Teaching Masked Autoencoder With Strong Augmentations.

Masked autoencoder (MAE) has been regarded as a capable self-supervised learner for various downstream tasks. Nevertheless, the model still lacks high-level discriminability, which results in poor linear probing performance. In view of the fact that strong augmentation plays an essential role in contrastive learning, can we capitalize on strong augmentation in MAE? The difficulty originates from the pixel uncertainty caused by strong augmentation that may affect the reconstruction, and thus, directly introducing strong augmentation into MAE often hurts the performance. In this article, we delve into the potential of strong augmented views to enhance MAE while maintaining MAE's advantages. To this end, we propose a simple yet effective masked Siamese autoencoder (MSA) model, which consists of a student branch and a teacher branch. The student branch derives MAE's advanced architecture, and the teacher branch treats the unmasked strong view as an exemplary teacher to impose high-level discrimination onto the student branch. We demonstrate that our MSA can improve the model's spatial perception capability and, therefore, globally favors interimage discrimination. Empirical evidence shows that the model pretrained by MSA provides superior performances across different downstream tasks. Notably, linear probing performance on frozen features extracted from MSA leads to 6.1% gains over MAE on ImageNet-1k. Fine-tuning (FT) the network on VQAv2 task finally achieves 67.4% accuracy, outperforming 1.6% of the supervised method DeiT and 1.2% of MAE. Codes and models are available at https://github.com/KimSoybean/MSA.

Self-Adaptive Perception of Object's Deformability with Multiple Deformation Attributes Utilizing Biomimetic Mechanoreceptors.

The perception of object's deformability in unstructured interactions relies on both kinesthetic and cutaneous cues to adapt the uncertainties of an object. However, the existing tactile sensors cannot provide adequate cutaneous cues to self-adaptively estimate the material softness, especially in non-standard contact scenarios where the interacting object deviates from the assumption of an elastic half-infinite body. This paper proposes an innovative design of a tactile sensor that integrates the capabilities of two slow-adapting mechanoreceptors within a soft medium, allowing self-decoupled sensing of local pressure and strain at specific locations within the contact interface. By leveraging these localized cutaneous cues, the sensor can accurately and self-adaptively measure the material softness of an object, accommodating variations in thicknesses and applied forces. Furthermore, when combined with a kinesthetic cue from the robot, the sensor can enhance tactile expression by the synergy of two relevant deformation attributes, including material softness and compliance. It is demonstrated that the biomimetic fusion of tactile information can fully comprehend the deformability of an object, hence facilitating robotic decision-making and dexterous manipulation.

A Low-Cost Instrumented Shoe System for Gait Phase Detection Based on Foot Plantar Pressure Data.

This paper presents a novel low-cost and fully-portable instrumented shoe system for gait phase detection. The instrumented shoe consists of 174 independent sensing units constructed based on an off-the-shelf force-sensitive film known as the Velostat conductive copolymer. A zero potential method was implemented to address the crosstalk effect among the matrix-formed sensing arrays. A customized algorithm for gait event and phase detection was developed to estimate stance sub-phases including initial contact, flat foot, and push off. Experiments were carried out to evaluate the performance of the proposed instrumented shoe system in gait phase detection for both straight-line walking and turning walking. The results showed that the mean absolute time differences between the estimated phases by the proposed instrumented shoe system and the reference measurement ranged from 45 to 58 ms during straight-line walking and from 51 to 77 ms during turning walking, which were comparable to the state of art.Clinical and Translational Impact Statement-By allowing convenient gait monitoring in home healthcare settings, the proposed system enables extensive ADL data collection and facilitates developing effective treatment and rehabilitation strategies for patients with movement disorders.

Surgical Continuum Manipulator Control Using Multiagent Team Deep Q Learning.

Continuum manipulator has shown great potential in surgical applications. The flexibility of the continuum manipulator helps it achieve many complicated surgeries, such as neurosurgery, vascular surgery, abdominal surgery, etc. In this paper, we propose a Team Deep Q learning framework (TDQN) to control a 2-DoF surgical continuum manipulator with four cables, where two cables in a pair form one agent. During the learning process, each agent shares state and reward information with the other one, which namely is centralized learning. Using the shared information, TDQN shows better targeting accuracy than multiagent deep Q learning (MADQN) by verifying on a 2-DoF cable-driven surgical continuum manipulator. The root mean square error during tracking with and without disturbance are 0.82mm and 0.16mm respectively using TDQN, whereas 1.52mm and 0.98mm using MADQN respectively.Clinical Relevance-The proposed TDQN shows a promising future in improving control accuracy under disturbance and maneuverability in robotic-assisted endoscopic surgery.

