ABSTRACT
Identifying an accurate dynamics model remains challenging for humanoid robots. The difficulty is mainly due to the following two points. First, a good initial model is required to evaluate the feasibility of motions for data acquisition. Second, a highly nonlinear optimization problem needs to be solved to design movements to acquire the identification data. To cope with the first point, in this paper, we propose a curriculum of identification to gradually learn an accurate dynamics model from an unreliable initial model. For the second point, we propose using a large-scale human motion database to efficiently design the humanoid movements for the parameter identification. The contribution of our study is developing a humanoid identification method that does not require the good initial model and does not need to solve the highly nonlinear optimization problem. We showed that our curriculum-based approach was able to more efficiently identify humanoid model parameters than a method that just randomly picked reference motions for identification. We evaluated our proposed method in a simulation experiment and demonstrated that our curriculum was led to obtain a wide variety of motion data for efficient parameter estimation. Consequently, our approach successfully identified an accurate model of an 18-DoF, simulated upper-body humanoid robot.
ABSTRACT
A physical trainer often physically guides a learner's limbs to teach an ideal movement, giving the learner proprioceptive information about the movement to be reproduced later. This instruction requires the learner to perceive kinesthetic information and store the instructed information temporarily. Therefore, (1) proprioceptive acuity to accurately perceive the taught kinesthetics and (2) short-term memory to store the perceived information are two critical functions for reproducing the taught movement. While the importance of proprioceptive acuity and short-term memory has been suggested for active motor learning, little is known about passive motor learning. Twenty-one healthy adults (mean age 25.6 years, range 19-38 years) participated in this study to investigate whether individual learning efficiency in passively guided learning is related to these two functions. Consequently, learning efficiency was significantly associated with short-term memory capacity. In particular, individuals who could recall older sensory stimuli showed better learning efficiency. However, no significant relationship was observed between learning efficiency and proprioceptive acuity. A causal graph model found a direct influence of memory on learning and an indirect effect of proprioceptive acuity on learning via memory. Our findings suggest the importance of a learner's short-term memory for effective passive motor learning.
Subject(s)
Memory, Short-Term , Psychomotor Performance , Adult , Humans , Young Adult , Proprioception , Learning , KinesthesisABSTRACT
Background: The aim of the study was to provide real-world data on the effectiveness and safety of a new fixed-ratio combination, insulin degludec/liraglutide (IDegLira) injection in Japanese patients with type 2 diabetes mellitus (T2DM). Methods: The primary endpoint was the change in glycated hemoglobin (HbA1c) level 6 months after the introduction of IDegLira. We also examined the rate of achievement of target HbA1c 7% and the individualized HbA1c targets set for each patient. Baseline characteristics associated with the change in HbA1c were also assessed. Seventy-five patients with T2DM were included in the analysis. Results: After the initiation of IDegLira, HbA1c decreased significantly from baseline with a change of -1.81% (baseline 9.61% and at 6 months 7.80%; P < 0.001). At baseline, the achievement rate of 7% HbA1c was 2.67% (n = 2), which increased to 36.0% (n = 27) after 6 months of IDegLira introduction (P < 0.05). The attainment rate of individualized HbA1c targets, which were set considering each patient's characteristics, improved from 2.67% (n = 2) to 49.3% (n = 37) (P < 0.001). Regardless of sex, body mass index, estimated glomerular filtration rate, duration of diabetes, or history of glucagon-like peptide-1 receptor agonist use, IDegLira significantly reduced HbA1c, but a higher C-peptide index was associated with a greater reduction in HbA1c. Conclusion: In this study, initiation of IDegLira in a real-world clinical setting was beneficial in lowering HbA1c in Japanese T2DM patients with inadequate glycemic control with existing therapy.
