Instrument localization is widely used in computer-assisted surgery. Most existing navigation systems monitor the position of the surgical instrument's stylus-tip in non-visible surgical areas. However, a gap still exists in meeting the surgeons' requirements for instrument monitoring. It is difficult for them to monitor the instrument's working direction and its contact with the lesion area without an intuitive silhouette or additional feature points on the instrument. This paper introduces a full-tool surgical navigation for enhancing instrument localization and visualization in surgery, providing silhouette monitoring and directional indications. The key to overcoming these challenges lies in reconstructing the surgical instrument, as well as calibrating the stylus-tip and direction of the key part on the instrument. Our method achieves this through a silhouette carving reconstruction method with a feature-dense position-sensing marker. In the experiment, the fiducial registration error of our method reaches 0.496 mm, and the relative orientation error is 1.401 degrees. These localization and orientation accuracies demonstrate the potential of full-tool navigation system in surgeries.

Pattern recognition models trained on low-density surface electromyography (sEMG) sensors are susceptible to signal quality degradation and source variability. This study addresses the critical challenge of reduced gesture recognition accuracy in armband-based sEMG systems caused by concurrent interference of electrode shift and damage. We propose a hybrid approach integrating a convolutional neural network (CNN), a squeeze-and-excitation (SE) attention block, and transfer learning (TL). Data from seven hand gestures performed by nine subjects under electrode shift/damage were analyzed. The SE-CNN TL model achieved accuracies of 96.32 +/- 1.29% (shift only), 94.98 +/- 3.82% (damage only), and 94.30 +/- 1.51% (concurrent interference)-significantly outperforming conventional and deep learning benchmarks. Notably, the accuracy under concurrent interference represents the highest level reported to date. This method demonstrates universality against diverse interferences and establishes a new state-of-the-art for concurrent interference mitigation in low-density sEMG systems. Our framework provides a generalized solution for robustness enhancement in sEMG-based pattern recognition.

Objective: coronary rotational atherectomy (CRA) is a vital method for the treatment of cardiovascular calcified blockages, but the lack of intravascular information may lead to improper setting of rotary parameters, posing risks of surgical complications. Forward-looking intravascular ultrasound (FL-IVUS) is a valuable tool for lumen imaging and has unique advantages in severely calcified or even occluded vessels. Methods: in this study, a visual rotary surgical instrument based on FL-IVUS imaging was proposed to achieve intraoperative feedback. Imaging catheters with forward viewing angles of $30^{\circ }$, $45^{\circ }$, and $60^{\circ }$ were designed and manufactured to investigate the influence of angle parameters of the transducer on image quality. The imaging performance of the FL-IVUS image feedback system was evaluated by wire phantoms, ex vivo porcine aorta, vascular phantom model and calcified tissue substitute. Results: the results showed that the transducers with different tilt angles had similar performance and resolution, and forward view ultrasound imaging demonstrated satisfactory performance for imaging ex vivo porcine aorta tissue. By increasing the angle between the beam and the axial direction, the spatial resolution of captured images improved, while the penetration depth decreased. Conclusion: the integrated forward-looking ultrasound rotary cutting devices have good ultrasound imaging performance and can become a useful tool for intracavitary information feedback, which is helpful to the parameter setting and evaluation of rotary grinding. Significance: miniaturized and integrated forward-looking ultrasound rotatory device can facilitate real-time ultrasound interventional feedback and reduce the risk of surgical complications.

Event sensors can capture fast-moving objects without motion blur, but their limited spatial resolution constrains tracking accuracy. To address this issue, we propose a method for achieving precise 6-DOF object tracking using lensless event sensors. A mask printed with marker fields replaces traditional lenses, and a component composed of multiple rigid beacons is used to mark the object. Data deconvolution and matching optimization algorithms are developed to calculate the positions of multiple beacons and estimate the object's pose, employing a dual-filter to smooth the tracking trajectory. Experimental results demonstrate that, compared with state-of-the-art methods, our method reduces positional error by 19.6%, from 1.68 cm to 1.35 cm, while achieving rotation accuracy, 2.94 degrees versus 3.08 degrees, at a tracking frequency of 500 Hz. Additionally, it maintains low computational cost with an average processing time of 1.71 ms per frame, thereby satisfying the requirements of real-time applications.

