Multisite electrophysiological monitoring of ex vivo tissues and organ models is essential for basic research and drug toxicity evaluation. However, conventional microelectrode arrays with fixed positions and rigid structures are insufficient for dynamic, curved tissue surfaces. Here, we present a magnetically actuated soft electrode (MSE) with precise navigation, adaptive attachment, and high-fidelity signal acquisition. Operating in a “locate–adhere–record–detach” cycle, the MSE enabled continuous multisite detection on beating ex vivo tissues. In isolated rat heart experiments, the MSE demonstrated millimeter-level navigation accuracy, stable contact, and high signal-to-noise ratio (average 28 dB). By integrating magnetic locomotion with electrophysiological sensing, this work establishes a programmable, actively addressable platform for multisite electrophysiological monitoring of organ models, tissue slices, and engineered constructs, offering broad potential for cardiotoxicity screening and cardiovascular research.

Sound target detection (STD) plays a critical role in modern acoustic sensing systems. However, existing automated STD methods show poor robustness and limited generalization, especially under low signal-to-noise ratio (SNR) conditions or when processing previously unencountered sound categories. To overcome these limitations, we first propose a brain–computer interface (BCI)-based STD method that utilizes neural responses to auditory stimuli. Our approach features the Triple-Region Spatiotemporal Dynamics Attention Network (Tri-SDANet), an electroencephalogram (EEG) decoding model incorporating neuroanatomical priors derived from EEG source analysis to enhance decoding accuracy and provide interpretability in complex auditory scenes. Recognizing the inherent limitations of stand-alone BCI systems (notably their high false alarm rates), we further develop an adaptive confidence-based brain–machine fusion strategy that intelligently combines decisions from both the BCI and conventional acoustic detection models. This hybrid approach effectively merges the complementary strengths of neural perception and acoustic feature learning. We validate the proposed method through experiments with 16 participants. Experimental results demonstrate that the Tri-SDANet achieves state-of-the-art performance in neural decoding under complex acoustic conditions. Moreover, the hybrid system maintains reliable detection performance at low SNR levels while exhibiting remarkable generalization to unseen target classes. In addition, source-level EEG analysis reveals distinct brain activation patterns associated with target perception, offering neuroscientific validation for our model design. This work pioneers a neuro-acoustic fusion paradigm for robust STD, offering a generalizable solution for real-world applications through the integration of noninvasive neural signals with artificial intelligence.

Insects have been integrated with electronic systems to create cyborg insects for various practical applications by utilizing their inherent adaptability and mobility. Nevertheless, most cyborg insects’ preparation depends on the invasive method, which can cause harm to critical sensory organs and restrict the obstacle-negotiating capabilities of cyborg insects. We present wearable devices with headgear and abdominal buckle that address these challenges using hooking mechanisms, multimaterial 3-dimensional printing, and selective electroless plating. These devices attach securely to the antenna scape and abdominal tergum without damaging functional organs, thereby preserving the insect’s natural sensory functions and physical intactness. Besides, the electrodes attach and detach easily without using adhesives, reducing the time required for cyborg insect preparation and enabling the reuse of insects. Experiments show that cyborg insects with wearable devices spend less time traversing obstacles than those prepared using invasive methods. Additionally, the potential for practical navigation tasks is further demonstrated by the cyborg insect’s capacity to navigate along the “S”-path. This work advances scalable, efficient, and ethical utilization of cyborg insects in the fields of robotics and biohybrid systems.

Biological imaging has revolutionized tissue analysis by revealing morphological and physiological dynamics, yet faces inherent limitations in penetration depth and resolution. Micro/nanomotors (MNMs), with autonomous propulsion and spatiotemporal control, offer transformative solutions to traditional static imaging paradigms. These dynamic contrast agents enhance detection sensitivity in ultrasound, fluorescence, photoacoustic, and magnetic resonance imaging via motion-amplified signal modulation, enabling real-time tracking of subcellular events and microenvironmental changes. While MNMs-enhanced bioimaging has advanced rapidly, systematic analysis of their mechanisms and challenges remains limited. Based on our research experience in this field, this paper first summarizes the signal-enhancing mechanisms of MNMs in single-modal imaging. It then explores multimodal applications through MNMs-probe design and discusses artificial intelligence-driven intelligent MNMs for precision imaging. Finally, challenges and outlook are outlined, aiming to provide a theoretical framework and research roadmap for MNMs-mediated bioimaging technologies.

