Search

Human movement intention recognition is important for human-robot interaction. Existing work based on motor imagery electroencephalogram (EEG) provides a non-invasive and portable solution for intention detection. However, the data-driven methods may suffer from the limited scale and diversity of the training datasets, which result in poor generalization performance on new test subjects. It is practically difficult to directly aggregate data from multiple datasets for training, since they often employ different channels and collected data suffers from significant domain shifts caused by different devices, experiment setup, etc. On the other hand, the inter-subject heterogeneity is also substantial due to individual differences in EEG representations. In this work, we developed two networks to learn from both the shared and the complete channels across datasets, handling inter-subject and inter-dataset heterogeneity respectively. Based on both networks, we further developed an online knowledge co-distillation framework to collaboratively learn from both networks, achieving coherent performance boosts. Experimental results have shown that our proposed method can effectively aggregate knowledge from multiple datasets, demonstrating better generalization in the context of cross-subject validation.

Egocentric vision is an emerging topic, which has demonstrated great potential in assistive healthcare scenarios, ranging from human-centric behavior analysis to personal social assistance. Within this field, due to the heterogeneity of visual perception from first-person views, egocentric pose estimation is one of the most significant prerequisites for enabling various downstream applications. However, existing methods for egocentric pose estimation mainly focus on predicting the pose represented in the camera coordinates from a single image, which ignores the latent cues in the temporal domain and results in less accuracy. In this paper, we propose Ego+X, an egocentric vision based system for 3D canonical pose estimation and human-centric social interaction characterization. Our system is composed of two head-mounted egocentric cameras, where one is faced downwards and the other looks outwards. By leveraging the global context provided by visual SLAM, we first propose Ego-Glo for spatial-accurate and temporal-consistent egocentric 3D pose estimation in the canonical coordinate system. With the help of an egocentric camera looking outwards, we then propose Ego-Soc by extending Ego-Glo to various social interaction tasks, e.g., object detection and human-human interaction. Quantitative and qualitative experiments have been conducted to demonstrate the effectiveness of our proposed Ego+X.

The success of most advanced facial expression recognition works relies heavily on large-scale annotated datasets. However, it poses great challenges in acquiring clean and consistent annotations for facial expression datasets. On the other hand, self-supervised contrastive learning has gained great popularity due to its simple yet effective instance discrimination training strategy, which can potentially circumvent the annotation issue. Nevertheless, there remain inherent disadvantages of instance-level discrimination, which are even more challenging when faced with complicated facial representations. In this paper, we revisit the use of self-supervised contrastive learning and explore three core strategies to enforce expression-specific representations and to minimize the interference from other facial attributes, such as identity and face styling. Experimental results show that our proposed method outperforms the current state-of-the-art self-supervised learning methods, in terms of both categorical and dimensional facial expression recognition tasks. Our project page: https://claudiashu.github.io/SSLFER.

Vision-based 3D human pose estimation and shape reconstruction play important roles in robot-assisted healthcare monitoring and personal assistance. However, 3D data captured from a single viewpoint always encounter occlusions and exhibit substantial heterogeneity across different views, resulting in significant challenges for both tasks. Extensive approaches have been proposed to perform each task separately, but few of them present a unified solution. In this paper, we propose a novel network based on signed distance functions, namely PoseSDF, to simultaneously reconstruct 3D lower limb shape and estimate gait pose by two dedicated branches. To promote multi-task learning, several strategies are developed to ensure that these two branches leverage the same latent shape code while exchanging information between them. More importantly, an auxiliary RotNet is incorporated into the inference phase, overcoming the inherent limitations of implicit neural functions under cross-view scenarios. Experimental results demonstrate that our proposed PoseSDF can achieve both high-quality shape reconstruction and precise pose estimation, generalizing well on the data from novel views, gait patterns, as well as real-world.

With the increasing awareness of high-quality life, there is a growing need for health monitoring devices running robust algorithms in home environment. Health monitoring technologies enable real-time analysis of users’ health status, offering long-term healthcare support and reducing hospitalization time. The propose of this work is twofold, the software focuses on the analysis of gait, which is widely adopted for joint correction and assessing any lower limb, or spinal problem. On the hardware side, a novel marker-less gait analysis device using a low-cost RGB camera mounted on a mobile tele-robot is designed. As gait analysis with a single camera is much more challenging compared to previous works utilizing multi-cameras, a RGB-D camera or wearable sensors, we propose using vision-based human pose estimation approaches. More specifically, based on the out-put of state-of-the-art human pose estimation models, we devise measurements for four bespoke gait parameters: inversion/eversion, dorsiflexion/plantarflexion, ankle and foot progression angles. We thereby classify walking patterns into normal, supination, pronation and limp. We also illustrate how to run the proposed machine learning models in low-resource environments such as a single entry-level CPU. Experiments show that our single RGB camera method achieves competitive performance compared to multi-camera motion capture systems, at smaller hardware costs.

