Browse Articles

LKDA-Net: Hierarchical transformer with large Kernel depthwise convolution attention for 3D medical image segmentation

Since Transformers have demonstrated excellent performance in the segmentation of two-dimensional medical images, recent works have also introduced them into 3D medical segmentation tasks. For example, hierarchical transformers like Swin UNETR have reintroduced several prior knowledge of convolutional networks, further enhancing the model’s volumetric segmentation ability on three-dimensional medical datasets. The effectiveness of these hybrid architecture methods is largely attributed to the large number of parameters and the large receptive fields of non-local self-attention. We believe that large-kernel volumetric depthwise convolutions can obtain large receptive fields with fewer parameters. In this paper, we propose a lightweight three-dimensional convolutional network, LKDA-Net, for efficient and accurate three-dimensional volumetric segmentation. This network adopts a large-kernel depthwise convolution attention mechanism to simulate the self-attention mechanism of Transformers. Firstly, inspired by the Swin Transformer module, we investigate different-sized large-kernel convolution attention mechanisms to obtain larger global receptive fields, and replace the MLP in the Swin Transformer with the Inverted Bottleneck with Depthwise Convolutional Augmentation to reduce channel redundancy and enhance feature expression and segmentation performance. Secondly, we propose a skip connection fusion module to achieve smooth feature fusion, enabling the decoder to effectively utilize the features of the encoder. Finally, through experimental evaluations on three public datasets, namely Synapse, BTCV and ACDC, LKDA-Net outperforms existing models of various architectures in segmentation performance and has fewer parameters. Code: https://github.com/zouyunkai/LKDA-Net.

Can creativity in science be learnt? These researchers think so

Research and application of an evaluation mechanism for enterprise information systems

Enhanced “Green Nudging”: Tapping the channels of cultural transmission

This paper shows how to increase the effectiveness of “green nudging” as a policy measure to induce sustainable preferences. Evidence indicates that the behavioral impact of “green nudges” is subject to decay. To address this problem, we propose “enhanced green nudges”, which incorporate learning biases as features of humans’ capacity for culture. These provide information for the formation of enduring “green” preferences. Several biases in cultural transmission are considered, such as direct bias, norm-following behavior, conformity, self-similarity, and the influence of role models. Moreover, a prerequisite for “enhanced green nudges” to be effective is that learning environments resemble the ones biases evolved from in humans’ evolutionary past. Based on a model of cultural evolution, several scenarios of preference acquisition through “green nudges” illustrate our arguments and indicate implications for policy.

Daily briefing: US researchers fight back on key climate report

Autumn leaf color brightness of Japanese alpine vegetation is projected to decrease under future climate change

Abstract Autumn leaf coloring is an essential cultural ecosystem service, but mechanisms of color brightness have scarcely been focused on, even though the autumn color crisis (leaf fall without coloring) has been reported in Japanese alpine shrubs. To approach the mechanisms, we analyzed the relationship between brightness and environmental conditions and projected possible future changes. Autumn color brightness was measured by repeated camera observations at three Japanese alpine sites. Environmental factors were analyzed using linear mixed model comparisons. Results showed earlier green-up timing was related to a duller autumn leaf color, possibly because a prolonged leaf period creates older leaves and reduces physiological vitality in autumn, affecting anthocyanin biosynthesis. Green-up timing based on snowmelt day and degree-day-based prediction of snowmelt day predicted earlier snowmelt and green-up in the future, with decreases in autumn color brightness. The amount of color reduction varied among climate models and scenarios, from a decline of several percent under MRI-CGCM3 RCP 2.6 to around 15% under MIROC5 RCP8.5 by the end of this century. Projected reductions in autumn color brightness could have an economic impact on tourism, and it could also be linked to modifications in material cycles and the ecosystem’s productivity.