Training Induced Improvement on Human Perception of the Force Feedback Provided by A Soft Robotic Glove.

Creating haptic interface by glove-based wearable robotic system has become an increasingly interested topic in the area of human robotic interaction. Many force feedback gloves are constructed based on soft actuators. However, the recent development of haptic and force feedback technology mainly focused on the advancement of the actuating components and mechanism, and innovation of the force feedback rendering algorithms. It seems that another important part of this human-robot-interaction loop, i.e. the human factors, were understudied. Here, this study focused on the learning effect in haptic perception. We designed a pneumatic muscle-based force feedback robotic glove which can provide customized force feedback to the dorsal surface of each finger. An experiment was carried out with a specifically training procedure and evaluation on the force feedback perceptions. The results show that practice-induced improvement can be achieved by this training, allowing people have a better perception of the force feedback provided by this glove.

Optical and MRI Multimodal Tracing of Stem Cells In Vivo.

Stem cell therapy has shown great clinical potential in oncology, injury, inflammation, and cardiovascular disease. However, due to the technical limitations of the in vivo visualization of transplanted stem cells, the therapeutic mechanisms and biosafety of stem cells in vivo are poorly defined, which limits the speed of clinical translation. The commonly used methods for the in vivo tracing of stem cells currently include optical imaging, magnetic resonance imaging (MRI), and nuclear medicine imaging. However, nuclear medicine imaging involves radioactive materials, MRI has low resolution at the cellular level, and optical imaging has poor tissue penetration in vivo. It is difficult for a single imaging method to simultaneously achieve the high penetration, high resolution, and noninvasiveness needed for in vivo imaging. However, multimodal imaging combines the advantages of different imaging modalities to determine the fate of stem cells in vivo in a multidimensional way. This review provides an overview of various multimodal imaging technologies and labeling methods commonly used for tracing stem cells, including optical imaging, MRI, and the combination of the two, while explaining the principles involved, comparing the advantages and disadvantages of different combination schemes, and discussing the challenges and prospects of human stem cell tracking techniques.

Molecular mechanisms underlying human spatial cognitive ability revealed with neurotransmitter and transcriptomic mapping.

Mental rotation, one of the cores of spatial cognitive abilities, is closely associated with spatial processing and general intelligence. Although the brain underpinnings of mental rotation have been reported, the cellular and molecular mechanisms remain unexplored. Here, we used magnetic resonance imaging, a whole-brain spatial distribution atlas of 19 neurotransmitter receptors, transcriptomic data from Allen Human Brain Atlas, and mental rotation performances of 356 healthy individuals to identify the genetic/molecular foundation of mental rotation. We found significant associations of mental rotation performance with gray matter volume and fractional amplitude of low-frequency fluctuations in primary visual cortex, fusiform gyrus, primary sensory-motor cortex, and default mode network. Gray matter volume and fractional amplitude of low-frequency fluctuations in these brain areas also exhibited significant sex differences. Importantly, spatial correlation analyses were conducted between the spatial patterns of gray matter volume or fractional amplitude of low-frequency fluctuations with mental rotation and the spatial distribution patterns of neurotransmitter receptors and transcriptomic data, and identified the related genes and neurotransmitter receptors associated with mental rotation. These identified genes are localized on the X chromosome and are mainly involved in trans-synaptic signaling, transmembrane transport, and hormone response. Our findings provide initial evidence for the neural and molecular mechanisms underlying spatial cognitive ability.

Road Narrow-Inspired Strain Concentration to Wide-Range-Tunable Gauge Factor of Ionic Hydrogel Strain Sensor.

The application of stretchable strain sensors in human movement recognition, health monitoring, and soft robotics has attracted wide attention. Compared with traditional electronic conductors, stretchable ionic hydrogels are more attractive to organization-like soft electronic devices yet suffer poor sensitivity due to limited ion conduction modulation caused by their intrinsic soft chain network. This paper proposes a strategy to modulate ion transport behavior by geometry-induced strain concentration to adjust and improve the sensitivity of ionic hydrogel-based strain sensors (IHSS). Inspired by the phenomenon of vehicles slowing down and changing lanes when the road narrows, the strain redistribution of ionic hydrogel is optimized by structural and mechanical parameters to produce a strain-induced resistance boost. As a result, the gauge factor of the IHSS is continuously tunable from 1.31 to 9.21 in the strain range of 0-100%, which breaks through the theoretical limit of homogeneous strain-distributed ionic hydrogels and ensures a linear electromechanical response simultaneously. Overall, this study offers a universal route to modulate the ion transport behavior of ionic hydrogels mechanically, resulting in a tunable sensitivity for IHSS to better serve different application scenarios, such as health monitoring and human-machine interface.

