机器人

This article examines the growing role of robotics in the cardiac catheterization laboratory as interventional cardiology moves toward more complex coronary and structural heart procedures. Robotic systems improve precision through motion scaling and tremor reduction while reducing operator fatigue and cumulative radiation exposure. The article argues that successful adoption depends not only on mechanical accuracy, but also on workflow integration, human-machine collaboration, safety design, regulatory readiness, and financial viability. It highlights the challenges of fitting robotic platforms into fast-paced cath lab environments where setup time, team coordination, and rapid conversion to manual control are critical. The discussion also explores remote robotic intervention to distribute specialist expertise across sites with limited staffing depth. Finally, the article considers how artificial intelligence, computer vision, and predictive analytics may extend robotic platforms from assistive tools into more intelligent systems that support safer, more scalable, and more accessible cardiovascular care.

Shape memory alloys (SMAs) are attractive for soft robotic actuation because of their compactness and high power density, yet their widespread application is limited by slow thermal recovery. Here, we introduce ionic wind cooling as a compact thermal-management strategy to overcome this bottleneck and improve the cyclic actuation performance of SMA-driven soft robotic systems. Two ionic wind configurations were designed and comparatively evaluated for localized cooling of SMA springs. Both significantly accelerated SMA recovery with only minimal additional power input, and the needle-ring configuration was selected for subsequent integration because of its favorable stability and compact geometry. By coupling SMA springs with an origami-based compression-twisting mechanism, an actuator-level soft twisting module capable of reversible bidirectional rotation exceeding 80° was developed. The cooling strategy was further extended to multiple SMA-driven modules and a reconfigurable soft robotic arm capable of coordinated twisting, bending, and multimodal grasping. The results show that ionic wind cooling not only enhances the recovery of individual SMA actuators but also enables faster motion switching and improved cyclic response at the system level. This work demonstrates ionic wind cooling as a compact and effective strategy for enhancing the dynamic performance of SMA-driven soft robotic systems.

While robotic surgery has expanded in living donor and transplant operations, it is unclear how the next generation of transplant surgeons will be trained to incorporate robotics into their practice. This study aims to assess current opinions regarding robotics training in transplant fellowship. This was a cross-sectional survey of 2025-2026 American Society of Transplant Surgeons (ASTS) abdominal transplant fellows and program directors (PDs), assessing experiences with robotics during fellowship, barriers to training, and future directions in robotic transplant education. Surveys were sent electronically across North America. This study included 2025-2026 ASTS abdominal transplant fellows and PDs. Eighty-one fellows and 19 PDs responded (response rates 55% and 45%, respectively). Seventy-three percent of fellows expected to participate in fewer than 50 cases during fellowship, 94% desired more robotic experience, and 35% expressed dissatisfaction with their training. Satisfaction correlated with higher transplant center volume, console exposure, and confidence in robotic procedures (all p&#x202f;<&#x202f;0.05). This study highlights limitations in robotic surgery training during abdominal transplant fellowships, revealing a gap between fellow expectations and current practice, and underscoring the need for standardized curricula and training.

The intelligent soft robotic gripper integrated with tactile sensors significantly enhances the robot's execution capabilities in complex tasks, resolving critical shortcomings of traditional mechanical grippers-namely, fragile item breakage from rigid impacts, irregular object slippage, and inefficiency due to recognition errors. While electrical sensors (e.g., piezoresistive, capacitive) struggle with structural complexity, signal crosstalk, and environmental interference, optical waveguide tactile sensing offers superior sensitivity, rapid dynamics, and electromagnetic immunity. However, existing waveguide tactile systems face two key limitations: millimeter-scale waveguides cause beam divergence, limiting deformation sensitivity and complicating heterogeneous integration. Additionally, critical gaps remain in adaptive grasping control and contextual object recognition during manipulation. Herein, we present a soft robotic gripper integrated with slender elastic optical waveguide sensors (EOWS) and equipped with a closed-loop feedback control module to achieve intelligent grasping and object attribute recognition. The hand comprises three flexible silicone fingers, each finger seamlessly integrates three EOWS for multi-modal tactile sensing. These sensors exhibit high sensitivity to bending angle (0.273%/&#xb0;), contact force (0.843%/N), and pressure (1.064%/N). Furthermore, a PID adaptive grasping control strategy and a long short-term memory (LSTM) deep learning algorithm are introduced to dynamically adjust the grasping force and intelligently recognize object attributes such as shape, size, and hardness, with accuracies exceeding 97% for each attribute. Ultimately, experimental validation via a smart fruit-sorting system highlights the platform's potential for precision agriculture, intelligent logistics, and medical robotics, demonstrating robust, adaptive manipulation in real-world applications. We present a soft robotic gripper seamlessly integrated with slender multi-modal elastic optical waveguide sensors (EOWS) and equipped with an adaptive control module to achieve intelligent grasping and object attribute recognition. Experimental validation via a smart fruit-sorting system highlights the platform's potential for precision agriculture, intelligent logistics, and medical robotics, demonstrating robust, adaptive manipulation in real-world applications.