An objective assessment of vision performance at different reading distances after smooth incision lenticular keratomileusis (SILK) for myopia

A condition diagnosis method for subway track structures employing distributed optical fiber sensing

An overview of coarse particle beneficiation of lithium ores

Abstract The increasing demand for lithium-ion batteries particularly for electric vehicles underscores the importance of improving the sustainability of lithium mining operations. The depletion of high-grade lithium ore deposits has necessitated the upgrading of medium to low-grade ores for lithium extraction. Spodumene is the most commercially exploited lithium-bearing mineral found in pegmatites due to its high lithium content. Ore sorting can be used for early rejection of up to 60% of gangue minerals prior to preconcentration. Dense media separation is a viable spodumene beneficiation method. However, as case studies have shown, flotation may still be required to process middlings and the undersized fraction, which falls outside the particle size range effective for dense media separation. Magnetic separation can be conducted during or after flotation to remove iron impurities in lithium concentrates. While fine particle flotation has historically achieved high recovery rates, their economic feasibility is increasingly questioned due to intensive comminution requirements. Coarse particle flotation in mechanical flotation cells for instance is inefficient due to turbulence-induced detachment of coarse particles. Coarse particle beneficiation using fluidized bed flotation cells can offer advantages such as reduced grind size and environmental footprint. Despite proven energy savings and recovery efficiencies in other mineral sectors, their application in lithium mining operations remains limited to pilot scale. Also, research in this area is underexplored. This review addresses this gap by evaluating the feasibility, potential benefits and challenges of integrating ore sorting, dense media separation, magnetic separation and fluidized bed flotation with the HydroFloat, NovaCell and Reflux cells into lithium ore beneficiation flowsheets. Key challenges identified include high water consumption and the inadvertent entrainment of fine particles requiring desliming steps. Furthermore, this review acknowledges the challenges in spodumene beneficiation due to the structural similarities among silicate minerals and highlights relevant pretreatment methods to improve selectivity, recovery and grade.

Rapid synthesis of graphitic carbon nitride nanosheets as an efficient adsorbent for removal of Methylene Blue and Rhodamine B from Aqueous Solutions

Study on acoustic emission evolution features of gas-bearing coal under different gas pressures

BCDCNN: breast cancer deep convolutional neural network for breast cancer detection using MRI images

Abstract Breast cancer (BC) is a kind of cancer that is created from the cells in breast tissue. This is a primary cancer that occurs in women. Earlier identification of BC is significant in the treatment process. To lessen unwanted biopsies, Magnetic Resonance Imaging (MRI) is utilized for diagnosing BC nowadays. MRI is the most recommended examination to detect and monitor BC and explain lesion areas as it has a better ability for soft tissue imaging. Even though, it is a time-consuming procedure and requires skilled radiologists. Here, Breast Cancer Deep Convolutional Neural Network (BCDCNN) is presented for Breast Cancer Detection (BCD) using MRI images. At first, the input image is taken from the database and subjected to a pre-processing segment. Adaptive Kalman filter (AKF) is utilized to execute the pre-processing phase. Thereafter, cancer area segmentation is conducted on filtered images by Pyramid Scene Parsing Network (PSPNet). To improve segmentation accuracy and adapt to complex tumor boundaries, PSPNet is optimized using the Jellyfish Search Optimizer (JSO). It is a recent nature-inspired metaheuristic that converges to an optimal solution in fewer iterations compared to conventional methods. Then, image augmentation is performed that includes augmentation techniques namely rotation, random erasing and slipping. Afterwards, feature extraction is done and finally, BCD is conducted employing BCDCNN, wherein the loss function is newly designed based on an adaptive error similarity. It improves the overall performance by dynamically emphasizing samples with ambiguous predictions, enabling the model to focus more on diagnostically challenging cases and enhancing its discriminative capability. Furthermore, BCDCNN acquired 90.2% of accuracy, 90.6% of sensitivity and 90.9% of specificity. The proposed method not only demonstrates strong classification performance but also holds promising potential for real-world clinical application in early and accurate breast cancer diagnosis.

Discovering activity transition patterns in social media check-in behavior via temporal activity motifs

Abstract Location-Based Social Network (LBSN) has produced a large quantity of user check-in data. A profound understanding of user behavior and intrinsic needs can be achieved by identifying patterns in activity type transitions, thereby enabling more intelligent location-based services. We proposed temporal activity motif and used this structure to identify frequent activity type transition patterns from check-in sequences, discovering the relationship and interaction between different activity types. 383 temporal activity motifs of 17 temporal topologies were extracted from the two-year Gowalla dataset of New York-Newark-Jersey City, NY-NJ-PA Metropolitan Statistical Area (MSA). These motifs are categorized into two groups: one is sequential motifs representing a complete activity process, while the other is non-sequential motifs representing the co-occurrence of two separate processes. They provide evidence of activity type recurrence, particularly in longer activity processes, highlights the cyclical nature of human mobility. Additionally, various activity types exhibit different influences on others and occupy different positions in activity processes. Furthermore, by leveraging non-sequential motifs, we specifically uncovered the co-occurrence patterns between two separate activity process. These findings bring new insights to optimize recommendation system and urban planning.

