belonging to a same semantic class to be more similar than embeddings from different classes. It raises a pixel-wise metric learning paradigm for semantic segmentation, by explicitly exploring the structures of labeled pixels, which were rarely studied before. Our training algorithm is compatible with modern segmentation solutions without extra overhead during testing. We experimentally show that, with famous segmentation models (i.e., DeepLabV3, HRNet, OCRNet, SegFormer, Segmenter, MaskFormer) and backbones (i.e., MobileNet, ResNet, HRNet, MiT, ViT), our algorithm brings consistent performance improvements across diverse datasets (i.e., Cityscapes, ADE20 K, PASCAL-Context, COCO-Stuff, CamVid). We expect that this work will encourage our community to rethink the current de facto training paradigm in semantic segmentation.

Risk of Acute Kidney Injury Following IV Iodinated Contrast Media Exposure: 2023 Update, From the AJR Special Series on Contrast Media.

Iodinated contrast material (ICM) has revolutionized the field of diagnostic radiology through improvements in diagnostic performance and the expansion of clinical indications for radiographic and CT examinations. Historically, nephrotoxicity was a feared complication of ICM use, thought to be associated with a significant risk of morbidity and mortality. Such fears often precluded the use of ICM in imaging evaluations, commonly at the expense of diagnostic performance and timely diagnosis. Over the past 20 years, the nephrotoxic risk of ICM has become a topic of debate, as more recent evidence from higher-quality studies now suggests that many cases of what was considered contrast-induced acute kidney injury (CI-AKI) likely were cases of mistaken causal attribution; most of these cases represented either acute kidney injury (AKI) caused by any of myriad other known factors that can adversely affect renal function and were coincidentally present at the time of contrast media exposure (termed "contrast-associated AKI" [CA-AKI]) or a manifestation of the normal variation in renal function that increases with worsening renal function. This Special Series Review discusses the current state of knowledge regarding CI-AKI and CA-AKI, including the incidence, risk factors, outcomes, and prophylactic strategies in the identification and management of these clinical conditions.

Hospital Allocation and Racial Disparities in Health Care

Abstract We develop a framework to measure the role of hospital allocation in racial disparities in health care and use it to study Black and white heart attack patients. Black patients receive care at lower-performing hospitals than white patients. However, over two decades, the performance gap between hospitals treating Black and white patients shrank by over two-thirds. This progress is due to more rapid performance improvement at hospitals that tended to treat Black patients rather than reallocation of patients. Hospital improvement is correlated with adoption of a productivity-raising input, beta blockers. Our work highlights reallocation and performance improvement as future disparity-reduction levers.

Interaction-and-Response Network for Distantly Supervised Relation Extraction.

Distantly supervised relation extraction (DSRE) aims to identify semantic relations from massive plain texts. A broad range of the prior research has leveraged a series of selective attention mechanisms over sentences in a bag to extract relation features without considering dependencies among the relation features. As a result, potential discriminative information existed in the dependencies is ignored, causing a decline in the performance of extracting entity relations. In this article, we focus on going beyond the selective attention mechanisms and propose a new framework termed interaction-and-response network (IR-Net) that adaptively recalibrates the features of sentence, bag, and group levels by explicitly modeling interdependencies among the features on each level. The IR-Net consists of a series of interactive and responsive modules throughout feature hierarchy, seeking to strengthen its power of learning salient discriminative features for distinguishing entity relations. We conduct extensive experiments on three benchmark DSRE datasets, including NYT-10, NYT-16, and Wiki-20m. The experimental results demonstrate that the IR-Net brings obvious improvements in performance when comparing ten state-of-the-art DSRE methods for entity relation extraction.

Bidirectional Probabilistic Subspaces Approximation for Multiview Clustering.

