Training Tasks Research Articles

Design and Implementation of Unity-based Special Unmanned Vehicle Operating Simulator

Special unmanned vehicles can replace personnel to perform military tasks with high risk coefficients, such as long-distance reconnaissance, guided irradiation, and weapons strikes. In order to ensure that the special unmanned vehicle can quickly and accurately complete the established task action, the operator should carry out a large number of trainings in advance for the task content, if the training is directly used in the early stage of the training to manipulate the training, which has the security risks, it’s easy to increase the operator's psychological pressure, and the actual training site can't realistically restore the task scene, the training effect is also single. In view of the above problems, this paper designs a special unmanned vehicle control simulator, which enables the trainers to intuitively understand the task content and key points, and at the same time can control the unmanned vehicle in the virtual task environment to carry out the established task training and result review and analysis, so as to rapidly improve the control level of the trainers and ensure the successful completion of the established task.

Construction and implementation of a laparoscopic skill training course based on a smartphone application and virtual reality

BackgroundTo develop a laparoscopic training course that combines a smartphone application (APP) and virtual reality (VR), and initially evaluate the feasibility and effectiveness of its implementation.MethodsThe Exploring Laparoscopy (Ex-Lap) app was developed to meet training demands. The course was designed by integrating the app with a VR simulator (LapSim®) and animal organ perfusion simulators. From January 2021 to December 2023, 91 participants were enrolled in the study and then divided into 5 separate batches to undergo the first stage of the course. The performance of the participants was evaluated by rating scale, the overall Training and Assessment of Basic Laparoscopic Techniques (TABLT) scores, and pass rates. Statistical analyses were conducted using SPSS 26.0, employing Kruskal-Wallis tests, Chi-squared analysis, and Fisher’s exact test, depending on the data type.ResultsThe Staged Training and Assessment of Laparoscopic Skills (STALS) course was developed, consisting of three stages. The overall pass rates for the first stage across the five batches ranged from 85 to 100%, with no significant difference (P = 0.387). No significant differences were found in the scale scores or TABLT scores for the training tasks among students from different batches (all P > 0.05).ConclusionsThe STALS course is applicable in residency training, demonstrating satisfactory teaching effectiveness and replicability.

Edge-assisted U-shaped split federated learning with privacy-preserving for Internet of Things

In the realm of the Internet of Things (IoT), deploying deep learning models to process data generated or collected by IoT devices is a critical challenge. However, direct data transmission can cause network congestion and inefficient execution, given that IoT devices typically lack computation and communication capabilities. Centralized data processing in data centers is also no longer feasible due to concerns over data privacy and security. To address these challenges, we present an innovative Edge-assisted U-Shaped Split Federated Learning (EUSFL) framework, which harnesses the high-performance capabilities of edge servers to assist IoT devices in model training and optimization process. In this framework, we leverage Federated Learning (FL) to enable data holders to collaboratively train models without sharing their data, thereby enhancing data privacy protection by transmitting only model parameters. Additionally, inspired by Split Learning (SL), we split the neural network into three parts using U-shaped splitting for local training on IoT devices. By exploiting the greater computation capability of edge servers, our framework effectively reduces overall training time and allows IoT devices with varying capabilities to perform training tasks efficiently. Furthermore, we proposed a novel noise mechanism called LabelDP to ensure that data features and labels can securely resist reconstruction attacks, eliminating the risk of privacy leakage. Our theoretical analysis and experimental results demonstrate that EUSFL can be integrated with various aggregation algorithms, maintaining good performance across different computing capabilities of IoT devices, and significantly reducing training time and local computation overhead.

Automated Transformation of Unstructured Cardiovascular Diagnostic Reports into Structured Datasets Using Sequentially Deployed Large Language Models.