Robot-assisted endobronchial intervention requires accurate localisation based on both intra- and pre-operative data. Most existing methods achieve this by registering 2D videos with 3D CT models according to a defined similarity metric with local features. Instead, we formulate the bronchoscopic localisation as a learning-based global localisation using deep neural networks. The proposed network consists of two generative architectures and one auxiliary learning component. The cycle generative architecture bridges the domain variance between the real bronchoscopic videos and virtual views derived from pre-operative CT data so that the proposed approach can be trained through a large number of generated virtual images but deployed through real images. The auxiliary learning architecture leverages complementary relative pose regression to constrain the search space, ensuring consistent global pose predictions. Most importantly, the uncertainty of each global pose is obtained through variational inference by sampling within the learned underlying probability distribution. Detailed validation results demonstrate the localisation accuracy with reasonable uncertainty achieved and its potential clinical value.

Acquisition of bimanual motor skills, critical in several applications ranging from robotic teleoperations to surgery, is associated with a protracted learning curve. Brain connectivity based on functional Near Infrared Spectroscopy (fNIRS) data has shown promising results in distinguishing experts from novice surgeons. However, it is less well understood how expertise-related disparity in brain connectivity is modulated by dynamic temporal demands experienced during a surgical task. In this study, we use fNIRS to examine the interplay between frontal and motor brain regions in a cohort of surgical residents of varying expertise performing a laparoscopic surgical task under temporal demand. The results demonstrate that prefrontal-motor connectivity in senior residents is more resilient to time pressure. Furthermore, certain global characteristics of brain connectomes, such as the small-world index, may be used to detect the presence of an underlying stressor.

Nature is a great source of inspiration for the development of solutions for biomedical problems. We present a novel biosensor design utilizing two-photon polymerisation and graphene to fabricate an enhanced biosensing platform for the detection of motile bacteria. A cage comprising venous valve-inspired directional micro-structure is fabricated around graphene-based sensing electronics. The asymmetric 3D micro-structure promotes motile cells to swim from outside the cage towards the inner-most chamber, resulting in concentrated bacteria surrounding the central sensing region, thus enhancing the sensing signal. The concentrating effect is proved across a range of cell cultures - from 101 CFU/ml to 109 CFU/ml. Fluorescence analysis shows a 3.38–3.5 times enhanced signal. pH sensor presents a 2.14–3.08 times enhancement via the detection of cellar metabolite. Electrical measurements demonstrate an 8.8–26.7 times enhanced current. The proposed platform provides a new way of leveraging bio-inspired 3D printing and 2D materials for the development of sensing devices for biomedical applications.

The current standard of intra-operative navigation during Fenestrated Endovascular Aortic Repair (FEVAR) calls for need of 3D alignments between inserted devices and aortic branches. The navigation commonly via 2D fluoroscopic images, lacks anatomical information, resulting in longer operation hours and radiation exposure. In this paper, a framework for real-time 3D robotic path planning from a single 2D fluoroscopic image of Abdominal Aortic Aneurysm (AAA) is introduced. A graph matching method is proposed to establish the correspondence between the 3D preoperative and 2D intra-operative AAA skeletons, and then the two skeletons are registered by skeleton deformation and regularization in respect to skeleton length and smoothness. Furthermore, deep learning was used to segment 3D pre-operative AAA from Computed Tomography (CT) scans to facilitate the framework automation. Simulation, phantom and patient AAA data sets have been used to validate the proposed framework. 3D distance error of 2mm was achieved in the phantom setup. Performance advantages were also achieved in terms of accuracy, robustness and time-efficiency. All the code will be open source.

Pre-operative Abdominal Aortic Aneurysm (AAA) 3D shape is critical for customized stent-graft design in Fenestrated Endovascular Aortic Repair (FEVAR). Traditional segmentation approaches implement expert-designed feature extractors while recent deep neural networks extract features automatically with multiple non-linear modules. Usually, a large training dataset is essential for applying deep learning on AAA segmentation. In this paper, the AAA was segmented using U-net with a small number (two) of training subjects. Firstly, Computed Tomography Angiography (CTA) slices were augmented with gray value variation and translation to avoid the overfitting caused by the small number of training subjects. Then, U-net was trained to segment the AAA. Dice Similarity Coefficients (DSCs) over 0.8 were achieved on the testing subjects. The PLZ, DLZ and aortic branches are all reconstructed reasonably, which will facilitate stent graft customization and help shape instantiation for intra-operative surgery navigation in FEVAR.