This paper proposes a novel visibility-guaranteed tracking control framework to address the dual objective of accurate target positioning and persistent endoscopic visual field maintenance. The method employs moment-based visual servoing for nominal target tracking. Based on input-to-state safety, a robust Control Barrier Function (RCBF) is defined to ensure targets remain within a visible set, whose boundary size is adjusted according to disturbances. The visibility constraint, along with the robot’s physical constraints, is enforced within a Quadratic Programming (QP)-based controller, explicitly designed to guarantee visibility maintenance while minimizing tracking error. Neurodynamic optimization is used to enable real-time solution of the constrained QP problem. Lyapunov-based stability analysis formalizes convergence guarantees. Simulation and experimental evaluations under disturbances demonstrate target visibility within the endoscopic field of view and accurately track targets. Comparative results confirm improved tracking accuracy and visibility over state-of-the-art methods.

This article proposes a novel 3-D Moir & eacute;-based visual marker with an explicit geometric model. The proposed marker is designed for out-of-plane rotation measurement, which comprises two periodic masks etched on the opposite sides of a glass wafer. The masks project a Moire pattern on the image plane of the observing camera, and this Moir & eacute; pattern is dramatically sensitive to out-of-plane rotation. The key contribution is that we first explicitly derive, for the first time, the geometric relationship between the rotation angle and the Moire pattern to build a straightforward measurement model, which reveals that the angle is simply determined by the Moire phase at the image principal point. The accuracy is independent of the observation distance and camera intrinsic parameters. Estimation and calibration algorithms are given. Experiment demonstrates the superiority, which shows an accuracy that is up to 7 times higher than traditional visual markers and at least 2 times higher than state-of-the-art 3-D Moire-based markers.

Rotational atherectomy (RA) is an effective minimally invasive treatment for severe coronary calcification, but it carries risks of vessel injury due to high mechanical and thermal loads. This study proposes a low-frequency vibration-assisted rotational atherectomy (LFVARA) technique that reduces cutting force and temperature rise by promoting cyclic disengagement between the tool and tissues. Experimental trials compared LFVARA against RA using bone cement as a calcified plaque phantom, with vibration parameters systematically varied across amplitudes (20–50 μm), frequencies (50–150Hz), and rotational speeds (60000–120000r/min). Response surface methodology based on a Box-Behnken design is employed to analyze parameter interactions and identify optimal conditions. Results show that increasing vibration frequency from 50 to 150Hz at 50 μm amplitude prolongs non-cutting time per cycle by 19.47%, while raising rotational speed from 60000 to 120000r/min at 50 μm/150Hz reduces it by 26.59%. Under 120000r/min/150Hz/50 μm conditions, LFVARA mitigates mean cutting force by 23.85% and temperature rise by 29.24%, as validated by force-thermal metrics. The optimized parameter set (60000r/min, 141.6Hz, 49 μm) is experimentally validated with less than 10% error in force and 5% in temperature prediction. The findings demonstrate that LFVARA can effectively reduce the average cutting force and temperature rise during the process of removing calcified tissues. This work presents a promising approach for coronary atherosclerosis treatment, offering a novel method to optimize therapeutic efficacy.