Rich data in cardiovascular diagnostic testing are often sequestered in unstructured reports, with the necessity of manual abstraction limiting their use in real-time applications in patient care and research. We developed a two-step process that sequentially deploys generative and interpretative large language models (LLMs; Llama2 70b and Llama2 13b). Using a Llama2 70b model, we generated varying formats of transthoracic echocardiogram (TTE) reports from 3,000 real-world echo reports with paired structured elements, leveraging temporal changes in reporting formats to define the variations. Subsequently, we fine-tuned Llama2 13b using sequentially larger batches of generated echo reports as inputs, to extract data from free-text narratives across 18 clinically relevant echocardiographic fields. This was set up as a prompt-based supervised training task. We evaluated the fine-tuned Llama2 13b model, HeartDx-LM, on several distinct echocardiographic datasets: (i) reports across the different time periods and formats at Yale New Haven Health System (YNHHS), (ii) the Medical Information Mart for Intensive Care (MIMIC) III dataset, and (iii) the MIMIC IV dataset. We used the accuracy of extracted fields and Cohen's Kappa as the metrics and have publicly released the HeartDX-LM model. The HeartDX-LM model was trained on randomly selected 2,000 synthetic echo reports with varying formats and paired structured labels, with a wide range of clinical findings. We identified a lower threshold of 500 annotated reports required for fine-tuning Llama2 13b to achieve stable and consistent performance. At YNHHS, the HeartDx-LM model accurately extracted 69,144 out of 70,032 values (98.7%) across 18 clinical fields from unstructured reports in the test set from contemporary records where paired structured data were also available. In older echo reports where only unstructured reports were available, the model achieved 87.1% accuracy against expert annotations for the same 18 fields for a random sample of 100 reports. Similarly, in expert-annotated external validation sets from MIMIC-IV and MIMIC-III, HeartDx-LM correctly extracted 201 out of 220 available values (91.3%) and 615 out of 707 available values (87.9%), respectively, from 100 randomly chosen and expert annotated echo reports from each set. We developed a novel method using paired large and moderate-sized LLMs to automate the extraction of unstructured echocardiographic reports into tabular datasets. Our approach represents a scalable strategy that transforms unstructured reports into computable elements that can be leveraged to improve cardiovascular care quality and enable research.

Analysis of tasks and training load during pre-season training in professional basketball

IntroductionIn the quest to improve basketball players' performance during the season, the analysis of the training processes during the pre-season is crucial to successfully face the competition period. The aim of this research was to characterize the training tasks of a professional basketball team of the ACB category during the 2022/23 pre-season, as well as to analyze the relationship between the game situation with the pedagogical variables and External Load variables. Materials and MethodsThe sample consisted of 107 tasks during 20 sessions of a Spanish first division professional basketball team. The dependent variables were the pedagogical variables and the external load. The independent variable was the Game Situation, understood as the organization of the players during the tasks. All variables were recorded using the Integral System for the Analysis of Training Tasks (SIATE) tool. A descriptive and inferential analysis was carried out to determine the relationships between the game situation with the pedagogical and external load variables. ResultsAll indicators of the pedagogical and external load variables show a statistically significant association with the game situation (p<.05). ConclusionsTherefore, the pedagogical and external load variables are conditioned by the game situation. Positioning itself as a variable of great relevance for the planning of the sessions, as its design facilitates the achievement of the proposed objectives and more effective training sessions.

Surface Electromyography and Gait Features in Patients after Anterior Cruciate Ligament Reconstruction.

An important reason for the poor recovery of anterior cruciate ligament (ACL) injuries is the poor recovery of muscle function. Therefore, we used surface electromyography (sEMG) and gait analysis to explore the muscle activation patterns and gait characteristics between lower limbs under different exercise states in patients, following anterior cruciate ligament reconstruction (ACLR). Forty-one adults with unilateral ACL injuries in Binzhou Medical University Hospital from October 2022 to June 2023 were allocated to three groups according to the time after ACL reconstruction: group A (≤3 months, 16), group B (3 months-1 year, 13), and group C (>1 year, 12). Patients were tested by sEMG and gait, while straight leg raising (SLR), walking at normal speed, fast walking, and walking up and down the stairs. Two related sample tests were performed for the normalized root mean square (RMS) values and gait parameters. Muscle function changes varied in different training tasks. The RMS value of the involved side was more than the uninvolved side in biceps femoris and semitendinosus of group A (p < 0.010), and for the bilateral rectus femoris (RS), vastus medialis (VM), and vastus lateralis in group B, only the comparison of the RS was significant in group C during fast walking and going up and down the stairs. The ground impact (0.90 [0.63, 1.33] vs. 0.71 [0.43, 1.02], p = 0.035) of the uninvolved side was significantly decreased compared to those of the involved side in patients with ACLR when going down the stairs. Different muscles need to be focused on at different stages of the postoperative period. sEMG and gait analysis can guide the development of a rehabilitation program.