ABSTRACT
This study introduces a body-weight-support (BWS) robot actuated by two pneumatic artificial muscles (PAMs). Conventional BWS devices typically use springs or a single actuator, whereas our robot has a split force-controlled BWS (SF-BWS), in which two force-controlled actuators independently support the left and right sides of the user's body. To reduce the experience of weight, vertical unweighting support forces are transferred directly to the user's left and right hips through a newly designed harness with an open space around the shoulder and upper chest area to allow freedom of movement. A motion capture evaluation with three healthy participants confirmed that the proposed harness does not impede upper-body motion during laterally identical force-controlled partial BWS walking, which is quantitatively similar to natural walking. To evaluate our SF-BWS robot, we performed a force-tracking and split-force control task using different simulated load weight setups (40, 50, and 60 kg masses). The split-force control task, providing independent force references to each PAM and conducted with a 60 kg mass and a test bench, demonstrates that our SF-BWS robot is capable of shifting human body weight in the mediolateral direction. The SF-BWS robot successfully controlled the two PAMs to generate the desired vertical support forces.
ABSTRACT
Glucagon-like peptide-1 receptor agonists (GLP-1RA) have a more potent glycated hemoglobin (HbA1c)-lowering effect than existing therapies and are widely used for treating type 2 diabetes mellitus (T2DM). Once-daily oral semaglutide is the world's first oral GLP-1RA. This study aimed to provide real-world data on oral semaglutide in Japanese patients with T2DM and its effects on cardiometabolic parameters. This was a single-center retrospective observational study. We examined changes in HbA1c and body weight (BW) and the rate of achieving HbA1c < 7% after 6 months of oral semaglutide treatment in Japanese patients with T2DM. Furthermore, we examined differences in the efficacy of oral semaglutide with multiple patient backgrounds. A total of 88 patients were included in this study. Overall, the mean (standard error of the mean) HbA1c at 6 months decreased by -1.24% (0.20%) from baseline, and BW at 6 months (n = 85) also decreased by -1.44 kg (0.26 kg) from baseline. The percentage of patients who achieved HbA1c < 7% changed significantly from 14% at baseline to 48%. HbA1c decreased from baseline regardless of age, sex, body mass index, chronic kidney disease, or diabetes duration. Additionally, alanine aminotransferase, total cholesterol, triglyceride, and non-high-density lipoprotein cholesterol were significantly reduced from baseline. Oral semaglutide may be an effective option for the intensification of therapy in Japanese patients with T2DM who have inadequate glycemic control with existing therapy. It may also reduce BW and improve cardiometabolic parameters.
ABSTRACT
AIM: Increasing evidence suggests that psychiatric disorders are linked to alterations in the mesocorticolimbic dopamine-related circuits. However, the common and disease-specific alterations remain to be examined in schizophrenia (SCZ), major depressive disorder (MDD), and autism spectrum disorder (ASD). Thus, this study aimed to examine common and disease-specific features related to mesocorticolimbic circuits. METHODS: This study included 555 participants from four institutes with five scanners: 140 individuals with SCZ (45.0% female), 127 individuals with MDD (44.9%), 119 individuals with ASD (15.1%), and 169 healthy controls (HC) (34.9%). All participants underwent resting-state functional magnetic resonance imaging. A parametric empirical Bayes approach was adopted to compare estimated effective connectivity among groups. Intrinsic effective connectivity focusing on the mesocorticolimbic dopamine-related circuits including the ventral tegmental area (VTA), shell and core parts of the nucleus accumbens (NAc), and medial prefrontal cortex (mPFC) were examined using a dynamic causal modeling analysis across these psychiatric disorders. RESULTS: The excitatory shell-to-core connectivity was greater in all patients than in the HC group. The inhibitory shell-to-VTA and shell-to-mPFC connectivities were greater in the ASD group than in the HC, MDD, and SCZ groups. Furthermore, the VTA-to-core and VTA-to-shell connectivities were excitatory in the ASD group, while those connections were inhibitory in the HC, MDD, and SCZ groups. CONCLUSION: Impaired signaling in the mesocorticolimbic dopamine-related circuits could be an underlying neuropathogenesis of various psychiatric disorders. These findings will improve the understanding of unique neural alternations of each disorder and will facilitate identification of effective therapeutic targets.