A Heuristically Accelerated Reinforcement Learning-Based Neurosurgical Path Planner.

The steerable needle becomes appealing in the neurosurgery intervention procedure because of its flexibility to bypass critical regions inside the brain; with proper path planning, it can also minimize the potential damage by setting constraints and optimizing the insertion path. Recently, reinforcement learning (RL)-based path planning algorithm has shown promising results in neurosurgery, but because of the trial and error mechanism, it can be computationally expensive and insecure with low training efficiency. In this paper, we propose a heuristically accelerated deep Q network (DQN) algorithm to safely preoperatively plan a needle insertion path in a neurosurgical environment. Furthermore, a fuzzy inference system is integrated into the framework as a balance of the heuristic policy and the RL algorithm. Simulations are conducted to test the proposed method in comparison to the traditional greedy heuristic searching algorithm and DQN algorithms. Tests showed promising results of our algorithm in saving over 50 training episodes, calculating path lengths of 0.35 after normalization, which is 0.61 and 0.39 for DQN and traditional greedy heuristic searching algorithm, respectively. Moreover, the maximum curvature during planning is reduced to 0.046 from 0.139 mm-1 using the proposed algorithm compared to DQN.

Characteristics of foot plantar pressure during turning in young male adults.

BACKGROUND: Turning gait is considered as a challenging motor task. However, only few existing studies reported turning biomechanics from the aspect of foot plantar pressure. RESEARCH QUESTION: This study aimed to investigate turning biomechanics by studying foot plantar pressure characteristics METHODS: Twelve young male participants were involved in this experimental study. They were instructed to perform turning tasks with different turning angles (i.e., 30°, 60°, and 90°). Foot plantar pressure was quantified by the force time integral (FTI) underneath seven plantar sub-areas. Analysis was carried out for different turning strategies (spin turns versus step turns), separately. RESULTS: The results showed that for small-angle spin turns, plantar pressure patterns changed at the early stage of the approaching step, suggesting a preparatory action for the increased lower limb range of motion in the transverse plane during turning; for step turns, an imbalance weight bearing mechanism was adopted when making large-angle turns to compensate for the centripetal force during turning. SIGNIFICANCE: The findings provide improved knowledge about turning biomechanics. They have practical implications for motion planning of lower-limb assistive devices for those with difficulties in turning.

Mechanically Interlocked Hydrogel-Elastomer Strain Sensor with Robust Interface and Enhanced Water-Retention Capacity.

Hydrogels are stretchable ion conductors that can be used as strain sensors by transmitting strain-dependent electrical signals. However, hydrogels are susceptible to dehydration in the air, leading to a loss of flexibility and functions. Here, a simple and general strategy for encapsulating hydrogel with hydrophobic elastomer is proposed to realize excellent water-retention capacity. Elastomers, such as polydimethylsiloxanes (PDMS), whose hydrophobicity and dense crosslinking network can act as a barrier against water evaporation (lost 4.6 wt.% ± 0.57 in 24 h, 28 °C, and ≈30% humidity). To achieve strong adhesion between the hydrogel and elastomer, a porous structured thermoplastic polyurethane (TPU) is used at the hydrogel-elastomer interface to interlock the hydrogel and bond the elastomer simultaneously (the maximum interfacial toughness is over 1200 J/m2). In addition, a PDMS encapsulated ionic hydrogel strain sensor is proposed, demonstrating an excellent water-retention ability, superior mechanical performance, highly linear sensitivity (gauge factor = 2.21, at 100% strain), and robust interface. Various human motions were monitored, proving the effectiveness and practicability of the hydrogel-elastomer hybrid.

Recurrent neural networks as kinematics estimator and controller for redundant manipulators subject to physical constraints.

Redundant manipulators could be efficient tools in industrial production as a result of their dexterity. However, existing kinematic control methods for redundant manipulators have two main disadvantages. On one hand, model uncertainties or unknown kinematic parameters may degrade the performance of existing model-based control methods subject to joint limits. On the other hand, existing model-free control methods ignore the existence of joint limits although they do not need to know kinematic models of redundant manipulators. In this paper, a quadratic programming (QP) scheme is elaborated to achieve the primary tracking control task of redundant manipulators as well as joint limits avoidance task. Besides, a gradient neurodynamics (GND) model is utilized to estimate the kinematics of redundant manipulators. Then, a primal dual neural network, which is employed to solve the QP problem, and the GND model are integrated towards developing a model-free control method constrained by joint angle and velocity limits for redundant manipulators. The visual sensory feedback is fed to the two neural networks. The efficacy of the proposed control method is demonstrated by extensive simulations and experiments, and the merits of the proposed method are also substantiated by comparisons.