Basolateral amygdala inputs to the nucleus accumbens shell modulate the consolidation of cued-response and inhibitory avoidance learning

The basolateral amygdala (BLA) modulates different types of memory consolidation via distinct projections to downstream brain regions in multiple memory systems. Prior studies indicate that the BLA projects to the nucleus accumbens shell (NAshell) and that these regions interact to influence some types of behavior. Moreover, previous pharmacological work suggests the BLA and NAshell interact to influence memory. However, the precise role of the BLA-NAshell pathway has never been directly investigated in the consolidation of different types of memory including cued-response, spatial, or inhibitory avoidance (IA) learning. To address this, male and female Sprague-Dawley rats received optogenetic manipulations of the BLA or BLA-NAshell pathway immediately following training in different learning tasks. An initial experiment found that optogenetically inhibiting the BLA itself immediately after training impaired cued-response retention in a Barnes maze task in males and females, confirming earlier pharmacological work in males alone. Subsequent experiments found that BLA-NAshell pathway inhibition impaired retention of cued-response and IA learning but had no effect on retention of spatial learning. However, the present work did not observe any effects of pathway stimulation immediately after cued-response or IA learning. Together, the present findings suggest the BLA modulates the consolidation of cued-response and IA, but not spatial, memory consolidation via NAshell projections.

Influence of punishments associated with task outcomes in physical and mental efforts performed by semi-elite soccer players during training tasks

ABSTRACT Punishing application associated with the task outcomes is a familiar strategy used by soccer coaches during training, but it has been mainly remained in the research background. To date, no previous studies have addressed the effects of punishments on soccer players’ physical and mental efforts during on-field training sessions. Voluntarily, 40 semi-elite male soccer players from two Spanish teams of the same club participated in the study (M age = 22.90 yrs, SD = 5.60 yrs), performing two training sessions with the same tasks, but different in the use or non-use of punishments associated with the task outcomes. Physical load was registered using Polar Team and Ratio of Perceived Exertion. Mental load was estimated by the NASA-Task Load Index and mental fatigue was collected using a Visual Analogue Scale. A portable eye tracker was used to seek the effects of punishment manipulation on ocular metrics. After training, players achieved higher values of peak and mean heart rate, peak and mean speed, distance/minute, and number of sprints at the session with punishments, compared to the session without them. The punishments also influenced participants’ eye movement metrics because changes were found in pupil diameter, latency and velocity of saccades, and number of saccades and blinks when compared to the two training sessions. However, no differences were found between sessions for subjective mental load and fatigue variables. These findings have practical implications to the design of soccer training tasks because this emotional constraint influenced the physical load and objective mental load of the soccer players, but not their subjective measures of mental load and fatigue.

Comparing laboratory and online settings: equivalence in training and transfer effects for training task-order coordination processes.

The literature on dual-task training suggests reductions in task-coordination costs with extensive practice, yet such regimens are resource-intensive. This study investigates the feasibility of online assessments for cognitive training studies by comparing training and transfer effects on task-order coordination (TOC) skills in laboratory versus online settings. We conducted a 5-day training regimen including pre-and post-test. Sixty-two participants completed training either in our laboratory or online via Pavlovia. They were assigned to one of two training order conditions, either practicing two visual-manual tasks in a dual-task situation with fixed task order or with random task order. Performance metrics included reaction time (RT) and error rates for trained and untrained tasks to assess TOC costs before and after the training. Data from both setting conditions (laboratory vs. online) were compared. Firstly, data of both settings revealed training-order specific training and transfer effects for TOC costs on RT level. Random task order training improved TOC for trained and untrained tasks, whereas fixed order training did not. Secondly, cross-setting analyses, both frequentists and Bayesian, confirmed these effects and revealed no reliable impact of setting on outcomes. This research carries two significant implications. Our findings demonstrate the acquisition of task-order coordination skills, extending prior research on improving task-coordination in dual-task situations. Additionally, the robust effects for such improvements were independent of specific tasks and setting (whether investigated online or in the laboratory), supporting the use of online testing in cognitive training regimens for resource savings without compromising quality.