Subject(s)
Autism Spectrum Disorder , Depressive Disorder, Major , Mental Disorders , Humans , Female , Male , Depressive Disorder, Major/diagnostic imaging , Dopamine , Bayes Theorem , Neural Pathways/diagnostic imaging , Magnetic Resonance Imaging , Prefrontal Cortex/diagnostic imaging , Mental Disorders/diagnostic imagingABSTRACT
BACKGROUND AND HYPOTHESIS: Dynamics of the distributed sets of functionally synchronized brain regions, known as large-scale networks, are essential for the emotional state and cognitive processes. However, few studies were performed to elucidate the aberrant dynamics across the large-scale networks across multiple psychiatric disorders. In this paper, we aimed to investigate dynamic aspects of the aberrancy of the causal connections among the large-scale networks of the multiple psychiatric disorders. STUDY DESIGN: We applied dynamic causal modeling (DCM) to the large-sample multi-site dataset with 739 participants from 4 imaging sites including 4 different groups, healthy controls, schizophrenia (SCZ), major depressive disorder (MDD), and bipolar disorder (BD), to compare the causal relationships among the large-scale networks, including visual network, somatomotor network (SMN), dorsal attention network (DAN), salience network (SAN), limbic network (LIN), frontoparietal network, and default mode network. STUDY RESULTS: DCM showed that the decreased self-inhibitory connection of LIN was the common aberrant connection pattern across psychiatry disorders. Furthermore, increased causal connections from LIN to multiple networks, aberrant self-inhibitory connections of DAN and SMN, and increased self-inhibitory connection of SAN were disorder-specific patterns for SCZ, MDD, and BD, respectively. CONCLUSIONS: DCM revealed that LIN was the core abnormal network common to psychiatric disorders. Furthermore, DCM showed disorder-specific abnormal patterns of causal connections across the 7 networks. Our findings suggested that aberrant dynamics among the large-scale networks could be a key biomarker for these transdiagnostic psychiatric disorders.
Subject(s)
Bipolar Disorder , Depressive Disorder, Major , Humans , Depressive Disorder, Major/diagnostic imaging , Magnetic Resonance Imaging/methods , Brain/diagnostic imaging , Bipolar Disorder/diagnostic imaging , Brain Mapping/methodsABSTRACT
Are leaders made or born? Leader-follower roles have been well characterized in social science, but they remain somewhat obscure in sensory-motor coordination. Furthermore, it is unknown how and why leader-follower relationships are acquired, including innate versus acquired controversies. We developed a novel asymmetrical coordination task in which two participants (dyad) need to collaborate in transporting a simulated beam while maintaining its horizontal attitude. This experimental paradigm was implemented by twin robotic manipulanda, simulated beam dynamics, haptic interactions, and a projection screen. Clear leader-follower relationships were learned only when strong haptic feedback was introduced. This phenomenon occurred despite participants not being informed that they were interacting with each other and the large number of equally-valid alternative dyadic coordination strategies. We demonstrate the emergence of consistent leader-follower relationships in sensory-motor coordination, and further show that haptic interaction is essential for dyadic co-adaptation. These results provide insights into neural mechanisms responsible for the formation of leader-follower relationships in our society.
Subject(s)
Haptic Technology , Learning , Humans , Acclimatization , Biological TransportABSTRACT
Genu recurvatum (knee hyperextension) is a common problem after stroke. It is important to promote the coordination between knee and ankle movements during gait; however, no study has investigated how multi-joint assistance affects genu recurvatum. We are developing a gait training technique that uses robotized knee-ankle-foot orthosis (KAFO) to assists the knee and ankle joints simultaneously. This report aimed to investigate the safety of robotized KAFO-assisted gait training (Experiment 1) and a clinical trial to treat genu recurvatum in a patient with stroke (Experiment 2). Six healthy participants and eight patients with chronic stroke participated in Experiment 1. They received robotized KAFO-assisted gait training for one or 10 sessions. One patient with chronic stroke participated in Experiment 2 to investigate the effect of robotized KAFO-assisted gait training on genu recurvatum. The patient received the training for 30 min/day for nine days. The robot consisted of KAFO and an attached actuator of four pneumatic artificial muscles. The assistance parameters were adjusted by therapists to prevent genu recurvatum during gait. In Experiment 2, we evaluated the knee joint angle during overground gait, Fugl-Meyer Assessment of lower extremity (FMA-LE), modified Ashworth scale (MAS), Gait Assessment and Intervention Tool (G.A.I.T.), 10-m gait speed test, and 6-min walk test (6MWT) before and after the intervention without the robot. All participants completed the training in both experiments safely. In Experiment 2, genu recurvatum, FMA-LE, MAS, G.A.I.T., and 6MWT improved after robotized KAFO-assisted gait training. The results indicated that the multi-joint assistance robot may be effective for genu recurvatum after stroke.
