154092 research outputs found
Sort by
Endovascular robotics:technical advances and future directions
Get PDFEndovascular interventions excel in treating cardiovascular diseases in a minimally invasive manner, showing improved outcomes over open techniques. However, challenges related to precise navigation–still relying on 2D fluoroscopy–persist. This review examines the role of robotics, highlighting commercial and research platforms, while exploring emerging trends like MRI compatibility, enhanced navigation, and autonomy. MRI-compatible systems offer radiation-free 3D imaging. Human-robot interaction evolves with task-specific interfaces, while autonomy ranges from partial to full, aiding clinical operators. Challenges include complexity and cost, emphasizing compatibility and navigation advancements. Integrating MRI-compatible robots, refining human-robot interaction, and enhancing autonomy promise advancements in endovascular surgery, fueled by AI and innovative imaging.</p
Deep learning-based phase calibration of airborne SAR Tomography
No full textSynthetic Aperture Radar (SAR) tomography (TomoSAR) is a proven technique for capturing multi-dimensional insights into the Earth&#039;s surface. Its operation relies on the principle that the phase of the backscattered signal reflects the sensor-to-target distance. However, in airborne SAR applications, residual platform motion often introduces phase errors, leading to challenges like increased side-lobes or signal blurring, which hinder information retrieval [1]. In forested or snow-covered regions, additional decorrelation sources, such as temporal, geometric, and volumetric factors, further degrade signal-to-noise ratios (SNR). This makes phase calibration a critical step to mitigate these disruptions and ensure data accuracy. Numerous methods have been developed for phase calibration in tomographic SAR data. Techniques like ALGAE [2] use algebraic approaches to isolate ground contributions but struggle in dense forests. Autofocusing methods [1, 3], based on optimizing entropy or contrast in tomograms, are adaptable but computationally demanding with large datasets. Techniques like [4,5] consider baseline error models for motion errors to estimate phase error while requiring phase triangularity assumption [4] to mitigate non-coherent scatterer effects. Nowadays, with the growing advancements in deep learning techniques and convolutional neural networks (CNNs) for remote sensing applications, we propose a novel CNN-based method for calibrating TomoSAR datasets. A key advantage of the proposed approach is its independence from coherent points, making it applicable to a broad range of environments, including forested and snowy regions. Moreover, this method operates within an unsupervised framework, eliminating the need for reference training data. The model is trained directly on the input TomoSAR data to predict phase errors using a carefully designed loss function that enforces a specific phase error model on the network&#039;s output. A critical aspect of this framework is its ability to bypass the need for generating training datasets comprising both distorted and clean TomoSAR data. In real-world scenarios, clean TomoSAR data are inaccessible, making traditional CNN based methods reliant on such data impractical or external source of information to generate reference data. Instead, the proposed approach does not depend on external reference datasets or additional information, addressing concerns regarding the generalizability of deep learning models for phase calibration. Indeed, while state-of-the-art deep learning methods are typically trained on extensive datasets, their applicability can be constrained by discrepancies between training data and the actual application scenarios, such as differences in study areas, baseline distributions, looking angles, and frequencies. In contrast, the proposed method offers a universally applicable framework for phase calibration across diverse TomoSAR datasets, irrespective of image characteristics, acquisition geometries, or study regions. The methodology is akin to unsupervised classification approaches, where the data itself is used to train a model capable of analyzing and calibrating the entire dataset. The proposed method was applied to the AfriSAR [6] dataset obtained from SETHI and compared against standard calibration techniques. Its performance was assessed through tomogram reconstruction along various range and azimuth profiles, with ground and canopy height models derived from LiDAR data overlaid onto the tomograms for validation. The results demonstrate the efficacy of the approach, underscoring its potential as a robust and user-friendly tool for phase calibration in TomoSAR data. Once trained on a universal set of tomographic images, the method can be readily deployed to streamline calibration processes. In particular, generating universal datasets from publicly available airborne SAR campaigns such as TropiSAR, UAVSAR, and others, and training the network on these datasets, makes it feasible to develop a ready-to-use tool for calibration applications. Experimental results, quality control measures, and evaluations will be presented at the conference