The existing multiview clustering models learn a consistent low-dimensional embedding either from multiple feature matrices or multiple similarity matrices, which ignores the interaction between the two procedures and limits the improvement of clustering performance on multiview data. To address this issue, a bidirectional probabilistic subspaces approximation (BPSA) model is developed in this article to learn a consistently orthogonal embedding from multiple feature matrices and multiple similarity matrices simultaneously via the disturbed probabilistic subspace modeling and approximation. A skillful bidirectional fusion strategy is designed to guarantee the parameter-free property of the BPSA model. Two adaptively weighted learning mechanisms are introduced to ensure the inconsistencies among multiple views and the inconsistencies between bidirectional learning processes. To solve the optimization problem involved in the BPSA model, an iterative solver is derived, and a rigorous convergence guarantee is provided. Extensive experimental results on both toy and real-world datasets demonstrate that our BPSA model achieves state-of-the-art performance even if it is parameter-free.

LGGNet: Learning From Local-Global-Graph Representations for Brain-Computer Interface.

Neuropsychological studies suggest that co-operative activities among different brain functional areas drive high-level cognitive processes. To learn the brain activities within and among different functional areas of the brain, we propose local-global-graph network (LGGNet), a novel neurologically inspired graph neural network (GNN), to learn local-global-graph (LGG) representations of electroencephalography (EEG) for brain-computer interface (BCI). The input layer of LGGNet comprises a series of temporal convolutions with multiscale 1-D convolutional kernels and kernel-level attentive fusion. It captures temporal dynamics of EEG which then serves as input to the proposed local- and global-graph-filtering layers. Using a defined neurophysiologically meaningful set of local and global graphs, LGGNet models the complex relations within and among functional areas of the brain. Under the robust nested cross-validation settings, the proposed method is evaluated on three publicly available datasets for four types of cognitive classification tasks, namely the attention, fatigue, emotion, and preference classification tasks. LGGNet is compared with state-of-the-art (SOTA) methods, such as DeepConvNet, EEGNet, R2G-STNN, TSception, regularized graph neural network (RGNN), attention-based multiscale convolutional neural network-dynamical graph convolutional network (AMCNN-DGCN), hierarchical recurrent neural network (HRNN), and GraphNet. The results show that LGGNet outperforms these methods, and the improvements are statistically significant ( ) in most cases. The results show that bringing neuroscience prior knowledge into neural network design yields an improvement of classification performance. The source code can be found at https://github.com/yi-ding-cs/LGG.

Crown Ether Electrolyte Additive Enables High-Rate and Stable Polyviologen Cathode Material for Chloride Ion Batteries.

A variety of inorganic and inorganic cathode materials for chloride ion storage are reported. However, their application in chloride ion batteries (CIB) is hindered by poor rate capability and cycling stability. Herein, an organic poly(butyl viologen dichloride) (PBVCl2) cathode material with significantly enhanced rate and cycling performance in the CIB is achieved using a crown ether (18-Crown-6) additive in the tributylmethylammonium chloride-based electrolyte. The as-prepared PBVCl2 cathodes exhibit impressive capacity increases from 149.4 to 179.1 mAh g-1 at 0.1 C and from 57.8 to 111.9 mAh g-1 at 10 C, demonstrating the best rate performance with the highest energy density among those of various reported cathodes for CIBs. This impressive performance improvement is a result of the great boosts in charge transfer, ion transport, and interface stability of the battery by the use of 18-Crown-6, which also contributes to a more than twofold increase in capacity retention after 120 cycles. The electrode reaction mechanism of the CIB based on highly reversible chloride ion transfer is revealed by Fourier transform infrared spectroscopy and X-ray photoelectron spectroscopy.

Balanced Federated Semisupervised Learning With Fairness-Aware Pseudo-Labeling.

Federated semisupervised learning (FSSL) aims to train models with both labeled and unlabeled data in the federated settings, enabling performance improvement and easier deployment in realistic scenarios. However, the nonindependently identical distributed data in clients leads to imbalanced model training due to the unfair learning effects on different classes. As a result, the federated model exhibits inconsistent performance on not only different classes, but also different clients. This article presents a balanced FSSL method with the fairness-aware pseudo-labeling (FAPL) strategy to tackle the fairness issue. Specifically, this strategy globally balances the total number of unlabeled data samples which is capable to participate in model training. Then, the global numerical restrictions are further decomposed into personalized local restrictions for each client to assist the local pseudo-labeling. Consequently, this method derives a more fair federated model for all clients and gains better performance. Experiments on image classification datasets demonstrate the superiority of the proposed method over the state-of-the-art FSSL methods.