ABSTRACT
Recent advances in deep neural networks have opened up new possibilities for visuomotor robot learning. In the context of human-robot or robot-robot collaboration, such networks can be trained to predict future poses and this information can be used to improve the dynamics of cooperative tasks. This is important, both in terms of realizing various cooperative behaviors, and for ensuring safety. In this article, we propose a recurrent neural architecture, capable of transforming variable-length input motion videos into a set of parameters describing a robot trajectory, where predictions can be made after receiving only a few frames. A simulation environment is utilized to expand the training database and to improve generalization capability of the network. The resulting architecture demonstrates good accuracy when predicting handover trajectories, with models trained on synthetic and real data showing better performance than when trained on real or simulated data only. The computed trajectories enable the execution of handover tasks with uncalibrated robots, which was verified in an experiment with two real robots.
ABSTRACT
Deep learning (DL) and reinforcement learning (RL) methods seem to be a part of indispensable factors to achieve human-level or super-human AI systems. On the other hand, both DL and RL have strong connections with our brain functions and with neuroscientific findings. In this review, we summarize talks and discussions in the "Deep Learning and Reinforcement Learning" session of the symposium, International Symposium on Artificial Intelligence and Brain Science. In this session, we discussed whether we can achieve comprehensive understanding of human intelligence based on the recent advances of deep learning and reinforcement learning algorithms. Speakers contributed to provide talks about their recent studies that can be key technologies to achieve human-level intelligence.
Subject(s)
Artificial Intelligence , Deep Learning , Algorithms , Humans , Reinforcement, PsychologyABSTRACT
AIM: Recently, a machine-learning (ML) technique has been used to create generalizable classifiers for psychiatric disorders based on information of functional connections (FCs) between brain regions at resting state. These classifiers predict diagnostic labels by a weighted linear sum (WLS) of the correlation values of a small number of selected FCs. We aimed to develop a generalizable classifier for gambling disorder (GD) from the information of FCs using the ML technique and examine relationships between WLS and clinical data. METHODS: As a training dataset for ML, data from 71 GD patients and 90 healthy controls (HCs) were obtained from two magnetic resonance imaging sites. We used an ML algorithm consisting of a cascade of an L1-regularized sparse canonical correlation analysis and a sparse logistic regression to create the classifier. The generalizability of the classifier was verified using an external dataset. This external dataset consisted of six GD patients and 14 HCs, and was collected at a different site from the sites of the training dataset. Correlations between WLS and South Oaks Gambling Screen (SOGS) and duration of illness were examined. RESULTS: The classifier distinguished between the GD patients and HCs with high accuracy in leave-one-out cross-validation (area under curve (AUC = 0.89)). This performance was confirmed in the external dataset (AUC = 0.81). There was no correlation between WLS, and SOGS and duration of illness in the GD patients. CONCLUSION: We developed a generalizable classifier for GD based on information of functional connections between brain regions at resting state.