Cervical arthrodesis requires precise pedicle screw placement to ensure safety and effectiveness. Traditional planning and execution are time-consuming and prone to variability. We developed a robot-assisted system integrating three components: an AI-based preoperative planning module, adapted from previous work, to generate patient-specific screw trajectory from 3D CT point-clouds; an intraoperative registration and motion compensation system with optical tracking to align the trajectory with patient anatomy in real time; and a comanipulation control strategy enforcing virtual fixtures and depth limits to guide the robotic arm safely. The system was tested on 3D-printed models and cadaveric specimens. Robotic assistance significantly improved the geometric accuracy of drilling, reducing transverse positional deviations by a factor of two and orientation deviations by a factor of eight compared with freehand procedures. In addition, the average drilling depth overshoot was reduced by 50%. Perforation rates were found to be of the same order as those observed with freehand techniques. The proposed workflow improves trajectory-following accuracy and depth control while preserving intuitive surgeon interaction. These results demonstrate the feasibility of integrating AI-based planning, intraoperative tracking, and collaborative robotics for cervical spine surgery.

Additive manufacturing enables fabrication of intricate electronic devices embedded in complex-shaped structural components. Here, we add a new member to the family of 3D-printed electronics - a high-performance 3D-printed magnetic field sensor featuring more than 300% magnetoimpedance effect at low frequencies and single point magnetic vector field reconstruction relying on 3D Hall effect magnetometry. The sensors are shaped as mechanically flexible and magnetically controllable springs as well as 3D crosses demonstrating abilities in tailoring the sensor response, operation field range, and operation frequency through rational design of the sensor geometry. The application potential of 3D-printed magnetoelectronics is featured via magnetic toggle switches for smart home, continuous joysticks for robotics control, three-axis magnetometers for volumetric multi-point detection, and self-aware 4D-printed mechatronic actuators. This technology enables 4D-printed structures that fold motion sensing directly into their design and enable each part to interact intelligently within a larger mechanism being aware of their environment and user&#xa0;interactions.

Smart hydrogels are capable of sensing and responding to external stimuli, demonstrating promising application in hydrogel actuators, soft robotics and other fields. The integration of hydrogel deformation with information encryption is important to help improve the storage density and security of information. In this study, a simple anodic electrical writing method is employed to graft catechol onto a chitosan/agarose double-network hydrogel, thereby enabling programmable spatial control of the patterned hydrogel. Meanwhile, the grafting process affects the physicochemical properties of the hydrogel, resulting in self-deformation in different solutions. Additionally, the introduction of dynamic boronate ester bonds endows the hydrogel with reversible multiple stimulus responsiveness (e.g., pH, glucose, and temperature), thus achieving reconfigurable deformation behaviors. The synergistic incorporation of electrically written information (Morse code) with the complex deformation of the hydrogel enables dual encryption. Owing to the redox property of catechol, different electrical signals can be generated in the written and unwritten areas. The security of the information is enhanced by using the electrical signals and Morse code for information reading. Furthermore, the hydrogel retains its deformation capability after 10&#xa0;cycles, and information stored within the written area remains readable via electrical signals even after 30&#xa0;days. This multi-responsive polysaccharide hydrogel prepared by electrical writing not only provides new insights for natural polysaccharide hydrogel actuators, but also stimulates their application potential in the field of information storage and encryption.