Background Epilepsy affects millions globally and is characterized by recurrent seizures from aberrant neuronal discharges. Reliable seizure classification is challenging due to the nonlinear complexity and high variability of electroencephalogram (EEG) signals. Objective To improve the accuracy of epilepsy classification using EEG signals, this study introduces a novel complex network framework—Multi-Span Transition Network with Riemannian Similarity Measure (RSMMTN). Methods The method synergizes three innovations: (1) A parameter-free second-order difference scatterplot to amplify transient signal features using zero-crossing dynamics; (2) Symbolic ordinal partition transition networks (OPTNs) to map scatterplot geometries into multi-span topological representations; (3) Affine-invariant Riemannian metrics to quantify kinetic resemblance through manifold-embedded adjacency matrices. Results Validated on synthetic and clinical datasets, RSMMTN achieves high accuracy in differentiating nonlinear time series, like ictal, preictal, and interictal states of epilepsy data. It outperforms existing methods, particularly excelling in the challenging task of preictal-interictal discrimination. Conclusions The RSMMTN framework is a computationally efficient and noise-robust method for epilepsy classification. Its high performance demonstrates its potential for clinical applications. This research provides a powerful new methodology for developing high-precision automated epilepsy diagnostic tools and potential real-time seizure prediction systems.

A visual-based point localization model achieving super-resolution in measurement is introduced, termed 3-D Moiré projection model. It leverages the 3-D Moiré effect generated by the shadow of the tandem masks fixed in front of a bare camera sensor to produce a highly sensitive response to the location of a target beacon. Using a new approach called casting map, the geometry relationship between the beacon and the Moiré pattern is derived. It is shown that the 3-D Moiré projection model can be interpreted as a pin-hole model with periodically extended scopes, and the frequencies of the mask patterns determine the analogical focal length. The calibration and localization algorithms are presented, and proof-of-concept prototypes are made. In experiments, the prototypes achieved a resolution of over 0.3 billion points using only 7 mm of thickness, which is more than 40 times that of pin-hole based counterparts. They also show potential and examples to surpass lens-based cameras, which are highly mature implementations of the pin-hole based model, despite having under one-tenth the thickness, lower cost, underdeveloped hardware and algorithms.

Traditional hypertonia diagnosis relies on the Modified Ashworth Scale (MAS), which is subjective and dependent on doctors’ experience. Although previous studies have explored the use of force sensors and surface electromyography (sEMG), finding a reliable and valid detection method remains a challenge. This study aims to develop a simple yet effective platform that integrates biomechanical and sEMG data for upper-limb muscle tone assessment, providing a more objective and quantitative evaluation approach. A detection platform was developed to collect biomechanical and sEMG data from 59 subjects, including 49 patients (MAS Ⅰ = 21, MAS Ⅰ + = 16, MAS Ⅱ = 12) and 10 healthy individuals, at different movement speeds (15°/s, 20°/s, and 25°/s). The acquired data underwent feature extraction, including signal processing and statistical analysis. Dimensionality reduction was applied to optimize the extracted features, and these features were then integrated into a classification algorithm for further analysis. The extracted features effectively distinguished patients from healthy individuals, with statistically significant differences (p < 0.01). Furthermore, the strong correlation between the extracted features and MAS scores (p < 0.01) confirmed the reliability of the proposed method. Finally, the classification algorithm demonstrated high consistency with clinical evaluations, validating its potential for clinical application in muscle tone assessment. This study introduces an objective and quantitative method for assessing muscle tone, shifting away from the traditional subjective MAS evaluation. By enhancing diagnostic accuracy, the proposed approach provides a more reliable basis for hypertonia diagnosis and treatment. The findings hold significant promise for optimizing clinical decision-making, ultimately improving patient management and therapeutic strategies.