Design and Control of a Highly Redundant Rigid-flexible Coupling Robot to Assist the COVID-19 Oropharyngeal-Swab Sampling.

The outbreak of novel coronavirus pneumonia (COVID-19) has caused mortality and morbidity worldwide. Oropharyngeal-swab (OP-swab) sampling is widely used for the diagnosis of COVID-19 in the world. To avoid the clinical staff from being affected by the virus, we developed a 9-degree-of-freedom (DOF) rigid-flexible coupling (RFC) robot to assist the COVID-19 OP-swab sampling. This robot is composed of a visual system, UR5 robot arm, micro-pneumatic actuator and force-sensing system. The robot is expected to reduce risk and free up the clinical staff from the long-term repetitive sampling work. Compared with a rigid sampling robot, the developed force-sensing RFC robot can facilitate OP-swab sampling procedures in a safer and softer way. In addition, a varying-parameter zeroing neural network-based optimization method is also proposed for motion planning of the 9-DOF redundant manipulator. The developed robot system is validated by OP-swab sampling on both oral cavity phantoms and volunteers.

Maximum Correntropy with Variable Center Unscented Kalman Filter for Robust Power System State Estimation.

The robust Kalman filter with correntropy loss has received much attention in recent years for forecasting-aided state estimation in power systems, since it efficiently reduces the negative influence of various abnormal situations, such as non-Gaussian communication, changing environment, and instrument failures, and obviously improves the stability of power systems. However, the existing correntropy-based robust Kalman filters usually use the Gaussian function with a fixed center as the kernel function in correntropy, which may not be a suitable choice in practical applications of power system forecasting-aided state estimation (PSSE). To address this issue, a new and robust unscented Kalman filter, called the maximum correntropy with variable center unscented Kalman filter (MCVUKF), is proposed in this paper for PSSE. Specifically, MCVUKF adopts an extended version of correntropy, whose center can be located at any position, to replace the original correntropy in an unscented Kalman filter to improve the performance in PSSE. Moreover, by using an exponential function of the innovation vector to adjust a covariance matrix, an enhanced MCVUKF (En-MCVUKF) method is also developed for suppressing the influence of bad data to the innovation vector and further improving the accuracy of PSSE. Finally, extensive simulations have been conducted on IEEE 14-bus, 30-bus, and 57-bus test power systems, and the simulation results have shown the superiority of the proposed MCVUKF and En-MCVUKF methods compared with several related state-of-the-art Kalman filter methods.

A collaborative robot for COVID-19 oropharyngeal swabbing.

The coronavirus disease 2019 (COVID-19) outbreak has increased mortality and morbidity world-wide. Oropharyngeal swabbing is a well-known and commonly used sampling technique for COVID-19 diagnose around the world. We developed a robot to assist with COVID-19 oropharyngeal swabbing to prevent frontline clinical staff from being infected. The robot integrates a UR5 manipulator, rigid-flexible coupling (RFC) manipulator, force-sensing and control subsystem, visual subsystem and haptic device. The robot has strength in intrinsically safe and high repeat positioning accuracy. In addition, we also achieve one-dimensional constant force control in the automatic scheme (AS). Compared with the rigid sampling robot, the developed robot can perform the oropharyngeal swabbing procedure more safely and gently, reducing risk. Alternatively, a novel robot control schemes called collaborative manipulation scheme (CMS) which combines a automatic phase and teleoperation phase is proposed. At last, comparative experiments of three schemes were conducted, including CMS, AS, and teleoperation scheme (TS). The experimental results shows that CMS obtained the highest score according to the evaluation equation. CMS has the excellent performance in quality, experience and adaption. Therefore, the proposal of CMS is meaningful which is more suitable for robot-sampling.

A Soft Robotic Intervention for Gait Enhancement in Older Adults.