Additive manufacturing has enabled highly sophisticated three-dimensional soft bodies through material distribution, opening new possibilities for architected soft matter systems. In parallel, rapid development in soft robotics has intensified the necessity of compliant, distributed, and deformation-driven sensing solutions. Among possible solutions, 3D-printed metamaterial-based soft sensors are promising due to unprecedented mechano-sensing performances. Specifically, by integrating material design and structural topology into a unified functional entity, these metamaterial-based sensing solutions enable tunable stiffness, controlled instability, and efficient electromechanical transduction. With these promising findings in mind, this review aims to provide a comprehensive framework addressing additive manufacturing technologies, material systems, and sensing mechanisms in 3D-printed metamaterial-based sensors. In this work, available 3D printing technologies are discussed, highlighting trade-offs in resolution, multi-material capability, and structural fidelity. In parallel, a variety of 3D printing materials, including polymer-based, functional composite, and smart responsive materials, is examined, emphasizing material-structure interactions determining sensing performance. Subsequently, metamaterial-based transduction mechanisms are classified into resistive, capacitive, and inductive modalities, together with emerging multifunctional and multimodal sensing modality. Conclusively, by synthesizing fabrication technologies, material systems, and sensing architectures within an additive manufacturing perspective, this review provides design frameworks and outlooks for perceptive soft machines and their applications in real-world scenarios.

The convergence of embodied intelligence and world models has catalyzed growing interest in integrating physical laws into AI systems. While prior surveys have examined world models and embodied intelligence separately, we focus on the progression that connects these capabilities as a unified developmental pathway from passive observation to active physical comprehension. This survey provides a systematic framework revealing how physical AI advances through four interconnected stages: perception transforms sensory data into structured physical representations, reasoning derives explanations from observed phenomena, modeling enables predictive simulation grounded in physical principles, and embodied interaction closes the loop through physical manipulation and environmental feedback. Each stage enables and enhances the next: perceptual grounding supports causal reasoning, reasoning unlocks predictive capabilities, and robust models drive genuine physical interaction. Through analysis of developments spanning architectural innovations, training methodologies, causal inference, and embodied systems, we synthesize how physical understanding emerges through cumulative integration across this progression. Our framework reveals the evolution from isolated, task-specific solutions toward integrated architectures that advance from pattern recognition toward causal reasoning and counterfactual prediction. This perspective provides foundations for next-generation physical AI systems with direct implications for safe, generalizable, and interpretable deployment across robotics, scientific discovery, and autonomous systems. We maintain a continuously updated taxonomy repository at https://github.com/AI4Phys/Awesome-AI-for-Physics.

暂无摘要（点击查看详情）

Laparoscopic common bile duct exploration (LCBDE) remains an effective single-stage strategy for choledocholithiasis, but declining utilization has raised concerns about diminishing operative experience and training exposure among contemporary surgeons. We evaluated institutional trends in LCBDE, intraoperative cholangiography (IOC), and resident participation and identified predictors of successful duct clearance. This retrospective single-center cohort study (2012-2021) included adults undergoing laparoscopic or robotic cholecystectomy. Patients were grouped as single-stage LCBDE with cholecystectomy or two-stage management with ERCP and endoscopic sphincterotomy followed by cholecystectomy. Cases were identified using CPT and ICD procedure codes cross-referenced with operating room logs. Data were abstracted from the electronic medical record. Primary outcomes were temporal trends in LCBDE, IOC utilization, and resident participation. Secondary outcomes were predictors of successful LCBDE. LCBDE succeeded in 240/291 cases (82.4%). Failures occurred exclusively during transcystic exploration, most commonly due to an inability to cannulate the cystic duct (43%). Median length of stay was shorter after successful LCBDE (1 vs 3&#xa0;days, p&#x2009;<&#x2009;0.001). Surgeons in successful cases had greater experience (median 17.5 vs 12.9&#xa0;years, p&#x2009;=&#x2009;0.03). On multivariable analysis, surgeon experience independently predicted success (OR 1.04 per year, 95% CI 1.004-1.078), whereas surgeon specialty was not significant after adjustment. Of 303 LCBDE patients, 291 met inclusion criteria; among 339 two-stage patients, 257 met inclusion criteria. Institutionally, annual LCBDE volume declined (trend p&#x2009;<&#x2009;0.05), IOC utilization decreased (66% in 2018 to 54% in 2021), and resident involvement fell (91.8% in 2014 to 69.6% in 2021), with per-resident exposure declining from 2.2 to 0.9 cases per year. LCBDE achieves high duct-clearance rates and shorter hospitalization when successful. Surgeon experience, rather than specialty, independently predicts success, underscoring a clinically meaningful learning curve. Declines in LCBDE volume, IOC utilization, and resident exposure highlight the need for training strategies&#xa0;including simulation, standardized workflows, and improved case access&#xa0;to preserve competency.