Two-dimensional forward-looking sonar (FLS) images are often severely degraded by speckle noise, impairing human interpretation and computer vision tasks. Existing denoising methods, including traditional multiplicative noise models and deep learning approaches, suffer from over-suppression of high-frequency information, leading to residual artifacts, incomplete noise removal, and blurred edges. To address these challenges, this letter proposes SonoLF-ASPPNet, a novel sonar image denoising network, along with a new simulated dataset construction method. The network incorporates three key contributions: the low-frequency structure-edge focusing module combines the traditional speckle reducing anisotropic diffusion (SRAD) with a Laplacian kernel to enable simultaneous low-frequency denoising and high-frequency edge preservation; the low-frequency attention module incorporates an atrous spatial pyramid pooling module to enhance contextual awareness and suppress high-frequency interference. The training utilizes a composite loss function to enhance denoising in smooth areas, suppress color spots, and homogenize backgrounds. The experimental results indicate that on the simulated dataset, sono low-frequency atrous spatial pyramid pooling network (SonoLF-ASPPNet) achieved peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) values of 27.76 dB and 0.817, respectively, along with excellent visual performance. On the real FLS dataset, the model outperformed comparative algorithms in terms of visual outcomes, demonstrating superior noise suppression and texture preservation. Moreover, it attained competitive quantitative scores in equivalent number of looks (ENL), edge preservation index (EPI), and M-score.

Accurate needle placement is critical in percutaneous thoracoabdominal interventions. Conventional CT-guided procedures are limited by iterative scans, operator-dependent variability, and two-dimensional visualization. This preclinical study evaluates a novel display-based augmented reality (AR) navigation system integrating virtual-to-real registration via deltille grid position sensing (DGPS) markers for enhanced needle guidance. Experiments were conducted on a CIRS abdominal phantom with three operators: one expert interventional radiologist and two novices. Puncture accuracy, angular deviation, procedure time, and number of verification scans were assessed. AR guidance achieved first-pass distance errors of 1.50 (0.46) mm for the expert and 1.66 (0.50) mm and 1.58 (1.06) mm for novices, with angular deviations of 1.41°, 1.30°, and 1.39°, respectively. No significant operator-related differences were observed. Compared with conventional step-and-shoot CT guidance, AR guidance reduced the first-pass and final-pass targeting error, while conventional CT guidance relies on iterative scan-adjust-advance steps to achieve acceptable final accuracy, with AR reducing intermediate re-scans while retaining a single final verification scan. The findings indicate that AR navigation enhances first-pass precision, reduces procedural variability, and improves workflow efficiency, supporting its potential clinical value for more consistent and efficient minimally invasive interventions.

Visual localization is essential in many vision-driven interaction domains, including automation, AR (augmented reality), and surgical navigation. This paper presents a visual marker specifically designed for deformable surface tracking. It has three advantages: 1) IFPs (inner false positives) detection is avoided through grayscale integration design, leading to more robust results. 2) Without relying on thresholds, our detection algorithm enhances the method’s reliability in deformable surface with complex shading. 3) We use a position-sensing marker design with higher information density than the self-identifing marker to ensure the supply of features. In the experiments, our marker achieved superior localization accuracy and zero IFPs. Additionally, we present two compelling case studies showcasing the marker’s practical applications in augmented reality and surgical instrument tracking. Our work offers a significant advancement in visual localization, especially in challenging scenarios involving deformable surfaces, providing valuable solutions for researchers and developers across various application domains.

Background and objective: In biopsy puncture procedures, successfully avoiding the randomly distributed subcutaneous arteries to reduce the incidence of postoperative complications and shorten patient recovery times remains a pressing challenge. To address this issue and achieve precise classification of arteries and their surrounding viscoelastic tissues, this study proposes a Deep Entropy-Weighted Fusion Network (DEHF-Net), which is utilized to classify a dataset comprising 2,200 ultrasound echo samples collected in the laboratory. Through comprehensive performance comparisons and subsequent model optimization, an automated identification system capable of accurately recognizing the structures of arteries and viscoelastic tissues was developed. Methods: In this study, a structurally optimized biopsy needle integrated with an ultrasonic transducer operating at a center frequency of 40 MHz was developed to acquire ultrasound echo signals from arteries and their surrounding viscoelastic tissues. A convolutional neural network (CNN) was combined with a quadratic support vector machine (Q-SVM) to construct the DEHF-Net framework. Subsequently, targeted optimizations were implemented to address identified limitations, resulting in a more robust and comprehensive DEHF-Net-based automatic recognition system. Results: The results indicate that the DEHF-Net automatic recognition system outperforms other fusion models across multiple performance metrics. After further optimization, all performance metrics exceeded 94%, achieving precise identification of arterial vessels and viscoelastic tissues. Conclusions: The DEHF-Net automatic recognition system developed in this study is capable of precisely classifying the ultrasonic echo signals from arterial vessels and the surrounding viscoelastic tissues. This provides a novel approach and a reliable technical foundation for ultrasound-assisted puncture techniques without imaging.