Subject(s)
Gambling , Algorithms , Brain/diagnostic imaging , Gambling/diagnostic imaging , Humans , Machine Learning , Magnetic Resonance Imaging/methodsABSTRACT
Model-based control has great potential for use in real robots due to its high sampling efficiency. Nevertheless, dealing with physical contacts and generating accurate motions are inevitable for practical robot control tasks, such as precise manipulation. For a real-time, model-based approach, the difficulty of contact-rich tasks that requires precise movement lies in the fact that a model needs to accurately predict forthcoming contact events within a limited length of time rather than detect them afterward with sensors. Therefore, in this study, we investigate whether and how neural network models can learn a task-related model useful enough for model-based control, that is, a model predicting future states, including contact events. To this end, we propose a structured neural network model predictive control (SNN-MPC) method, whose neural network architecture is designed with explicit inertia matrix representation. To train the proposed network, we develop a two-stage modeling procedure for contact-rich dynamics from a limited number of samples. As a contact-rich task, we take up a trackball manipulation task using a physical 3-DoF finger robot. The results showed that the SNN-MPC outperformed MPC with a conventional fully connected network model on the manipulation task.
Subject(s)
Robotics , Learning , Motion , Neural Networks, Computer , Robotics/methodsABSTRACT
Improving human motor performance via physical guidance by an assist robot device is a major field of interest of the society in many different contexts, such as rehabilitation and sports training. In this study, we propose a Bayesian estimation method to predict whether motor performance of a user can be improved or not by the robot guidance from the user's initial skill level. We designed a robot-guided motor training procedure in which subjects were asked to generate a desired circular hand movement. We then evaluated the tracking error between the desired and actual subject's hand movement. Results showed that we were able to predict whether a novel user can reduce the tracking error after the robot-guided training from the user's initial movement performance by checking whether the initial error was larger than a certain threshold, where the threshold was derived by using the proposed Bayesian estimation method. Our proposed approach can potentially help users to decide if they should try a robot-guided training or not without conducting the time-consuming robot-guided movement training.
ABSTRACT
Our brain can be recognized as a network of largely hierarchically organized neural circuits that operate to control specific functions, but when acting in parallel, enable the performance of complex and simultaneous behaviors. Indeed, many of our daily actions require concurrent information processing in sensorimotor, associative, and limbic circuits that are dynamically and hierarchically modulated by sensory information and previous learning. This organization of information processing in biological organisms has served as a major inspiration for artificial intelligence and has helped to create in silico systems capable of matching or even outperforming humans in several specific tasks, including visual recognition and strategy-based games. However, the development of human-like robots that are able to move as quickly as humans and respond flexibly in various situations remains a major challenge and indicates an area where further use of parallel and hierarchical architectures may hold promise. In this article we review several important neural and behavioral mechanisms organizing hierarchical and predictive processing for the acquisition and realization of flexible behavioral control. Then, inspired by the organizational features of brain circuits, we introduce a multi-timescale parallel and hierarchical learning framework for the realization of versatile and agile movement in humanoid robots.
Subject(s)
Artificial Intelligence , Robotics , Behavior Control , Computer Simulation , Humans , LearningABSTRACT
Machine learning classifiers for psychiatric disorders using resting-state functional magnetic resonance imaging (rs-fMRI) have recently attracted attention as a method for directly examining relationships between neural circuits and psychiatric disorders. To develop accurate and generalizable classifiers, we compiled a large-scale, multi-site, multi-disorder neuroimaging database. The database comprises resting-state fMRI and structural images of the brain from 993 patients and 1,421 healthy individuals, as well as demographic information such as age, sex, and clinical rating scales. To harmonize the multi-site data, nine healthy participants ("traveling subjects") visited the sites from which the above datasets were obtained and underwent neuroimaging with 12 scanners. All participants consented to having their data shared and analyzed at multiple medical and research institutions participating in the project, and 706 patients and 1,122 healthy individuals consented to having their data disclosed. Finally, we have published four datasets: 1) the SRPBS Multi-disorder Connectivity Dataset 2), the SRPBS Multi-disorder MRI Dataset (restricted), 3) the SRPBS Multi-disorder MRI Dataset (unrestricted), and 4) the SRPBS Traveling Subject MRI Dataset.