Robotic assistance may overcome the limitations of intraocular visualization during glaucoma surgery, particularly in eyes with corneal opacity. This preclinical study investigated the feasibility and surgical utility of a robotic-assisted intraocular endoscope-holding system for real-time visualization during Preserflo MicroShunt implantation. An endoscope-holding robot (OQrimo) enabled stable, hands-free intraocular imaging throughout the procedure. Porcine eyes were used, and corneal opacity was simulated with a black-painted contact lens. Surgical feasibility, tube-endothelium (TE) distance, and intraoperative complications were evaluated. Continuous endoscopic visualization using the robotic system enabled successful MicroShunt implantation in all eyes without complications, including intraoperative mechanical trauma to the cornea, iris, or lens. The TE distance was significantly greater in the robotic group than in the conventional group under both normal visualization conditions (1.96&#x2009;&#xb1;&#x2009;0.27&#xa0;mm vs. 1.13&#x2009;&#xb1;&#x2009;0.40&#xa0;mm, p&#x2009;=&#x2009;0.009) and simulated corneal opacity (1.85&#x2009;&#xb1;&#x2009;0.32&#xa0;mm vs. 1.12&#x2009;&#xb1;&#x2009;0.28&#xa0;mm, p&#x2009;=&#x2009;0.015), indicating more favorable tube positioning. These results demonstrate that robotic-assisted intraocular endoscopic visualization provides stable and precise guidance for MicroShunt implantation independent of corneal clarity. This technology has the potential to enhance surgical accuracy and expand the feasibility of minimally invasive glaucoma surgery in eyes with compromised anterior segment visualization.

The evolution of soft robots into embodied intelligent systems relies fundamentally on precise proprioception. However, a universal solution for capturing continuous deformations during diverse interactions, particularly in spatially confined interventional scenarios, remains lacking. Here, we introduce a deep learning-enabled versatile shape perception method based on a single-ended multimode fiber (MMF). By leveraging the intrinsic integration advantages of optics, our minimalist reflective architecture physically eliminates the dependence on complex demodulation units and distal devices. Furthermore, treating chaotic optical speckle fields as data streams encoding high-dimensional shape information, reconfigurable neural decoders resolve a single physical channel into versatile perception modes tailored to heterogeneous tasks: discrete state confirmation on soft grippers (>99% accuracy), continuous shape tracking on bionic dexterous hands (~5-fold spatial resolution enhancement), and intuitive 3D morphological reconstruction of soft surgical robots (IoU>0.93). Overall, our work establishes a versatile framework for breaking hardware adaptability limits via computation, laying a solid foundation for closed-loop control in digital twins of soft robots.

Large language models (LLMs) are driving a paradigm shift in molecular science, transitioning from molecule understanding to autonomous scientific discovery. By learning from multimodal molecular representations through advanced training techniques, LLMs address traditional limitations in molecular research, including expert dependency and limited experimental scalability. We analyze strategies for cross-modal alignment and domain adaptation that enable LLMs to interpret complex molecular semantics. This deep semantic understanding translates into broad versatility across key downstream applications, encompassing retrieval tasks like molecule-text matching, prediction tasks like reaction outcome inference, and generative tasks including zero-shot molecule design. Crucially, we highlight how LLMs integrate with robotic platforms to establish closed-loop autonomous discovery systems, where AI agents automate hypothesis generation, experimental planning, and iterative validation. While these advances accelerate exploration of molecular space, persistent challenges remain in scaling experimental validation and bridging symbolic reasoning with physical experimentation. This review provides a comprehensive roadmap for leveraging LLMs to redefine scientific discovery in molecular science.