In-Page Navigation Aids for Screen-Reader Users with Automatic Topicalisation and Labelling

Navigation aids such as headers and internal links provide vital support for screen-reader users on web documents to grasp a document’s structure. However, when such navigation aids are unavailable or not appropriately marked up, this situation can cause serious difficulties. This article presents the design and evaluation of a tool for automatically generating navigation aids with headers and internal links for screen readers with topicalisation and labelling algorithms. The proposed tool uses natural language processing techniques to divide a web document into topic segments and label each segment in two cycles based on its content. We conducted an initial user study in the first cycle with eight blind and partially-sighted screen reader users. The evaluation involved tasks with questions answered by participants with information from texts with and without automatically generated headers. The results in the first cycle provided preliminary indicators of performance improvement and cognitive load reduction. The second cycle involved co-designing an improved version with two blind experts in web accessibility, resulting in a browser extension which injects automatically generated headers and in-page navigation with internal links, along with improvements in the generation of labels using OpenAI’s ChatGPT. The browser extension was evaluated by seven blind participants using the same four texts used to evaluate the preliminary prototype developed in the first cycle. With the two development cycles, the study provided important insights into the design of navigation aids for screen-reader users using natural language processing techniques, including the potential use of generative artificial intelligence for assistive technologies and limitations that need to be explored in future research.

Analysis of the key factors for small and medium-sized enterprises growth using principal component analysis

Small and Medium-sized Enterprises (SMEs) significantly contribute to national economic development worldwide. However, numerous factors affect the growth of SMEs, particularly in developing economies. Due to the complexity of these variables, it is challenging to determine where to begin improving SMEs. This study aims to identify and analyse the factors that hinder the overall performance of (SMEs) to gain insights into the principal variables constituting problems. A literature survey identified and abstracted 36 variables that influence SME growth. A structured questionnaire was created using the variables and administered to the respondents in the SME sector. We collated and converted the respondents' scores into primary data. The data was analysed using Kendall's Coefficient of Concordance (W) and Principal Component Analytics (PCA) tools, and the results were presented. The result of Kendall's Coefficient of Concordance analysis was (W) = 0.52, indicating that the judges strongly agree that the 36 variables affect SME growth. The PCA analysis identified the critical variables affecting SME growth. These variables, along with their corresponding factor loading, include Marketing information = 0.80, Cost of transportation = 0.81, Information technology = 0.83, Economic initiatives = −0.83, Financial constraints = −0.80, Cultural change = 0.80, Technical know-how = 0.822, Economic factors = −0.80, and Business information system = 0.813. The implication of the study's findings for management is that SMEs should begin process improvement for effective overall performance by addressing the problems associated with the critical variables. The perceived limitation of the study is that the respondents' opinions may not reflect 100 % of the opinions of the unsampled population in the SME sector. This study's originality includes (i) holistic documentation of the myriad of variables influencing SMEs, which are fragmented in the literature, and (ii) pinpointing the critical variables affecting SME development with measurable evidence not found in the literature.

Hydration, phase assemblage and microstructure of cementitious blends with low-grade limestone