A data driven framework for optimizing droplet microfluidics with residual block and Fourier enhanced networks

Uncertainty management in multiobjective electric vehicle integrated optimal power flow based hydrothermal scheduling of renewable power system for environmental sustainability

Spatiotemporal analysis and forecasting of public attention to China’s five major religions

Abstract In the digital era, leveraging search engine data to gain insights into public interests and attitudes towards religion has become increasingly important. To examine the spatiotemporal evolution and develop a forecasting model of public attention to China’s five major religions (Buddhism, Taoism, Catholicism, Christianity and Islam), this study introduces geographic information system technology, forecasting technology and spatiotemporal analysis methods into religious research, using Baidu Index data from 2020 to 2024. The results show that most religions exhibit stable annual public attention, although significant changes occur during specific seasons, festivals or events. Spatial variation in public attention is moderate, with a relatively balanced regional distribution. However, a distinct east–west clustering pattern is evident, reflecting spatial aggregation. The SARIMA-based forecasting model effectively captures temporal dynamics and demonstrates strong forecasting performance. Findings reveal the spatiotemporal distribution patterns of religious public attention and highlight the practical value of the forecasting model, thereby emphasising the importance of considering spatial factors and forecasting models when addressing the dissemination of religion in the digital age. This study provides new methods and perspectives for religious research, promoting an interdisciplinary synthesis of religious studies, sociology and geography. It offers new insights into global religious studies, religious communication strategies and cultural exchange. Furthermore, it contributes to advancing cross-cultural and cross-regional religious research and offers valuable references for religious organisations and policymakers in managing religious affairs, adjusting policies and optimising communication strategies in a globalised context.

ERNIE-TextCNN: research on classification methods of Chinese news headlines in different situations

Deep learning approach for automated hMPV classification

Nanoparticle conjugation of ginsenoside Rh2 enhanced antitumor efficacy on hepatocellular carcinoma

Integrating data-driven and physics-based approaches for robust wind power prediction: A comprehensive ML-PINN-Simulink framework

Abstract This study presents a comprehensive hybrid forecasting framework that synergizes machine learning algorithms, MATLAB Simulink-based physical modeling, and Physics-Informed Neural Networks (PINNs) to advance wind power prediction accuracy for a 10 kW Permanent Magnet Synchronous Generator (PMSG)-based Wind Energy Conversion System (WECS). Using a complete annual dataset of 8,760 hourly wind speed observations from the MERRA-2 platform, ten machine learning algorithms were systematically evaluated, including Random Forest, XGBoost, and an advanced Stacking ensemble model. The Stacking ensemble demonstrated superior performance, achieving an exceptional R2 of 0.998 and RMSE of 0.11, significantly outperforming individual algorithms. A detailed MATLAB Simulink model was developed to replicate turbine behaviour under identical wind conditions, physically, providing robust validation for ML predictions. The Simulink model achieved satisfactory performance under nominal wind conditions but exhibited computational constraints during extreme wind scenarios, leading to compromised output reliability. To bridge the gap between pure data-driven learning and physical realism, a Physics-Informed Neural Network was subsequently integrated to combine data-driven learning with physical constraints, using both observational data and physics-based synthetic datasets. Comparative analysis revealed that ML models deliver superior speed and accuracy for operational forecasting, while the PINN framework maintains physical consistency with competitive predictive performance. The framework’s practical applicability was demonstrated through a 2026 case study for southern Tamil Nadu, which incorporated projected environmental changes, including a 0.6% annual decline in wind speed. This real-world validation showcased the framework’s adaptability to evolving climatic conditions and long-term forecasting capabilities. This integrated methodology provides a robust foundation for enhancing wind power integration into modern energy systems, while maintaining both computational accuracy and physical interpretability, thereby supporting sustainable energy transition goals.