Dementia's rising prevalence places an immense burden on caregivers. Knowledge Graphs (KGs) and Large Language Model (LLM)-augmented KGs are emerging AI approaches that organize complex dementia care knowledge and enable personalized, context-aware support, yet this field remains nascent. We aimed to map and synthesize research on KGs and LLM-augmented KGs in dementia caregiving, identifying system types, applications, outcomes, challenges, and ethical considerations. Following the JBI framework, a comprehensive search was conducted across six academic databases (PubMed, Scopus, Web of Science, IEEE Xplore, PsycINFO, CINAHL) and grey literature. Eligibility criteria included studies detailing the design, development, or evaluation of KGs or LLM-augmented KGs for dementia caregiving. Twelve articles representing 11 unique studies met the inclusion criteria. All 11 studies used KG or ontology components; eight were KG-only systems, often supporting personalized meal planning, care plan recommendations, knowledge management, robotic assistance, or virtual assistants. Three studies described LLM-augmented KGs (3/11), primarily using retrieval-augmented generation to enhance conversational AI for caregivers or persons with dementia. Reported benefits included improved usability, personalized support, more accurate or relevant recommendations, and potential improvements in quality of life and independence. Key challenges involved technical complexity, KG maintenance, data quality, limited real-world evaluation, and underdeveloped ethical analysis. Integrating KGs with LLMs for dementia caregiving is a promising yet nascent interdisciplinary field. While early systems demonstrate potential, significant gaps remain in clinical validation, comprehensive ethical guidelines development, and responses to caregivers' diverse and evolving needs.

A numerical research work is carried out to examine fluid-structure interaction in flow over six inline square cylinders undergoing transverse oscillations via the lattice Boltzmann method. The impact of the oscillation frequency ratio ([Formula: see text]), where imposed oscillation frequency is represented as [Formula: see text] and [Formula: see text]is the natural vortex shedding frequency of a stationary cylinder, on wake dynamics and flow patterns is examined. Simulations are performed at Reynolds number of Re&#x2009;=&#x2009;80, with a non-dimensional gap spacing (s/d&#x2009;=&#x2009;1.0) and an oscillation amplitude ratio of A/d&#x2009;=&#x2009;0.2. The frequency ratio is varied in the range 0.1 < [Formula: see text]&#x2264; 2.0. Three distinct flow patterns are identified: synchronous lock-on (0.8 &#x2264; [Formula: see text]&#x2264; 1.4), quasi-periodic lock-on-I (1.6 &#x2264; [Formula: see text] &#x2264; 2.0) and quasi-periodic non-lock-on-I (0.1 &#x2264; [Formula: see text]&#x2264; 0.6). The synchronization range is wider than that reported for single oscillating cylinder. A single coherent wake envelops the entire cylinder array, with wake recovery occurring at higher oscillation frequencies due to merging of vortex and formation of multi-polar vortices downstream of final cylinder. The first cylinder experiences highest mean drag, followed by a reduction and gradual increase along the downstream cylinders.

The objectives of this study were to determine the accuracy of component positioning, patient satisfaction, functional outcomes, component survivorship, and complications of robotic total hip arthroplasty (THA) in patients who have atypical hip anatomy. This study included 192 robotic THAs performed in 182 patients for developmental dysplasia of the hip (n = 122), Leg-Calve-Perthes disease (n = 27), slipped capital femoral epiphysis (n = 20), previous acetabular fracture (n = 12), and skeletal dysplasia (n = 11). Predefined radiological outcomes, patient satisfaction, University of California at Los Angeles (UCLA) activity score, Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), Oxford Hip Score (OHS), Forgotten Joint Score (FJS), component survivorship, and any complications were recorded. The mean follow-up time was 3.8 &#xb1; 1.7 years (range, 2.1 to 5.4). Robotic THA was associated with high levels of accuracy in executing the planned horizontal (root mean square error (RMSE) 1.5 &#xb1; 1.4 mm) and vertical centers of rotation (RMSE: 1.8 &#xb1; 1.7 mm), combined offset (RMSE: 2.9 &#xb1; 3.8 mm), and leg-length correction (RMSE: 1.5 &#xb1; 1.4 mm). Acetabular component positioning within Lewinnek's safe zones was 94.7%, and Callanan's safe zones was 93.8%. The median patient satisfaction score was 90 (interquartile range (IQR), 85 to 95), the median WOMAC score was 18 (IQR, 12 to 20), and the mean FJS score was 77.8 &#xb1; 10.8 at final follow-up. Robotic THA was associated with improvements in the mean UCLA (P < 0.001) and OHS (P < 0.001) at final follow-up. The five-year survivorship was 98.8% (95% CI [confidence interval]: 95.8 to 100) with implant revision for any reason as the end point. Robotic THA in patients who have atypical hip anatomy was associated with high levels of accuracy in executing the planned component positioning. In this challenging patient population, robotic THA was associated with encouraging early component survivorship, satisfactory functional outcomes, and low risk of complications at short-term follow-up.

