Accurate boundary segmentation of volumetric images is a critical task for image-guided diagnosis and computer-assisted intervention. It is challenging to address the boundary confusion with explicit constraints. Existing methods of refining boundaries overemphasize the slender structure while overlooking the dynamic interactions between boundaries and neighboring regions. In this paper, we reconceptualize the mechanism of boundary generation via introducing Pushing and Pulling interactions, then propose a unified network termed PP-Net to model shape characteristics of the confused boundary region. Specifically, we first propose the semantic difference module (SDM) from the pushing branch to drive the boundary towards the ground truth under diffusion guidance. Additionally, the class clustering module (CCM) from the pulling branch is introduced to stretch the intersected boundary along the opposite direction. Thus, pushing and pulling branches will furnish two adversarial forces to enhance representation capabilities for the faint boundary. Experiments are conducted on four public datasets and one in-house dataset plagued by boundary confusion. The results demonstrate the superiority of PP-Net over other segmentation networks, especially on the evaluation metrics of Hausdorff Distance and Average Symmetric Surface Distance. Besides, SDM and CCM can serve as plug-and-play modules to enhance classic U-shape baseline models, including recent SAM-based foundation models. Source codes are available at https://github.com/EndoluminalSurgicalVision-IMR/PnPNet.

Conventional drug delivery routes, such as oral and injectable administration, often suffer from poor targeting, systemic sideeffects, and limited control over drug release. Magnetically guided microrobots have emerged as a promising solution by enablingminimally invasive navigation, precise localization, and controllable cargo delivery in complex biological environments. Recentadvances in microfabrication, biomaterials, and magnetic actuation have led to diverse microrobot designs, including helicalswimmers and biohybrid systems incorporating red blood cells, bacteria, algae, or exosomes. These microrobots can transportdrugs, cells, or theranostic agents, with release triggered by magnetic, chemical, optical, or acoustic stimuli. In parallel, progressin imaging and tracking technologies, such as ultrasound, photoacoustic imaging, X-ray, and MRI, has enabled real-time guidanceand monitoring in vivo. Despite encouraging proof-of-concept studies, several challenges hinder clinical translation. These includelimited biocompatibility, reduced locomotion efficiency under physiological flow, insufficient release precision, and regulatoryconstraints. Future development requires safer and more robust materials, multimodal imaging strategies for accurate navigation,and scalable fabrication methods that meet medical standards. With continued integration of engineering and biomedical research,magnetically guided microrobots hold strong potential for targeted therapy, regenerative medicine, and minimally invasivetreatment, representing a significant advance in precision drug delivery.

Background and aims: Paediatric myocardial strain analysis through echocardiography is often characterized by high variance and limited precision, highlighting the need for a standardized and vendor-agnostic approach applicable for diverse image qualities and populations, which could enhance cardiac function evaluation and enable early detection of cardiac impairment. Methods: The Motion-Echo system was proposed, a semi-supervised deep learning framework built on 11 096 paediatric and 11 297 adult echocardiograms spanning diverse image qualities and vendors. It integrated context compensation and motion estimation modules for temporally coherent segmentation, myocardial motion estimation, and global strain assessment with minimal manual annotations. Clinical utility was further evaluated through downstream applications. Results: Motion-Echo achieved mean absolute errors of 2.099% [95% confidence interval (CI) 1.803-2.401] and 2.665% (95% CI 2.339-3.026) for global longitudinal and circumferential strain assessments, with Pearson correlation coefficients of 0.799 (95% CI 0.715-0.871) and 0.781 (95% CI 0.687-0.844), respectively. To validate the clinical utility, automated strain values achieved an area under the curve (AUC) of 0.906 (95% CI 0.816-0.981) for cancer therapy-related cardiac dysfunction risk prediction. For late gadolinium enhancement detection, automated global longitudinal strain reached an AUC of 0.782 (95% CI 0.666-0.885). For left ventricular ejection fraction decline forecasting, the system outperformed manual strain values (DeLong P < .001). In addition, incorporating estimated motion flows yielded a remarkable AUC improvement to 0.952 (95% CI 0.917-0.980) for myocardial infarction detection. Conclusions: Leveraging a large-scale paediatric dataset, Motion-Echo provided a reliable and generalizable framework for myocardial strain analysis, demonstrating potential to facilitate earlier detection of cardiac dysfunction and generate digital cardiac function profiles of children. Keywords: Deep learning; Echocardiography; Motion estimation; Myocardial strain; Paediatric cardiology; Semi-supervised learning.