Limestone of the purity mandated by the standards for use in cement manufacture (as raw material or as performance improver) is only a fraction of the mined material. Sometimes, limestone contains impurities like silica, clay, dolomite etc. beyond the levels specified for use as an additive or for cement production. This limestone, commonly known as low-grade limestone, also may have a potential to be used as a supplementary cementitious material. This paper reports a study conducted to understand the hydration behaviour of ordinary Portland cement in combination with low-grade limestone. Low-grade limestone samples with dolomite, silica and clay impurities were collected from various limestone mines in India and used in cementitious blends at various replacement levels after suitable processing. Reaction kinetics was studied using calorimetry and phase assemblage using techniques like X-ray diffraction and thermogravimetry. Mercury intrusion porosimetry was employed to study the microstructure of cementitious blends. Dilution effect of low-grade limestone was studied by measuring compressive strength of cement mortar cubes for various replacement levels. Shortening of induction period, increased peak heat rate and reduced time to peak were observed in the blends of ordinary Portland cement with low-grade limestone. Carboaluminate formation, ettringite stabilisation and higher bound water content were also seen. The studies showed that low-grade limestone also can be used as a partial replacement for ordinary Portland cement without compromising the strength up to 15% replacement. Higher replacement levels, although showing a reduction in strength due to dilution effect, could still be considered for non-structural applications.

Gender-related differences in kinematic parameters of the start and performance in a 50m breaststroke swimming

Introduction: In swimming, the competitions are divided by age categories, with each swimming event beginning at the start. For optimal improvements in swimming performance in the sprint events, the performance at the start and during swimming in each age category should be analyzed. Aim: The aim of our study is to show the differences in kinematic parameters of the start and performance in the sprint event of 50 m breaststroke from the perspective of gender in different age groups at international competitions organized in Slovakia. Materials and Methods: The females were categorized by age into K3 (2010-2011; n=36; 282 WA points), K2 (2008-2009; n=26; 409 WA points) and K1 (2007 and older; n=34; 503 WA points) groups. Males were categorized into K3 (2009-2010; n=27; 265 WA points), K2 (2007-2008; n=33; 372 WA points) and K1 (2006 and older; n=39; 207 WA points) groups. We used SwimPro cameras and Dartgish software to monitor kinematic parameters in the above-water, underwater, and swim phases. The parameters monitored were block time (BT), time (FT), and length (FD) of flight, time (UWT) and length (UWD) underwater, swim time (ST) and distance (SD) to 15 m, time to 15 m (T15), 25 m (T25), and 50 m (T50) distance. Data were tested by Shapiro-Wilk, Kruskal-Wallis ANOVA, and Mann-Whitney U test in Statistica software (version 13.5). Results: Statistically significant differences (p<0.01) in females were found in the above-water phase in BT, FT, and FD between K3 and K1. The underwater phase in UWT showed differences between K3 to K1, K2 to K1 whereas for UWD - between K3 and K2 and between K2 and K1. For T15 and T25, there were differences (p<0.01) between all categories, but in the final time (T50), the differences occurred between K3 and K2 and between K3 and K1. In males, there was a significant difference in the above-water phase between K3 and K1 and between K2 and K1. In UWD, ST, SD, T15, T25, and T50, there were significant differences between all categories. Differences (p>0.01) between males and females by category in the above-water phase were measured in FT (K1) and FD (K2, K1). In the underwater phase, the differences were found in UWT (K2) and UWD (K1). For T15, T25, and T50, significant differences were shown in all categories. Conclusions: The kinematic analysis revealed differences not only in the start phases and the 50 m breaststroke swimming in the male and female categories but also between genders. Differences in swimming performance may be due to performance at the start and therefore more attention should be paid to the start phases in the different categories of swimmers.

A mixture of free and microencapsulated essential oils combined with turmeric and tannin in the diet of cattle in the growing and finishing phase: A new tool to enhance productivity