Instrument tracking serves as the backbone of sur gical navigation systems, enabling precise real-time localization of surgical tools. However, most visual tracking systems adopt f ixed camera configurations that passively observe the surgical instrument. Such static architectures have limited measurement volumes and cannot fully accommodate complex instrument trajectories during surgery, leading to tracking failures. To avoid this situation, we propose a multi-view active visual tracking system to maintain consistent tracking of moving surgical in struments. Multiple cameras observing from different positions localize the instrument by constructing multi-view multi-point perspective equations. Within this positioning framework, an online viewpoint planning algorithm is developed to actively adjust non-ideal viewpoints by modeling suboptimal-view condi tions and formulating corresponding inequality constraints. The system employs a visual servoing based camera control mecha nism to execute viewpoint adjustments in real time. Following each adjustment, the extrinsic parameters for the camera are recalibrated using an online multi-frame extrinsic calibration method, ensuring consistent spatial accuracy. Both simulation and real-world experiments are conducted, and the proposed system achieves excellent localization performance and exhibits superior robustness against viewpoint limitations compared with f ixed-camera systems.

BACKGROUND AND OBJECTIVE:Percutaneous renal biopsy is an essential diagnostic procedure for renal diseases. However, puncture trajectory planning still relies heavily on the operator's experience, and standardized quantitative planning methods remain lacking. To improve procedural safety, reproducibility, and standardization, this study proposes a hybrid sampling-based RRT* algorithm (HS-RRT*) based on reverse path planning for efficient and safe trajectory planning in different renal puncture regions. METHODS:An ex vivo renal puncture phantom was developed using three-dimensional (3D) printing, and the puncture environment was reconstructed from 3D medical images. Building upon the conventional reverse RRT algorithm, HS-RRT* incorporates a sampling-channel strategy, a sampling-angle evaluation mechanism, multi-path selection, and Bézier curve-based trajectory smoothing. The proposed method was systematically compared with representative path-planning algorithms under identical spatial configurations. RESULTS:Experimental results demonstrate that HS-RRT* outperforms the compared algorithms in terms of path length, search efficiency, and trajectory smoothness. The generated puncture trajectories satisfy clinical safety requirements while effectively reducing the randomness associated with single-path planning. Moreover, the optimized trajectories exhibit improved smoothness, with the maximum bending angle maintained below 4° CONCLUSIONS: HS-RRT* provides a precise, efficient, and standardized solution for preoperative trajectory planning in percutaneous renal biopsy. By integrating reverse path planning, hybrid sampling, and trajectory optimization, the proposed framework provides a reliable technical basis for intelligent puncture navigation and standardized clinical procedures.