Depth and odometry estimation for endoscopic imaging is an essential task for robot assisted endoluminal intervention. Due to the difficulty of obtaining sufficient in vivo ground truth data, unsupervised learning is preferred in practical settings. Existing methods, however, are hampered by imaging artifacts and the paucity of unique anatomical markers, coupled with tissue motion and specular reflections, leading to the poor accuracy and generalizability. In this work, a trimming-then-augmentation framework is proposed. It uses a “mask-then-recover” training strategy to firstly mask out the artifact regions and then reconstruct the depth and pose information based on the global perception of a convolutional network. Subsequently, an augmentation module is used to provide stable correspondence between endoscopic image pairs. A task-specific loss function guides the augmentation module to adaptively establish stable feature pairs for improving the overall accuracy of subsequent 3D structural reconstruction. Detailed validation has been performed with results showing that the proposed method can significantly improve the accuracy of existing state-of-the-art unsupervised methods, demonstrating the effectiveness of the method and its resilience to image artifacts, in addition to its stability when applied to in vivo settings.

Background: Epilepsy is a heterogeneous syndrome. Personalised localisation of epileptogenic zone (EZ) is critical for diagnosis and treatment of drug-resistant focal epilepsy. Multichannel stereoelectroencephalography (SEEG) monitoring acquired over a period of two to three weeks was collected in different patients, resulting in comprehensive epileptogenic information and terabytes of high dimensional data. Consequently, there is a need for high-throughput data analytical methods to enable data-driven, personalised seizure detection and EZ localisation. Methods: Here, a seizure detection and EZ localisation AI system — SEEGformer is proposed, by utilising SEEG data from 61 patients acquired across two centres and three cohorts capturing tens of thousands of abnormal discharges and around ten seizures per person on average. SEEGformer employs a parallel transformer architecture to analyse multiple representations of multichannel SEEG signals, including the real part, imaginary part, and amplitude of the analytic signal after Fourier transform. The MRI information was encoded in SEEGformer to construct the structural dependence of the brain areas. Inter-channel dependencies and interactions were captured for seizure detection. A cross-channel attention mechanism computed the epileptogenic risk score for each channel to localise EZ using ictal SEEG data. Each patient’s SEEG data was used to train and validate their individual-specific SEEGformer model. Findings: In three clinical cohorts, SEEGformer achieved an average AUROC of 0.937 (95% CI, 0.922–0.950) for seizure detection and 0.798 (95% CI, 0.749–0.847) for EZ localisation. Localisation performance surpassed state-of the-art methods by over 5%. SEEGformer further revealed distinct phase synchronisation patterns in dynamically evolving epileptogenic zone networks, with a significance level of P < 0.0001. Interpretation: Due to its high interpretability and visualisation capabilities, SEEGformer can enhance clinical decision-making by providing an objective, data-driven reference to optimise epileptogenic zone delineation and surgical strategy development. Currently, the improved SEEGformer is being developed to construct a dedicated SEEG atlas for epilepsy.

Through fusing materials into thin, elongated strands with structured cross-sections, recent advances in multimaterial fibers have enabled myriad applications for simultaneous sensing of many different inputs. Addressing the underlying challenges of preform fabrication, thermal compatibility of dissimilar materials with heterogeneous properties, and functional material deposition enables the creation of flexible embodiments such as smart fabrics or miniaturized devices for in vivo applications. By leveraging micro-nano fabrication techniques, microscale devices can be integrated at the tip of the fiber with photo, thermal, magnetic, electro, pneumatic, or hydraulic actuation, leading to thin, tethered fiberbots for targeted therapy and endoluminal intervention.