Surface electromyography (sEMG) is widely adopted in upper-limb rehabilitation for decoding motor intent. However, current wearable systems typically require eight or more channels, imposing penalties on device complexity, power consumption, and patient comfort. Reducing channel count without sacrificing recognition accuracy remains an open challenge for user-independent rehabilitation scenarios. We hypothesize that systematically optimizing channel configurations under high-fidelity acquisition conditions can identify a minimal viable topology suitable for clinical deployment. We present a hierarchical validation framework integrating three pillars of evidence. First, we developed a custom acquisition system achieving a common-mode rejection ratio exceeding 112&#xa0;dB, ensuring that observed inter-channel differences reflect physiology rather than hardware artifacts. Second, we performed an exhaustive subset search across all 63 non-empty combinations of six channels in 24 healthy subjects under leave-one-subject-out (LOSO) cross-validation. Third, a dual-track sensitivity analysis was conducted by training on physiologically augmented data while testing under hardware-degraded conditions. Results reveal a tiered structure: the 3-channel configuration serves as the efficiency knee point (91.65%), exceeding the 90% clinical reliability threshold, although its cross-subject generalizability is limited by a maximum regret of 10.5% and only 8/24 subjects achieving zero regret under the global 3-channel topology. The 4-channel configuration acts as the recommended optimal (94.44%), eliminating worst-case blind spots ([Formula: see text]). An overlap ratio analysis confirms high alignment between global and individual optima, with 67% of subjects achieving zero regret under the global 4-channel topology. The dual-track sensitivity analysis demonstrates a 16% robustness gain under low-SNR conditions, validating that high-fidelity acquisition is the functional prerequisite for channel reduction. Within the six-site candidate pool studied here, the globally fixed 4-channel topology achieved performance close to the 6-channel baseline under rigorous LOSO validation. While the 3-channel configuration meets the 90% clinical threshold on average, its substantial inter-subject regret (up to 10.5%) limits its suitability as a universal solution. The optimal 4-channel topology converges on four forearm muscles spanning the wrist's two kinematic degrees of freedom, providing an anatomically grounded, patient-independent design principle. These findings offer a practical pathway toward lightweight, comfortable, and clinically viable rehabilitation interfaces that could improve patient compliance and therapeutic outcomes.

This paper addresses the scaled containment control problem (SCCP) for first/second-order stochastic multi-agent systems (SMASs) in a noisy environment. The agents receive neighbor's state information subjected to nonzero scaling factors, which causes the followers to finally converge to a scaled deterministic constant formed by the leaders. A stochastic approximation protocol with time-varying gains is designed to attenuate the noise effects. Using a state decomposition method, some sufficient and necessary conditions for the SCCP of first/second-order SMASs are given under the topological requirement of containing a directed spanning forest. For first-order systems, the followers converge to the scaled deterministic constant formed by the leaders. For second-order systems, two interaction modes are analyzed: under constant velocities mode, followers' positions converge to the unbounded scaled deterministic constant formed by the leaders' positions and followers' velocities converge to the scaled deterministic constant formed by the leaders' velocities; under zero velocity mode, the convergence behavior of the second-order systems is analogous to that of first-order systems, where followers' positions converge to the scaled deterministic constant formed by the leaders' positions and followers' velocities converge to zero. The numerical examples demonstrate the validity of the theoretical findings.

Composite magnetoelectric (piezoelectric/magnetostrictive) materials are gaining an increased interest due to their appealing wireless actuation capabilities. For many applications, such as sensing, energy harvesting or cell electrostimulation, flexible and adaptable heterostructures are required. However, some of the proposed structures suffer from poor magnetoelectric performance due to their inherent stiffness and the resulting substrate clamping effects. Here, a highly flexible and conformable nanostructured ZnO(piezoelectric)/FeGa(magnetostrictive) magnetoelectric heterostructure embedded in a low Young's modulus elastomer (polydimethylsiloxane; PDMS) is presented. This heterostructure is designed to&#xa0;enhance wound healing&#xa0;via electric stimulation induced by low-intensity and low-frequency AC magnetic fields. In vitro testing using normal human dermal fibroblasts (NHDF) and human keratinocytes (HaCaT) cell cultures demonstrated high cell viability and no adverse effects under magnetoelectric stimulation. Moreover, the magnetoelectric stimulation significantly enhanced keratinocyte stratification and fibroblast collagen production, with remarkable improvements in cell migration compared to non-stimulated cells. These results underscore the potential of highly flexible magnetoelectric heterostructures as active dressings to improve wound healing processes, especially relevant for chronic wounds and other epithelial disorders.