Deep learning relies on learning from extensive data to generate prediction results. This approach may inadvertently capture spurious correlations within the data, leading to models that lack interpretability and robustness. Researchers have developed more profound and stable causal inference methods based on cognitive neuroscience. By replacing the correlation model with a stable and interpretable causal model, it is possible to mitigate the misleading nature of spurious correlations and overcome the limitations of model calculations. In this survey, we provide a comprehensive and structured review of causal inference methods in deep learning. Brain-like inference ideas are discussed from a brain-inspired perspective, and the basic concepts of causal learning are introduced. The article describes the integration of causal inference with traditional deep learning algorithms and illustrates its application to large model tasks as well as specific modalities in deep learning. The current limitations of causal inference and future research directions are discussed. Moreover, the commonly used benchmark datasets and the corresponding download links are summarized.

Robo-pigeons, a novel class of hybrid robotic systems developed using brain–computer interface technology, hold marked promise for search and rescue missions due to their superior load-bearing capacity and sustained flight performance. However, current research remains largely confined to laboratory environments, and precise control of their flight behavior, especially flight altitude regulation, in a large-scale spatial range outdoors continues to pose a challenge. Herein, we focus on overcoming this limitation by using electrical stimulation of the locus coeruleus (LoC) nucleus to regulate outdoor flight altitude. We investigated the effects of varying stimulation parameters, including stimulation frequency (SF), interstimulus interval (ISI), and stimulation cycles (SC), on the flight altitude of robo-pigeons. The findings indicate that SF functions as a pivotal switch controlling the ascending and descending flight modes of the robo-pigeons. Specifically, 60 Hz stimulation effectively induced an average ascending flight of 12.241 m with an 87.72% success rate, while 80 Hz resulted in an average descending flight of 15.655 m with a 90.52% success rate. SF below 40 Hz did not affect flight altitude change, whereas over 100 Hz caused unstable flights. The number of SC was directly correlated with the magnitude of altitude change, enabling quantitative control of flight behavior. Importantly, electrical stimulation of the LoC nucleus had no significant effects on flight direction. This study is the first to establish that targeted variation of electrical stimulation parameters within the LoC nucleus can achieve precise altitude control in robo-pigeons, providing new insights for advancing the control of flight animal–robot systems in real-world applications.

After years of research and development, flexible sensors are gradually evolving from the traditional “electronic” paradigm to the “ionic” dimension. Smart flexible sensors derived from the concept of ion transport are gradually emerging in the flexible electronics. In particular, ionic hydrogels have increasingly become the focus of research on flexible sensors as a result of their tunable conductivity, flexibility, biocompatibility, and self-healable capabilities. Nevertheless, the majority of existing sensors based on ionic hydrogels still mainly rely on external power sources, which greatly restrict the dexterity and convenience of their applications. Advances in energy harvesting technologies offer substantial potential toward engineering self-powered sensors. This article reviews in detail the self-powered mechanisms of ionic hydrogel self-powered sensors (IHSSs), including piezoelectric, triboelectric, ionic diode, moist-electric, thermoelectric, potentiometric transduction, and hybrid modes. At the same time, structural engineering related to device and material characteristics is discussed. Additionally, the relevant applications of IHSS toward wearable electronics, human–machine interaction, environmental monitoring, and medical diagnostics are further reviewed. Lastly, the challenges and prospective advancement of IHSS are outlined.

Ultrasound localization microscopy (ULM) is a novel imaging technique that overcomes the diffraction limit to achieve super-resolution imaging at the 10-μm scale. Despite its remarkable progress, challenges persist in enhancing the precision of microbubble tracking and fulfilling the requirements for high frame rates in practical circumstances, especially in moving organs. To address these issues, an enhanced ULM approach (shorten as vc-Kalman) integrating rapid motion compensation was developed to achieve excellent image quality. Unlike traditional methods relying on observed bubble positions, the proposed algorithm combined statistical information derived from historical data with Kalman-filter-predicted positions to enable more accurate bubble localization. Meanwhile, microbubble brightness in adjacent frames was incorporated as multidimensional feature to further improve the matching efficacy. Furthermore, velocity constraint was applied to minimize possible erroneous traces and enhance the contrast-to-noise ratio of ULM images, while ensuring the continuity of vascular reconstruction and the accuracy of the blood flow analysis to generate a reduced normalized root mean square error in velocity estimation, even at a relatively low frame rate of 146 Hz. More important, to effectively suppress the impact of physiological movements in moving organs like kidneys, this algorithm fulfilled subpixel displacement vector identification through parabolic fitting and expedited motion compensation via dynamic programming-based cross-correlation search. The results indicated that this advanced vc-Kalman method substantially boosted both the robustness and accuracy of ULM imaging, thereby opening more opportunities for clinical applications of super-resolution ULM technology.

Ultraviolet (UV) light, invisible to the human eye, possesses both benefits and risks. To harness its potential, UV photodetectors (PDs) have been engineered. These devices can convert UV photons into detectable signals, such as electrical impulses or visible light, enabling their application in diverse fields like environmental monitoring, healthcare, and aerospace. Wide bandgap semiconductors, with their high-efficiency UV light absorption and stable opto-electronic properties, stand out as ideal materials for UV PDs. This review comprehensively summarizes recent advancements in both traditional and emerging wide bandgap-based UV PDs, highlighting their roles in UV imaging, communication, and alarming. Moreover, it examines methods employed to enhance UV PD performance, delving into the advantages, challenges, and future research prospects in this area. By doing so, this review aims to spark innovation and guide the future development and application of UV PDs.