Growing evidence highlights the importance of body composition (BC), including bone, muscle, and adipose tissue (AT), as a critical biomarker for cardiometabolic risk stratification. However, conventional methods for quantifying BC components using medical images are hindered by labor-intensive workflows and limited anatomical coverage. This study developed BioCompNet—an end-to-end deep learning workflow that integrates dual-parametric magnetic resonance imaging (MRI) sequences (water/fat) with a hierarchical U-Net architecture to enable fully automated quantification of 15 biomechanically critical BC components. BioCompNet targets 10 abdominal compartments (vertebral bone, psoas muscles, core muscles, subcutaneous AT [SAT], superficial SAT, deep SAT, intraperitoneal AT, retroperitoneal AT, visceral AT, and intermuscular AT [IMAT]) and 5 thigh compartments (femur, muscle, SAT, IMAT, and vessels). The workflow was developed on 8,048 MRI slices from a community-based cohort (n = 503) and independently validated on 240 MRI slices from a tertiary hospital (n = 30). The model’s performance was benchmarked against expert annotations. On internal and external validation datasets, BioCompNet achieved average Dice similarity coefficients of 0.944 and 0.938 for abdominal compartments and 0.961 and 0.936 for thigh compartments, respectively. Excellent interreader reliability was observed (intraclass correlation coefficient ≥ 0.881) across all quantified features, and IMAT quantification showed a strong linear trend (Ptrend < 0.001) compared to physician-rated assessments. The workflow substantially reduced processing time from 128.8 ± 5.6 to 0.12 ± 0.001 min per case. By enabling rapid, accurate, and comprehensive volumetric analysis of BC components, BioCompNet establishes a scalable framework for precision cardiometabolic risk assessment and clinical decision support.

With the increasing use of computed tomography (CT), concerns about radiation dose have grown. Deep-learning-based methods have shown great promise in improving low-dose CT image quality while further reducing patient dose. However, most deep-learning-based methods are trained on vendor-specific CT datasets with varying imaging conditions and dose levels, which results in poor generalizability across vendors due to marked data heterogeneity. Moreover, the centralization of multicenter datasets is restricted by the high costs of data collection and privacy regulations. To overcome these challenges, we propose FedM2CT, a federated metadata-constrained method with mutual learning for all-in-one CT reconstruction. This method enables simultaneous reconstruction of multivendor CT images with different imaging geometries and sampling protocols in one framework. Specifically, FedM2CT consists of 3 modules: task-specific iRadonMAP (TS-iRadonMAP), condition-prompted mutual learning (CPML), and federated metadata learning (FMDL). TS-iRadonMAP performs task-specific low-dose reconstruction, CPML shares condition-prompted knowledge between clients and the server, and FMDL aggregates model parameters with a metamodel to effectively mitigate the effect of data heterogeneity. Extensive experiments under 3 different settings demonstrate that the proposed FedM2CT achieves outstanding results compared to other methods, both qualitatively and quantitatively, showing the potential to achieve the goal of all-in-one CT reconstruction with different low-dose tasks, i.e., low-milliampere-second, sparse-view, and limited-angle.

C-tactile afferents are low-threshold mechanoreceptors that innervate the hairy skin of mammals, essential for emotional interactions. Replication of such a mechanism could facilitate emotional interactions between humans and embodied intelligence robotic systems. Herein, we demonstrate a monolithic synaptic device that replicates and integrates tactile sensing and neuromorphic processing functions for in-sensor computing. The device is operable by both mechanical and electrical inputs, with the mechanoelectrical operation mechanism stemming from the synergistic effect of dynamic ionic migration and injection. As a proof of concept, the device effectively converts spatiotemporal tactile stimuli into distinct electrical signals, which are subsequently encoded to enable the microcomputer to classify multiple discrete emotional states, such as happiness, calmness, and excitement. This monolithic integrated device, which converges mild tactile perception with neuromorphic processing, with high tactile sensitivity and low-energy consumption, establishes an approach for emotional interaction between intelligent robots and human beings.