Stable three-dimensional hand landmark reconstruction using low-cost RGB-D sensors is important for human–computer interaction, robot teleoperation, and vision-based motion analysis. RGB-based hand landmark detectors provide stable semantic 2D landmarks, but their depth output is not a metric measurement in the physical camera coordinate system. Stereo cameras can provide metric depth, but direct landmark-level back-projection is sensitive to invalid pixels, local depth holes, boundary noise, and partial occlusion. To address these problems, this paper presents a lightweight RGB-D sensing front-end that combines MediaPipe semantic hand landmarks with ZED2 stereo depth. The proposed pipeline detects 21 semantic hand landmarks in the RGB image, obtains landmark-level metric depth from the aligned ZED2 depth map using local median sampling, reconstructs 3D landmarks by camera back-projection, and further applies exponential moving average filtering and a bone-length consistency constraint. Experiments were conducted on a self-collected SVO dataset containing 13 hand actions and 26 recorded sequences, and an additional checkerboard-based reference-distance validation was performed to evaluate the metric depth sampling and 3D back-projection component. Compared with single-pixel sampling, the 5×5 local median strategy slightly increased the valid-depth ratio from 0.9731 to 0.9738 and reduced the temporal smoothness metric from 1.7163 mm to 1.6902 mm. To further justify the temporal filtering choice, an additional comparison with the 1 Euro Filter was conducted using the reconstructed win5 trajectories. The 1 Euro Filter produced stronger smoothing, reducing the temporal smoothness metric to 0.196 mm, but also reduced the path-length ratio to 0.484, indicating substantial motion attenuation. EMA0.7 was therefore retained as a more balanced setting, reducing the temporal smoothness metric to 0.826 mm while maintaining a path-length ratio of 0.803. The BL0.5 bone-length constraint reduced the bone-length standard deviation from 2.0727 mm to 1.1995 mm with limited trajectory modification. The final configuration provides a practical low-cost RGB-D front-end for stable 3D hand landmark reconstruction under controlled indoor conditions.

Edentulous implant surgery presents significant educational challenges due to its procedural complexity and the limitations of traditional teaching methods. Conventional approaches often lack sufficient hands-on practice and real-time feedback, which hinders students’ skill mastery. The objective is to address these limitations by developing an innovative virtual reality (VR)-based teaching system. This system is designed to provide an immersive training environment to enhance students' procedural understanding, practical skills, and overall surgical competency. We developed a VR-based teaching system that integrates advanced modeling with interactive technologies. The system provides real-time, dynamic feedback on critical surgical parameters, allowing students to practice and improve their skills in an immersive environment. Preliminary assessments demonstrated a marked improvement in students’ operational proficiency and surgical competency. The system successfully bridged the gap between theoretical knowledge and practical execution. This VR-powered system offers a novel and effective solution for dental surgical education. It has the potential to elevate learning outcomes and better equip students with the expertise and confidence needed to navigate the complexities of clinical practice.

Marker cubes are widely used for pose measurement, but their performance degrades under singular viewpoints and partial occlusion. This paper presents an interactive sphere with a seamless self-identifying graph marker field. By extending the conventional per-face marker scheme to a continuous spherical surface, the proposed design enables pose measurement from arbitrary viewing angles through locally identifiable features. A bottom-up pipeline is developed for marker decoding and pose measurement. Experiments show that, compared with cube-and icosahedron-based carriers, the proposed sphere achieves stronger viewpoint robustness and more stable pose measurement. These results demonstrate the effectiveness of the proposed method for robust pose measurement in interactive scenarios. The demo and source code are available at https://github.com/ Lilin2015/Interactive-Sphere.

Benign Paroxysmal Positional Vertigo (BPPV) is a prevalent vestibular disorder encountered in clinical settings. Diagnosis of this condition primarily relies on the observation of nystagmus, which involves monitoring the eye movements of patients. However, existing medical equipment for collecting and analyzing nystagmus data has notable limitations and deficiencies. To address this challenge, a comprehensive BPPV nystagmus data collection and intelligent analysis framework has been developed. Our framework leverages a neural network model, Egeunet, in conjunction with mathematical statistical techniques like Fast Fourier Transform (FFT), enabling precise segmentation of eye structures and accurate analysis of eye movement data. Furthermore, an eye movement analysis method has been introduced, designed to enhance clinical decision-making, resulting in more intuitive and clear analysis outcomes. Benefiting from the high sensitivity of our eye movement capture and its robustness in the face of environmental conditions and noise, our BPPV nystagmus data collection and intelligent analysis framework has demonstrated outstanding performance in BPPV detection.