Edge contact is essential for achieving the ultimate device pitch scaling of stacked nanosheet transistors with monolayer 2-dimensional (2D) channels. However, due to large edge-contact resistance between 2D channels and contact metal, there is currently a lack of high-performance edge-contact device technology for 2D material channels. Here, we report high-performance edge-contact monolayer molybdenum disulfide (MoS₂) field-effect transistors (FETs) utilizing well-controlled plasma etching techniques. Plasma etching with pure argon improves the edge dangling bonds and thus improves the edge-contact quality. Edge-contact monolayer MoS₂ FET shows good ohmic contact even at cryogenic temperatures (20 K), achieving a record-low contact resistance (R_c) of 1.25 kΩ·μm among all edge-contact MoS₂ devices. The record-high on-state current of 436 μA/μm and transconductance of 123 μS/μm at V_ds = 1 V are achieved on an edge-contact monolayer MoS₂ FET with L_ch = 120 nm. This work highlights the great potential of edge contacts for high-performance monolayer transition metal dichalcogenide (TMD) material electronics.

Nanozymes are a class of nanomaterials that exhibit catalytic functions analogous to those of natural enzymes. They demonstrate considerable promise in the biomedical field, particularly in the treatment of bone infections, due to their distinctive physicochemical properties and adjustable catalytic activities. Bone infections (e.g., periprosthetic infections and osteomyelitis) are infections that are challenging to treat clinically. Traditional treatments often encounter issues related to drug resistance and suboptimal anti-infection outcomes. The advent of nanozymes has brought with it a new avenue of hope for the treatment of bone infections.

Tryptophan (Trp), an essential amino acid, performs as a precursor for synthesizing various bioactive molecules primarily metabolized through the kynurenine (Kyn), serotonin, and indole pathways. The diverse metabolites were deeply implicated in multiple physiological processes. Emerging research has revealed the multifaceted contribution of Trp in skeletal health and pathophysiology of bone-related disease with the involvement of specific receptors including aryl hydrocarbon receptor (AhR), which modulated the downstream signaling pathways to manage the expression of pivotal genes and thereby altered cellular biological processes, such as proliferation and differentiation. Accompanied by distinct alterations in immune function, inflammatory responses, endocrine balance, and other physiological aspects, their impact and efficacy in osteochondrogenic disorders have also been well documented. Nevertheless, a thorough understanding of Trp metabolism within bone biology is currently lacking. In this review, we elucidate the complexities of Trp metabolic pathway and several metabolites, delineating their versatile modulatory roles in the physiology and pathology of osteoblasts (OBs), osteoclasts (OCs), chondrocytes, and intercellular coupling effects, as well as in the progression of osteochondral disorder. Moreover, we comprehensively delineate the regulatory mechanisms by which gut microbiota-generated indole derivatives mediate bidirectional crosstalk along the gut–bone axis. The establishment of an elaborate governing network about bone homeostasis provides a novel insight on therapeutic interventions.

Ocean acidification is becoming more prevalent and may contribute to coral reef degradation, yet our understanding of its role in global reef decline remains limited. Therefore, there is an urgent need to study the impact of reduced pH levels on the growth patterns of major reef-building corals. Here, we studied the skeleton-forming strategies of 4 widely distributed coral species in a simulated acidified habitat with a pH of 7.6 to 7.8. We reconstructed and visualized the skeleton-forming process, quantified elemental calcium loss, and determined gene expression changes. The results suggest that different reef-building corals have diverse growing strategies in lower pH conditions. A unique “cavity-like” forming process starts from the inside of the skeletons of Acropora muricata, which sacrifices skeletal density to protect its polyp–canal system. The forming patterns in Pocillopora damicornis, Montipora capricornis, and Montipora foliosa were characterized by “osteoporosis”, exhibiting disordered skeletal structures, insufficient synthesis of adhesion proteins, and low bone mass, correspondingly. In addition, we found that damage from acidification particularly affects pre-existing skeletal structures in the colony. These results enhance our understanding of skeleton-forming strategies in major coral species under lower pH conditions, providing a foundation for coral reef protection and restoration amidst increasing ocean acidification.

Ferroptosis plays a role in wound healing during the maturation of senescent endothelial cells. This study explores the modulation of ferroptosis in senescent human umbilical vein endothelial cells (HUVECs) and wound-healing processes by Piezo1 activation at the molecular, cellular, and tissue levels. Elevated Piezo1 expression was observed in HUVECs treated with the senescence inducer doxorubicin (Doxo) and the ferroptosis inducer erastin and in aged wound tissue. Pharmacological inhibition or knockdown of Piezo1 protected senescent HUVECs and aged wound tissue from ferroptosis. Additionally, Piezo1 channel activity was found to promote ferroptosis in senescent HUVECs by increasing intracellular Ca²⁺ levels. The calmodulin-dependent kinase II (CaMKII)/activating transcription factor 3 (ATF3)/SLC7A11 signaling axis was activated upon stimulation with erastin and Doxo, driving Piezo1-induced ferroptosis. CaMKII directly interacted with ATF3, which could be modulated through Piezo1 channel regulation. Notably, Piezo1 knockout mice or adeno-associated virus 9-mediated silencing of ATF3 attenuated ferroptosis in senescent cells and accelerated wound repair. Mechanistically, both genetic and pharmacological inhibition of Piezo1 promoted wound healing in aged tissues and regulated ferroptosis in senescent HUVECs through the CaMKII/ATF3/SLC7A11 pathway. In conclusion, these findings suggest that targeting Piezo1-mediated ferroptosis in senescent HUVECs offers a promising therapeutic approach for improving wound healing in the elderly.