This paper presents the use of tubing to store clinical microdialysis samples for delayed analysis with high temporal resolution, offering an alternative to traditional discrete offline microdialysis sampling. Samples stored in this way were found to be stable for up to 72 days at −80 °C. Examples of how this methodology can be applied to glucose and lactate measurement in a wide range of in vivo monitoring experiments are presented. This paper presents a general model, which allows for an informed choice of tubing parameters for a given storage time and flow rate avoiding high back pressure, which would otherwise cause the microdialysis probe to leak, while maximising temporal resolution.

Mental health disorders are rank ed among the top twenty main causes of disability worldwide. It was found that there is an intriguing relationship between mental health problems and financial difficulties. Current technology uses mobile apps for self - monitoring of mental health conditions with a potential to avoid debt crisis caused by mental illness. In this paper, we propose the use of a wearable sensor as an objective evaluation tool for monitoring the e motional well - being of the subject. By fusing the sensory data with financial data, an intelligent self - guard system is proposed for preventing excessive spending caused by mental condition.

The diagnosis of CSF hypovolemia remains controversial. The primary diagnostic factor relies on confirmation of leakage of the CSF based on reduced spinal fluid pressure. Determining the specific leakage site is the most important issue for effective treatment but remains a difficult task. Although CT myelography, radioisotope cisternography, and MRI are commonly performed in the diagnosis of CSF hypovolemia, these techniques can rarely identify the precise leakage site. Therefore, an epidural blood patch is performed in the lumbar spine in many cases. This study reports a new diagnostic modality that can help to confirm the leakage site. Fat-suppressed T2-weighted sagittal images were compared before and after the infusion of 20 ml of saline into the subarachnoid space of the lumbar region to detect the specific leakage site with high probability. Three patients were successfully treated by the epidural blood patch based on data obtained with the new diagnostic modality. Two patients were treated in the cervical region and 1 in the lumbar region. The use of fat-suppressed T2-weighted sagittal images after saline infusion could be a relevant diagnostic modality compared with images obtained by CT myelography, radioisotope cisternography, and ordinary MRI to achieve accurate diagnosis and effective treatment of patients with CSF hypovolemia.

Background and Aim There are currently limited training and assessment tools available to novice endoscopists. A potential tool for the objective assessment of endoscopist visual search strategy is eye‐tracking technology. The aim of this study is to assess whether eye‐tracking technology can be used to differentiate the visual gaze patterns (VGP) of experienced and novice endoscopists, and to characterize any differences arising between the two groups. Method With the use of portable eye‐tracking glasses, VGP of novice (n = 20) and experienced (n = 14) endoscopists were compared while viewing a colonoscopy withdrawal through the hepatic flexure. Analysis was performed by comparing the central versus peripheral distribution and the horizontal distribution of novice and experienced endoscopist fixations, along with comparison of basic eye‐tracking metrics. Results This study found that experienced endoscopists had a significantly higher percentage of fixations within the periphery of the screen (13.4% vs 23.0%, P = 0.013). Experienced endoscopists also had a significantly greater percentage of fixations on the left side of the screen (18.6% vs 33.5%, P = 0.005) that displayed the poorly visualized “inside bend” of the hepatic flexure. Conclusion This study has detailed specific VGP acquired through expertise, which can potentially explain why adenomas are regularly missed at the hepatic flexure during colonoscopy. These may be useful for the training of novice colonoscopists, and further validation may utilize VGP in the development of an objective proficiency based curriculum to improve the detection of pathology and overall quality in endoscopy.

Understanding the solid biomechanics of the human body is important to the study of structure and function of the body, which can have a range of applications in health care, sport, well-being, and workflow analysis. Conventional laboratory-based biomechanical analysis systems and observation-based tests are designed only to capture brief snapshots of the mechanics of movement. With recent developments in wearable sensing technologies, biomechanical analysis can be conducted in less-constrained environments, thus allowing continuous monitoring and analysis beyond laboratory settings. In this paper, we review the current research in wearable sensing technologies for biomechanical analysis, focusing on sensing and analytics that enable continuous, long-term monitoring of kinematics and kinetics in a free-living environment. The main technical challenges, including measurement drift, external interferences, nonlinear sensor properties, sensor placement, and muscle variations, that can affect the accuracy and robustness of existing methods and different methods for reducing the impact of these sources of errors are described in this paper. Recent developments in motion estimation in kinematics, mobile force sensing in kinematics, sensor reduction for electromyography, and the future direction of sensing for biomechanics are also discussed.