Subject(s)
Brain/diagnostic imaging , Databases, Factual , Magnetic Resonance Imaging , Mental Disorders/diagnostic imaging , Neuroimaging , Adult , Female , Humans , Machine Learning , Male , Middle Aged , Young AdultABSTRACT
BACKGROUND: The major surgical treatment for Parkinson's disease (PD) is deep brain stimulation (DBS), but a less invasive treatment is desired. Vagus nerve stimulation (VNS) is a relatively safe treatment without cerebral invasiveness. In this study, we developed a wireless controllable electrical stimulator to examine the efficacy of VNS on PD model rats. METHODS: Adult female Sprague-Dawley rats underwent placement of a cuff-type electrode and stimulator on the vagus nerve. Following which, 6-hydroxydopamine (6-OHDA) was administered into the left striatum to prepare a PD model. VNS was started immediately after 6-OHDA administration and continued for 14 days. We evaluated the therapeutic effects of VNS with behavioral and immunohistochemical outcome assays under different stimulation intensity (0.1, 0.25, 0.5 and 1 mA). RESULTS: VNS with 0.25-0.5 mA intensity remarkably improved behavioral impairment, preserved dopamine neurons, reduced inflammatory glial cells, and increased noradrenergic neurons. On the other hand, VNS with 0.1 mA and 1 mA intensity did not display significant therapeutic efficacy. CONCLUSIONS: VNS with 0.25-0.5 mA intensity has anti-inflammatory and neuroprotective effects on PD model rats induced by 6-OHDA administration. In addition, we were able to confirm the practicality and effectiveness of the new experimental device.
ABSTRACT
Parkinson's disease (PD) patients often suffer from spinal diseases requiring surgeries, although the risk of complications is high. There are few reports on outcomes after spinal surgery for PD patients with deep brain stimulation (DBS). The objective of this study was to explore the data on spinal surgery for PD patients with precedent DBS. We evaluated 24 consecutive PD patients with 28 spinal surgeries from 2007 to 2017 who received at least a 2-year follow-up. The characteristics and outcomes of PD patients after spinal surgery were compared to those of 156 non-PD patients with degenerative spinal diseases treated in 2013-2017. Then, the characteristics, outcomes, and spinal alignment of PD patients receiving DBS were analyzed in degenerative spinal/lumbar diseases. The mean age at the time of spinal surgery was 68 years. The Hoehn and Yahr score regarding PD was stage 1 for 8 patients, stage 2 for 2 patients, stage 3 for 8 patients, stage 4 for 10 patients, and stage 5 for 0 patient. The median preoperative L-DOPA equivalent daily dose was 410 mg. Thirteen patients (46%) received precedent subthalamic nucleus (STN) DBS. Lumbar lesions with pain were common, and operation and anesthesia times were long in PD patients. Pain and functional improvement of PD patients persisted for 2 years after surgery with a higher complication rate than for non-PD patients. PD patients with STN DBS maintained better lumbar lordosis for 2 years after spinal surgery. STN DBS significantly maintained spinal alignment with subsequent pain and functional amelioration 2 years after surgery. The outcomes of spinal surgery for PD patients might be favorably affected by thorough treatment for PD including DBS.
Subject(s)
Deep Brain Stimulation , Parkinson Disease , Subthalamic Nucleus , Animals , Humans , Pain , Parkinson Disease/therapy , Retrospective Studies , Treatment OutcomeABSTRACT
An amendment to this paper has been published and can be accessed via a link at the top of the paper.
ABSTRACT
Sports trainers often grasp and move trainees' limbs to give instructions on desired movements, and a merit of this passive training is the transferring of instructions via proprioceptive information. However, it remains unclear how passive training affects the proprioceptive system and improves learning. This study examined changes in proprioceptive acuity due to passive training to understand the underlying mechanisms of upper extremity training. Participants passively learned a trajectory of elbow-joint movement as per the instructions of a single-arm upper extremity exoskeleton robot, and the performance of the target movement and proprioceptive acuity were assessed before and after the training. We found that passive training improved both the reproduction performance and proprioceptive acuity. We did not identify a significant transfer of the training effect across arms, suggesting that the learning effect is specific to the joint space. Furthermore, we found a significant improvement in learning performance in another type of movement involving the trained elbow joint. These results suggest that participants form a representation of the target movement in the joint space during the passive training, and intensive use of proprioception improves proprioceptive acuity.