Medical image registration, which establishes spatial correspondences between different medical images, serves as a fundamental process in numerous clinical applications and diagnostic workflows. Despite significant advancement in unsupervised deep learning-based registration methods, these approaches consistently yield suboptimal results compared to their weakly-supervised counterparts. Recent advancements in universal segmentation models have made it easier to obtain anatomical labels from medical images. However, existing registration methods have not fully leveraged the rich anatomical and structural prior information provided by segmentation labels. To address this limitation, we propose a BIdirectional feature Alignment and Semantics-guided Network, referred to as BIASNet, for weakly-supervised image registration. Specifically, starting from multi-scale features extracted from the pre-trained VoCo, fine-tuned using Low-Rank Adaptation (LoRA), we propose a dual-attribute learning scheme, incorporating a novel BIdirectional Alignment and Fusion (BIAF) module for extracting both semantics-wise and intensity-wise features. These two types of features are subsequently fed into a semantics-guided progressive registration framework for accurate deformation field estimation. We further propose an anatomical region deformation consistency learning to regularize the target anatomical regions deformation. Comprehensive experiments conducted on three typical yet challenging datasets demonstrate that our method achieves consistently better results than other state-of-the-art deformable registration approaches. The source code is publicly available at https://github.com/xiehousheng/BIASNet.

Deformation of flexible robots can be practically assessed using extension/compression, shear, curvature, and torsion. Sensing based on one or more of the above characteristics enables closed-loop control for delicate tasks that require precision and dexterity. Due to the increasing popularity of flexible robotics in recent years, significant research effort has been directed to this burgeoning field. Although numerous studies have addressed soft sensing technologies, their successful integration into flexible robotic systems remains limited. This article provides a comprehensive review of sensing methods, from multidimensional deformation to the underlying principles of deriving hard-to-measure deformation from surrogate parameters. It focuses on sensing modalities such as strain measurement via piezoelectric, capacitive, resistive, and optical techniques. The applications of deformation sensing in industrial and service robotics are described. Future challenges and potential research issues including resolution, conformability, multifunctionality, crosstalk, and miniaturization are discussed. The need for a synergistic approach across disciplines is highlighted, emphasizing the integration of new materials, microstructures, advanced manufacturing technologies, and state-of-the-art signal processing techniques.

Intraoperative and spatiotemporal monitoring of neuroinflammatory indices during brain-computer interface (BCI) implantation is essential for ensuring safety and efficacy of the procedure. Current biomolecular detection approaches are unable to obtain spatiotemporally resolved inflammatory profiling, which is important for guiding the placement of microelectrodes intraoperatively. This study presents an intraoperative spatiotemporal acute inflammation detector (ISAID) that harnesses droplet-based sampling and multiplexed titanium oxynitride (TiNO) plasmonic biosensing to assess local inflammation during the insertion of intracortical microelectrodes. Through freestanding sampling droplets and fine-tuned TiNO-based biosensors, the ISAID achieved precise, sensitive, and integrated sampling to biosensing for cytokine detection with a spatial resolution down to 610 μm and a fast equivalent bioassay time of 1.25 min. The proposed system also allows multiple ISAID biosensing modes, enabling both spatial inflammation mapping and multi-cytokine analysis. Quantitative analyses of inflammatory cytokines with in vivo mouse models demonstrate the accuracy and practical advantages of the system.

Precision-controlled microscale manipulation tasks—including neural probe implantation, ophthalmic surgery, and cell membrane puncture— often involve minimally invasive membrane penetration techniques with real-time force feedback to minimize tissue trauma. This imposes rigorous design requirements on the corresponding miniaturised instruments with robotic assistance. This paper proposes an integrated piezoelectric module (IPEM) that combines high-frequency vibration-assisted penetration with real-time in-situ force sensing. The IPEM features a compact piezoelectric actuator integrated with a central tungsten probe, generating axial micro-vibration (4652 Hz) to enable smooth tissue penetration while simultaneously measuring contact and penetration forces via the piezoelectric effect. Extensive experiments were conducted to validate the effectiveness and efficacy of the pro-posed IPEM. Both static and dynamic force sensing tests demonstrate the linearity, sensitivity(9.3 mV/mN), and accuracy (MAE < 0.3mN, MAPE < 1%) of the embedded sensing unit. In gelatin phantom tests, the module reduced puncture and insertion forces upon activation of vibration. In vivo experiments in mouse brains further confirmed that the system could reduce penetration resistance (from an average of 11.67mN without vibration to 7.8mN with vibration, decreased by 33%) through the pia mater and accurately mimic the electrode implantation-detachment sequence, leaving a flexible electrode embedded with minimal trauma. This work establishes a new paradigm for smart surgical instruments by integrating a com-pact actuator-sensor design with real-time in-situ force feedback capabilities, with immediate applications in brain-machine interfaces and microsurgical robotics.