The segmentation of slender structures in images faces challenges of discontinuous segmentation and insufficient recognition. These slender structures, such as arteries and veins, are particularly important in 3D medical images that are sensitive to the computational complexity of segmentation networks. Therefore, in order to balance the computational complexity of the network and adaptive perception of slender structures, this paper proposes a multi-directional attention module for slender structures, which is a lightweight attention module that can be inserted into the encoder or decoder unit. At the same time, we propose a contour loss function to address the class imbalance phenomenon that may arise in the joint segmentation task of slender and ordinary structures. This function improves the balance by converting traditional class mask labels into contour mask labels. The effectiveness of our proposed module has been validated through training on segmentation tasks on 2D and 3D images.

To provide reliable and high-quality services in the sixth-generation (6G) systems, movable antennas (MAs) have attracted much attention since they can use the spatial degree of freedom adequately. Compared to the traditional fixed position arrays, MAs give much better performance in multi-user and multi-antenna scenarios, which implement efficient beamforming and interference suppression in various communication cases. However, the MA array design strategy and the associated channel estimation problems require high-complexity iterative computation algorithms, making it difficult to be exploited in practical applications. In this work, a novel channel estimation method with the MA arrays is proposed based on the convolutional neural network (CNN), which considers the complexity of the algorithm and time consumption while accomplishing the optimal channel estimation. By comparing it with different benchmarks, especially for the orthogonal matching tracking, the CNN-based channel estimation method implements a better trade-off between the mean square error and the computational complexity and the designed examples are provided to verify the effectiveness of the proposed approaches.

Embodied artificial intelligence has emerged as a transformative paradigm, marking a fundamental shift in artificial intelligence research toward systems that tightly couple perception, cognition, and action within real-world environments. This editorial emphasizes the growing significance of embodied artificial intelligence, introduces the key contributions presented in this Special Issue, and provides an overview of the current challenges and prospective research directions shaping the future of the field.

Micro-electromechanical system (MEMS) microrobots have significant potential in various fields, including healthcare, assembly, and industrial monitoring. This paper provides a comprehensive review of microrobot components, including materials, sensing, actuation, and power systems. It also discusses microrobot classifications - such as swimming, walking, aerial types, microgrippers, and micromanipulators - and summarizes the significance and primary applications of each category. Finally, key challenges, particularly those related to sensing and powering microrobots, are examined, along with current perspectives and potential solutions.