Falls continue to be a major safety and health concern for older adults. Researchers reported that increased gait variability was associated with increased fall risks. In the present study, we proposed a novel wearable soft robotic intervention and examined its effects on improving gait variability in older adults. The robotic system used customized pneumatic artificial muscles (PAMs) to provide assistive torque for ankle dorsiflexion during walking. Twelve older adults with low fall risks and twelve with medium-high fall risks participated in an experiment. The participants were asked to walk on a treadmill under no soft robotic intervention, inactive soft robotic intervention, and active soft robotic intervention, and their gait variability during treadmill walking was measured. The results showed that the proposed soft robotic intervention could reduce step length variability for elderly people with medium-high fall risks. These findings provide supporting evidence that the proposed soft robotic intervention could potentially serve as an effective solution to fall prevention for older adults.

Nanosecond pulsed electric fields impair viability and mucin expression in mucinous colorectal carcinoma cell.

Nanosecond pulsed electric fields (nsPEFs) are a non-thermal technology that can induce a myriad of biological responses and changes in cellular physiology. nsPEFs have gained significant attention as a novel cancer therapy. However, studies investigating the application of nsPEF in mucinous carcinomas are scarce. In this study, we explored several biological responses in two mucinous colorectal adenocarcinoma cell lines, LS 174T and HT-29, to nsPEF treatment. We determined the overall cell survival and viability rates following nsPEF treatment using CCK-8 and colony formation assays. We measured the intracellular effects of nsPEF treatment by analyzing cell cycle distribution, cell apoptosis and mitochondrial potential. We also analyzed mucin production at both mRNA and protein levels. Our results showed that nsPEF treatment significantly reduced mucinous cell viability in a dose-dependent manner. nsPEF treatment increased cell cycles arrest at G0/G1 while the proportion of G2/M cells gradually decreased. Cell apoptosis increased following nsPEF treatment with a clear loss in mitochondrial membrane potential. Furthermore, the protein expression of functional mucin family members decreased after nsPEF treatment. In conclusion, nsPEF treatment reduced MCRC cell viability, cell proliferation, and mucin protein production while promoted apoptosis. Our work is a pilot study that projects some insights into the potential clinical applications of nsPEFs in treating mucinous colorectal carcinoma.

A hybrid learning-based hysteresis compensation strategy for surgical robots.

BACKGROUND: The hysteretic forces arising from the electric cables that externally run along the robotic joints are the main disturbance to the precise parameter estimation of gravity compensation model, for the Master Tool Manipulator (MTM) of the da Vinci Research Kit (dVRK). Because such nonlinear disturbance forces and the gravitational forces are often hybrid and in the same magnitude. METHODS: A strategy is proposed to separate these two hybrid forces, and model them by individual learning-based algorithms. A specially designed Elastic Hysteresis Neural Network model is employed to capture the hysteresis nature of disturbance forces. RESULTS: The experimental results show that our proposed strategy has higher compensation accuracy (78.64%-93.32%), and fewer real samples are required for model estimation (100 samples for each joint). CONCLUSIONS: Our proposed gravity compensation strategy for the MTM of the dVRK shows great improvement over existing state-of-the-arts methods through conducted comparative experiments.

High-fidelity structured illumination microscopy by point-spread-function engineering.

Structured illumination microscopy (SIM) has become a widely used tool for insight into biomedical challenges due to its rapid, long-term, and super-resolution (SR) imaging. However, artifacts that often appear in SIM images have long brought into question its fidelity, and might cause misinterpretation of biological structures. We present HiFi-SIM, a high-fidelity SIM reconstruction algorithm, by engineering the effective point spread function (PSF) into an ideal form. HiFi-SIM can effectively reduce commonly seen artifacts without loss of fine structures and improve the axial sectioning for samples with strong background. In particular, HiFi-SIM is not sensitive to the commonly used PSF and reconstruction parameters; hence, it lowers the requirements for dedicated PSF calibration and complicated parameter adjustment, thus promoting SIM as a daily imaging tool.

Measurement of expansion factor and distortion for expansion microscopy using isolated renal glomeruli as landmarks.

Expansion microscopy has enabled super resolution imaging of biological samples. The accurate measurement of expansion factor and distortion typically requires locating and imaging the same region of interest in the sample before and after expansion, which is often time-consuming to achieve. Here we introduce a convenient method for relocation by utilizing isolated porcine glomeruli as landmarks during expansion. Following heat denaturation and proteinase K digestion protocols, the glomeruli exhibit expansion factor of 3.5 to 4 (only 7%-16% less expanded than the hydrogel), and 1% to 2% of relative distortion. Due to its appropriate size of 100 to 300 µm, the location of the glomerulus in the sample are visible to eyes, while its detailed shape only requires bright field microscopy. For expansion factors ranging from 3 to 10, the region in the vicinity of the glomerulus can be easily re-identified, and sometimes allows quantification of expansion factor and distortion under bright field without fluorescent labels.