21 - Do eight weeks of experience with hypopressive exercise training impact pelvic floor muscle strength, stiffness or task performance? An interventional cohort.

21 - Do eight weeks of experience with hypopressive exercise training impact pelvic floor muscle strength, stiffness or task performance? An interventional cohort.

Multisession tDCS combined with intrastimulation training improves emotion recognition in adolescents with autism spectrum disorder

Previous studies indicate that transcranial direct current stimulation (tDCS) is a promising emerging treatment option for autism spectrum disorder (ASD) and its efficacy could be augmented using concurrent training. However, no intrastimulation social cognition training for ASD has been developed so far. The objective of this two-armed, double-blind, randomized, sham-controlled clinical trial is to investigate the effects of tDCS combined with a newly developed intrastimulation social cognition training on adolescents with ASD. Twenty-two male adolescents with ASD were randomly assigned to receive 10 sessions of either anodal or sham tDCS at F3/right supraorbital region together with online intrastimulation training comprising basic and complex emotion recognition tasks. Using baseline magnetic resonance imaging data, individual electric field distributions were simulated, and brain activation patterns of the training tasks were analyzed. Additionally, questionnaires were administered at baseline and following the intervention. Compared to sham tDCS, anodal tDCS significantly improved dynamic emotion recognition over the course of the sessions. This task also showed the highest activations in face processing regions. Moreover, the improvement was associated with electric field density at the medial prefrontal cortex and social awareness in exploratory analyses. Both groups showed high tolerability and acceptability of tDCS, and significant improvement in overall ASD symptoms. Taken together, multisession tDCS improved dynamic emotion recognition in adolescents with ASD using a task that activates brain regions associated with the social brain network. The variability in the electric field might diminish tDCS effects and future studies should investigate individualized approaches.

Physiological and perceptual responses of firefighters wearing protective clothing under various training environment and activity conditions

To enhance operational capabilities of firefighters in live-fire scenarios, daily training is absolutely important which involves various types of training tasks in diverse thermal environments. Fire training in heat may impose severe heat stress on firefighters. Yet, the thermal burden associating with different training has rarely been studied. In this study, twelve healthy male firefighters performed 50 min of treadmill exercise while wearing firefighting protective clothing (FPC) in two environmental conditions (34°C and 37°C) with three training intensities (3 km/h, 4.5 km/h and 6 km/h). Results showed that in hot environments, training intensity exhibited a more pronounced effect on physiological responses than ambient temperature. Exposure to 37°C with heavy activity produced high level of core temperature (38.8±0.2°C), near maximal heart rate (164±18 bpm), increased energy expenditure (27.10±1.42 ml/min/kg) coupled with mild dehydration (2.08%±0.35%). The forehead, back and chest which showed the highest level of thermal sensation vote, wetness sensation vote and thermal comfort vote are suggested to be primarily considered when designing cooling strategies. Moreover, when providing recovery and cooling for firefighters working in extreme heat while wearing fully encapsulated FPC, the priority should be given to work intensity and duration, followed by environmental conditions.

Meta-Reinforcement Learning in Nonstationary and Nonparametric Environments.