Attitude estimation from miniature inertial and magnetic sensors has been used in a wide variety of applications, ranging from virtual reality, underwater vehicles, handheld navigation devices, to biomotion analysis. However, appropriate sensor calibrations for accurate sensor measurements are essential to the performance of attitude estimation algorithms. In this paper, we present a robust sensor calibration method for accurate attitude estimation from three-axis accelerometers, gyroscopes, and magnetometer measurements. The proposed calibration method only requires a simple pan-tilt unit. A unified sensor model for inertial and magnetic sensors is used to convert the sensor readings to physical quantities in metric units. Based on the sensor model, a cost function is constructed, and a two-step iterative algorithm is then proposed to calibrate the inertial sensors. Due to the difficulties of acquiring the ground-truth of the Earth magnetic field, a simplified pseudomagnetometer calibration method is also presented based on an ellipsoid fitting algorithm. The calibration method is then applied to our sensor nodes, and the good performance of the orientation estimation has illustrated the effectiveness of the proposed sensor calibration method.

This report describes the development of a deformable model for the automatic delineation of coronary artery cross-sectional areas with magnetic resonance imaging. The method is validated with coronary artery phantoms of varying diameters and images with different levels of signal-to-noise ratios. The reproducibility of the technique was examined with simulated geometrical shifts and motions during data acquisition. The experimental results indicate a very high reproducibility and low inter-observers variability of the technique, suggesting its suitability for non-invasive assessment of serial changes of vessel dilatation following pharmacological intervention.

Superconducting thick films (200$\mu$m-300$\mu$m) of YBCO have been fabricated successfully on MgO substrates by a new approach. The precursor powders (YBa2Cu3O7-$\delta$(Y123)+0.1molY2BaCuO5(Y211)) are placed between two MgO (10mm x 10mm) slices to form sandwich structure. The YBCO thick films have been obtained from the precursors by modified melt growth process. Resistance measurements of YBCO thick films show Tcon of 87.3K and $\Delta$T of 4.5K. Estimates using hysteresis loops and the Bean model give a value of 2.78 x 10$^{3}$ A/cm2 (77K, 0T) for the critical current density. The observations of scanning electron microscopy (SEM) and x-ray diffraction patterns (XRD) show the supercoducting Y123 phase matrix containing discrete Y211 inclusions. Keywords: superconducting thick films, YBCO, Y211 phase, peritectic temperature, Comment: 12 pages, 4 figure

Superconducting MgB2 thin films were fabricated on Al2O3(0001) substrates under an ex-situ processing conditions. Boron thin films were deposited by pulsed laser deposition followed by a post-annealing process. Resistance measurements of the deposited MgB2 films show Tc of about 39K, while scanning electron microscopy and x-ray diffraction analysis indicate that the films consist of well-crystallized grains with a highly c-axis-oriented structure., Comment: 7 Pages, 3 figures

This paper outlines the macroscopic anatomy and physiology of the heart, linking the micro and macroscopic structure of cardiac muscle fibre to its function during contraction and dilatation. The properties of cardiac muscle cells and the process of contraction at a cellular level are described. The macroscopic structure of the myocardium is further explained as one muscle band wound into a double twist. This helps to elucidate the muscle architecture and structure of the ventricles. Ventricular dynamics are also described as the twisting and untwisting of this muscle band to produce shortening and lengthening. Myocardial perfusion and causes of disease are discussed. Coronary artery disease and its effect on contractility is then described and ways of measuring contractility are introduced.

Optical imaging and sensing modalities have been used in medical diagnosis for many years. An obvious example is endoscopy, which allows remote wide-field imaging of internal tissues using optical fibers and/or miniature charge-coupled device (CCD) cameras. While techniques such as endoscopy provide useful tools for clinicians, they do not typically allow a complete diagnosis to be made. Instead, physical biopsies may be required to confirm or refute the presence of disease. Furthermore, endoscopic procedures are both invasive and time-consuming. As such, much research is currently directed toward the development of devices that can provide a complete in vivo diagnosis without the requirement for a physical biopsy. Ideally, such devices should also be minimally or non-invasive, and they should provide immediate identification of disease at the point of care. Additionally, there is significant interest in the development of implantable diagnostic devices that can be left within patients’ bodies for extended periods of time (for several days or longer). Such systems could be used for automated disease diagnosis, and example applications include the detection of post-surgical infections as well as monitoring of the health status of patients undergoing chemotherapy. This chapter focuses on the development of optical instruments that can provide in situ diagnosis at the point of care, with an emphasis on progress towards miniature devices that may function as implants in the future.

In Minimally Invasive Robotic Surgery (MIRS) a robot is interposed between the surgeon and the surgical site to increase the precision, dexterity, and to reduce surgeon's effort and cognitive load with respect to the standard laparoscopic interventions. However, the modern robotic systems for MIRS are still based on the traditional telemanipulation paradigm, e.g. the robot behaviour is fully under surgeon's control, and no autonomy or assistance is implemented. In this work, supervised and shared controllers have been developed in a vision-free, human-in-the-Ioop, control framework to help surgeon during a surgical suturing procedure. Experiments conducted on the da Vinci Research Kit robot proves the effectiveness of the method indicating also the guidelines for improving results.