Mechanical Constraint Identification in Model-Mediated Teleoperation
No full textModel-Mediated Teleoperation (MMT) is a form of teleoperation where a model is used to display operator commands to the environment and environmental feedback to the operator. This circumvents the performance-stability trade-off to which many other teleoperation methods are prone under high time delays. An aspect of MMT which has received little attention is the modelling of rigid-body constraints, which are important for many manipulation tasks. In this paper, we detail a description of rigid body constraints and highlight a specific type of contact: the Lower Pair, of which the contacts found in hinged doors and household drawers, for instance, are examples. We also present a method for estimating these Lower Pairs and test it in a real world scenario. Results show that it is robust to measurement noise, as well as small amounts of movement in the constrained directions.</p
Counterexample-Guided Abstraction Refinement for Generalized Graph Transformation Systems
No full textThis paper addresses the following verification task: Given a graph transformation system and a class of initial graphs, can we guarantee (non-)reachability of a given other class of graphs that characterizes bad or erroneous states? Both initial and bad states are characterized by nested conditions (having first-order expressive power). Such systems typically have an infinite state space, causing the problem to be undecidable. We use abstract interpretation to obtain a finite approximation of that state space, and employ counter-example guided abstraction refinement to iteratively obtain suitable predicates for automated verification. Although our primary application is the analysis of graph transformation systems, we state our result in the general setting of reactive systems.</p
Advanced serial analysis of the diaphragm surface EMG:insights into the effect of pressure support on the neuro-ventilatory response during the ICU stay
Get PDFBackground: Ventilatory support levels in ICU patients should be tailored to both optimal gas exchange and respiratory muscle loading, as over- and underassistance may cause diaphragm dysfunction. The diaphragm’s capacity to overcome mechanical load and deliver ventilatory output is reflected by the patient’s neural respiratory drive (NRD), tidal volume (TV) and respiratory rate (RR). Surface electromyography of the diaphragm (sEMGdi) offers a continuous, non-invasive measure of NRD. We investigated the effect of pressure support (PS) level on the coupling of diaphragm electrical activity (sEAdi) and ventilatory output during the ICU stay. Methods: In clinically stable ICU patients (N = 17), four PS-levels were applied on alternate days, based on the clinical value (− 3, + 0, + 3, and + 6 cmH2O). sEAdi time-product (ETPdi) was calculated from high-quality sEAdi waveforms, using a novel, advanced signal analysis approach. The breath-by-breath correlation between ETPdi and TV was defined as neuro-ventilatory coupling (NVC), enabling quantification of the neuro-ventilatory response. Results: On group level (13 patients, 26 PS-trials), ETPdi and RR increased with decreasing PS-levels (2.4 and 1.6 percentage point (pp)/cmH2O), whereas TV decreased (2.5 pp/cmH2O). Longitudinal analysis (4 patients, 14 PS-trials) showed strengthened coupling between ETPdi and TV during weaning, reflected by an increase in median NVC from 3.4% (IQR 2.9) to 26.3% (IQR 21.7) between the first and last PS-trial. Conclusion: Advanced sEMGdi analysis allows for non-invasive quantification of NVC, reflecting the diaphragm’s capacity to overcome mechanical load. In patients approaching liberation from MV, increasing NVC indicates the shift from near-passive to active breathing. This study demonstrates the potential of NVC to inform tailoring of ventilatory support levels. Trial registration number: Dutch Trial Register NL9654. Registered August 05, 2021.</p
Goo-y:A Robot’s Shape-Changing Capability for Fast, Non-Anthropomorphic Communication with People (and Animals)
No full textEffective interaction with robots relies heavily on their ability to convey their purpose and intentions. While people are familiar with communication methods like (human) speech, this may lead to unnecessary anthropomorphisation or deception. On the other hand, robot-specific signals such as movements or colours are often difficult to interpret. Drawing inspiration from interspecies communication and animation principles offers a broader design space for creating expressive, easily understood signals through shape-changing capabilities. We detail the development of a shape-changing capability for a robot to support non-anthropomorphic, visceral communication. Starting with a virtual reality study to gauge people’s responses to various shape-changing behaviours, followed by developing a full-scale version suitable for standard robotic platforms and in-situ experimentation. This low-cost, easy-to-build capability opens new possibilities for researchers in human-robot interaction by making shape-changing technology accessible to the CHI research community.</p