The study was conducted to evaluate whether adding a blend containing free and microencapsulated essential oils, combined with turmeric and tannin, can replace monensin as a performance improver, positively influencing animal health and meat quality. Fourteen male cattle with an average initial body weight of 253 kg and ten females with an average initial body weight of 239 kg were included in the experiment in a confinement system. The animals were divided into four independent groups: Males (M-PHYTO, inclusion of the phytogenic mixture; M-MONEN, inclusion of monensin); Females (F-PHYTO, inclusion of phytogenic; F-MONEN, inclusion of monensin). The diets were formulated for an average daily gain of 1.5 kg, with food provided twice daily. Data and biological samples were collected on days 33, 100, and 153. M-PHYTO and F-PHYTO showed greater average daily gain and body weight, with a tendency to consume more dry matter and feed efficiency. However, the apparent digestibility of nutrients was higher in animals in the MONEN group than those in the PHYTO group, regardless of sex. The total number of leukocytes in cattle in the M-PHYTO and F-PHYTO groups was higher due to the higher lymphocyte count in the blood of these animals. The activity of gamma-glutamyltransferase and aspartate aminotransferase in animals that consumed the phytogenic mixture was lower compared to animals that consumed monensin. TBARS levels were lower in the serum and meat of cattle in the M-PHYTO and F-PHYTO groups, as were the total thiol concentration and glutathione S-transferase activity in the blood and meat of these animals. Bacterial activity was greater in the rumen fluid of cattle that consumed the phytogenic, as demonstrated by the methylene blue reduction test. The total production of short-chain fatty acids and acetic acid showed an interaction between treatment and days for both sex classes and an effect of treatment for females, with a higher concentration in the rumen in the M-PHYTO and F-PHYTO groups. Propionic acid affected treatment and the treatment x day interaction for males and females, with the M-PHYTO and F-PHYTO groups showing higher concentrations. There was an interaction between treatment and day in the acetate/propionate ratio, being lower in M-PHYTO and F-PHYTO cattle. Treatment between ruminal microbiota groups had no effect. The rib eye area had more significant gains when consumed phytogenic by males and females. The subcutaneous and sirloin fat thickness was greater in F-PHYTO animals than in the control group; however, in males, the result was the opposite. The sum of polyunsaturated fatty acids in meat was higher in M-PHYTO compared to M-MONEN and lower in F-PHYTO meat compared to F-MONEN. These results indicate that the phytogenic mixture is a potential performance improver in the diet of growing and finishing cattle, regardless of sex, but meat characteristics differ between males and females.

Cooperative Control and Performance Evaluation of a Linear MIMO Parabolic Spatiotemporal Dynamic System on Directed and Switching Topologies.

The current work concerns on cooperative control of exponential stabilization and control performance improvement in the spatial domain for a linear spatiotemporal dynamic system associated with multiple control actuators and multiple collaborative measurement sensors. By assuming that the system dynamics is modeled by a MIMO parabolic partial differential equation (PPDE) and each sensor can share measurement information with its topological neighbors in directed and switching topological networks, a cooperative control protocol is proposed to achieve the control aim of this article. With the help of a combination of multiagent consensus theory, Lyapunov's method, and integral inequality technique, sufficient conditions are presented for the closed-loop exponential stability of the PPDE in the norm |·|2 . Moreover, some performance indexes are defined to evaluate the closed-loop control performance improvement in the spatial domain. Extensive simulation results are finally presented for a simple numerical example and a practical heat treatment process to verify the effectiveness and performance improvement capability of the proposed cooperative control protocol.

Durability of Effects of Cognitive Remediation on Cognition and Psychosocial Functioning in Schizophrenia: A Systematic Review and Meta-Analysis of Randomized Clinical Trials.

Cognitive remediation provides substantial improvements in cognitive performance and real-world functioning for people living with schizophrenia, but the durability of these benefits needs to be reassessed and better defined. The aims of this study were to provide a comprehensive assessment of the durability of the benefits of cognitive remediation for cognition and functioning in people living with schizophrenia and evaluating potential moderators of effects. A systematic search was conducted in PubMed, Scopus, and PsycINFO, and reference lists of included articles and Google Scholar were inspected. Eligible studies were randomized clinical trials of cognitive remediation in patients diagnosed with schizophrenia spectrum disorders in which follow-up assessments were included. Screening and data extraction were performed by at least two independent reviewers. Cohen's d was used to measure outcomes. Primary outcomes were changes in cognition and functioning from baseline to conclusion of follow-up. Moderators of the durability of effects were assessed. Of 2,840 identified reports, 281 full texts were assessed and 130 reports on 67 studies with 5,334 participants were included. Cognitive remediation produced statistically significant positive effects that persisted at the end of follow-up in global cognition (d=0.23) and in global functioning (d=0.26). Smaller study samples and single-center studies were associated with better cognitive outcomes; longer treatment and follow-up duration, techniques for transferring cognitive gains to the real world, integration with psychiatric rehabilitation, group format of delivery, and more female participants in the sample were associated with better functional outcomes. Cognitive remediation provides durable improvements in cognition and functioning in schizophrenia. This finding corroborates the notion that cognitive remediation should be implemented more widely in clinical and rehabilitation practice.

Fast Incomplete Multi-View Clustering With View-Independent Anchors.

Multi-view clustering (MVC) methods aim to exploit consistent and complementary information among each view and achieve encouraging performance improvement than single-view counterparts. In practical applications, it is common to obtain instances with partially available information, raising researches of incomplete multi-view clustering (IMC) issues. Recently, several fast IMC methods have been proposed to process the large-scale partial data. Though with considerable acceleration, these methods seek view-shared anchors and ignore specific information among single views. To tackle the above issue, we propose a fast IMC with view-independent anchors (FIMVC-VIA) method in this article. Specifically, we learn individual anchors based on the diversity of distribution among each incomplete view and construct a unified anchor graph following the principle of consistent clustering structure. By constructing an anchor graph instead of pairwise full graph, the time and space complexities of our proposed FIMVC-VIA are proven to be linearly related to the number of samples, which can efficiently solve the large-scale task. The experiment performed on benchmarks with different missing rate illustrates the improvement in complexity and effectiveness of our method compared with other IMC methods. Our code is publicly available at https://github.com/Tracesource/ FIMVC-VIA.

A Survey of Visual Transformers.

Transformer, an attention-based encoder-decoder model, has already revolutionized the field of natural language processing (NLP). Inspired by such significant achievements, some pioneering works have recently been done on employing Transformer-liked architectures in the computer vision (CV) field, which have demonstrated their effectiveness on three fundamental CV tasks (classification, detection, and segmentation) as well as multiple sensory data stream (images, point clouds, and vision-language data). Because of their competitive modeling capabilities, the visual Transformers have achieved impressive performance improvements over multiple benchmarks as compared with modern convolution neural networks (CNNs). In this survey, we have reviewed over 100 of different visual Transformers comprehensively according to three fundamental CV tasks and different data stream types, where taxonomy is proposed to organize the representative methods according to their motivations, structures, and application scenarios. Because of their differences on training settings and dedicated vision tasks, we have also evaluated and compared all these existing visual Transformers under different configurations. Furthermore, we have revealed a series of essential but unexploited aspects that may empower such visual Transformers to stand out from numerous architectures, e.g., slack high-level semantic embeddings to bridge the gap between the visual Transformers and the sequential ones. Finally, two promising research directions are suggested for future investment. We will continue to update the latest articles and their released source codes at https://github.com/liuyang-ict/awesome-visual-transformers.

Pure large kernel convolutional neural network transformer for medical image registration

Deformable medical image registration is a fundamental and critical task in medical image analysis. Recently, deep learning-based methods have rapidly developed and have shown impressive results in deformable image registration. However, existing approaches still suffer from limitations in registration accuracy or generalization performance. To address these challenges, in this paper, we propose a pure convolutional neural network module (CVTF) to implement hierarchical transformers and enhance the registration performance of medical images. CVTF has a larger convolutional kernel, providing a larger global effective receptive field, which can improve the network’s ability to capture long-range dependencies. In addition, we introduce the spatial interaction attention (SIA) module to compute the interrelationship between the target feature pixel points and all other points in the feature map. This helps to improve the semantic understanding of the model by emphasizing important features and suppressing irrelevant ones. Based on the proposed CVTF and SIA, we construct a novel registration framework named PCTNet. We applied PCTNet to generate displacement fields and register medical images, and we conducted extensive experiments and validation on two public datasets, OASIS and LPBA40. The experimental results demonstrate the effectiveness and generality of our method, showing significant improvements in registration accuracy and generalization performance compared to existing methods. Our code has been available at https://github.com/fz852/PCTNet.