Age Against the Machine:How Age Relates to Listeners' Ability to Recognize Emotions in Robots' Semantic-Free Utterances
No full textSemantic-Free Utterances (SFUs, sounds conveying intention without using words) are being increasingly adopted for human-robot interaction (HRI) to communicate affect. In healthcare, where older adults are overrepresented, affective robotics are becoming more common to reduce healthcare professionals' workload. Hence, understanding how older adults perceive and communicate with robots is crucial. Although previous studies have demonstrated a decline in older adults' ability to categorize emotions, it remains unclear how this impacts their comprehension of SFUs used in HRI. This paper investigates the effect of age (and other factors) on listeners' ability to categorize emotions in SFUs designed for HRI. Additionally, we explore listeners' preferences of SFUs for a healthcare robot. Listeners indicated that SFUs' similarities to natural language, the need for a distinction between human and robot, and their expectations of how a hospital robot should sound like, influenced their preferences. Furthermore, we conducted an online emotion categorization task to investigate how age, emotion category, type of SFU (with varying degrees of robot-likeness), listeners' gender, and their experience with robots relate to listeners' ability to categorize emotions. Results confirm that as age increases, there is a decline in emotion categorization performance of SFUs varying by emotion category and type of SFU.</p
Ask and You Shall Find:How Suggestions by a Conversational Robot Assist Children with Information Search
No full textWith the rapid advancement of AI, we anticipate that children will increasingly turn to conversational search systems instead of traditional search engines. We developed a speech-based conversational robot and tested its effectiveness in helping children formulate queries through suggestions. The study took place in a museum with 48 children. Two ways of providing suggestions were compared. No significant effect was found on children’s suggestion usage, nor on their self-reported attitudes. Additionally, we explored two novel research instruments to measure children’s attitude toward the robot’s actions. We found that children provided significantly more feedback, and more positive feedback, with one of the instruments. Our contributions to developing conversational search technology for children include advancements in conversational search design, the exploration of appropriate research instruments, and suggestions for future work in this field.</p
Perception and Practice of Data Management Plans in Health:An Exploratory Study
Get PDFThis paper discusses the quality of Data Management Plans (DMPs) in the health sector and assesses remove the researchers’ perceptions of DMPs. We applied qualitative methods to examine publicly available DMPs in healthcare, analyzing researchers’ views and practices for creating these plans. The study combines three research methods: analysis of DMPs in the health sector, semi-structured questionnaires, and interviews. Our findings reveal that researchers are generally unaware of the importance and usefulness of DMPs, and acknowledge various inconsistencies and challenges in their development. In this paper, we identified that data management practices need to be improved and advocate for automating them and making DMPs machine-actionable. We also recommend more educational programs, such as workshops and courses, in data management especially for researchers. Finally, we recommend defining clear, accessible guidelines for researchers to effectively elaborate DMPs, and institutionalizing data management within organizations by establishing data (or digital) competence centers.</p
Dual-Piston Hoop Gear Driven MR Safe Pneumatic Stepper Motor
No full textPneumatic stepper motors are useful in actuating magnetic resonance (MR) safe robotics. A novel design is presented, which consists of two pistons in cross-configuration, one hoop gear and one rotor composed of a cycloid gear on a shaft. Realizations are presented in two different sizes, both employing 13 teeth and a step size of 6.9∘. The larger motor, RC-80, delivers 6.3 N m torque and 24 W output power. The smaller motor, RC-45, delivers 0.47 N m and 3.0 W. Both motors can continuously operate at approximately 50% of the maximum load for 60 min without breaking down. The presented motors provide highest torque among all state-of-art MR safe pneumatic stepper motors of comparable size, and are significantly more robust. Especially RC-45 is suitable for actuating the joints of compact MR safe surgical robots.</p