Recent state-of-the-art artificial agents lack the ability to adapt rapidly to new tasks, as they are trained exclusively for specific objectives and require massive amounts of interaction to learn new skills. Meta-reinforcement learning (meta-RL) addresses this challenge by leveraging knowledge learned from training tasks to perform well in previously unseen tasks. However, current meta-RL approaches limit themselves to narrow parametric and stationary task distributions, ignoring qualitative differences and nonstationary changes between tasks that occur in the real world. In this article, we introduce a Task-Inference-based meta-RL algorithm using explicitly parameterized Gaussian variational autoencoders (VAEs) and gated Recurrent units (TIGR), designed for nonparametric and nonstationary environments. We employ a generative model involving a VAE to capture the multimodality of the tasks. We decouple the policy training from the task-inference learning and efficiently train the inference mechanism on the basis of an unsupervised reconstruction objective. We establish a zero-shot adaptation procedure to enable the agent to adapt to nonstationary task changes. We provide a benchmark with qualitatively distinct tasks based on the half-cheetah environment and demonstrate the superior performance of TIGR compared with state-of-the-art meta-RL approaches in terms of sample efficiency (three to ten times faster), asymptotic performance, and applicability in nonparametric and nonstationary environments with zero-shot adaptation. Videos can be viewed at https://videoviewsite.wixsite.com/tigr.

Visual Perceptual Learning of Form-Motion Integration: Exploring the Involved Mechanisms with Transfer Effects and the Equivalent Noise Approach.

Background: Visual perceptual learning plays a crucial role in shaping our understanding of how the human brain integrates visual cues to construct coherent perceptual experiences. The visual system is continually challenged to integrate a multitude of visual cues, including form and motion, to create a unified representation of the surrounding visual scene. This process involves both the processing of local signals and their integration into a coherent global percept. Over the past several decades, researchers have explored the mechanisms underlying this integration, focusing on concepts such as internal noise and sampling efficiency, which pertain to local and global processing, respectively. Objectives and Methods: In this study, we investigated the influence of visual perceptual learning on non-directional motion processing using dynamic Glass patterns (GPs) and modified Random-Dot Kinematograms (mRDKs). We also explored the mechanisms of learning transfer to different stimuli and tasks. Specifically, we aimed to assess whether visual perceptual learning based on illusory directional motion, triggered by form and motion cues (dynamic GPs), transfers to stimuli that elicit comparable illusory motion, such as mRDKs. Additionally, we examined whether training on form and motion coherence thresholds improves internal noise filtering and sampling efficiency. Results: Our results revealed significant learning effects on the trained task, enhancing the perception of dynamic GPs. Furthermore, there was a substantial learning transfer to the non-trained stimulus (mRDKs) and partial transfer to a different task. The data also showed differences in coherence thresholds between dynamic GPs and mRDKs, with GPs showing lower coherence thresholds than mRDKs. Finally, an interaction between visual stimulus type and session for sampling efficiency revealed that the effect of training session on participants' performance varied depending on the type of visual stimulus, with dynamic GPs being influenced differently than mRDKs. Conclusion: These findings highlight the complexity of perceptual learning and suggest that the transfer of learning effects may be influenced by the specific characteristics of both the training stimuli and tasks, providing valuable insights for future research in visual processing.

Learning Effects of Mechanical Ventilator/Tracheal Suctioning XR Simulators and Extracting Decision Making Criteria to Introduce a Novel Simulator

Researchers have developed Simmar+ESTE-SIM, a XR simulator that can train the multitask of “endotracheal suctioning for a patient with a mechanical-ventilator’’ as part of support for Technology-dependent children in the community. In this study, we conducted a questionnaire after using the simulator with the aim of extracting the learning effects of the simulator and judgment indicators when introducing a new simulator, based on feedback from 4th grade university students and the faculty members. As a result, both groups tended to score low on the question of whether the content was appropriate for learning in the third year, but the scores were generally high, at 3.0 or higher. The training task “endotracheal suctioning for a patient with mechanical respirator” was slightly early in 3rd grade students, and the appropriate timing was shown to be for 4th grade students. Principal component analysis extracts two principal components: “Balance between learning content difficulty and motivation (Achievability)” and “Balance between learning time, cost, and learning effect (Feasibility)”. These two factors were thought to be the promoting and inhibiting factors when introducing a simulator. The future challenge is to popularize the simulator education program which is both achievable and feasible to improve the quality of care for children with home-ventilators and their families.

Narrative Review of the Theoretical-Methodological Foundations of the TREINI Program.

Scientific knowledge has advanced in the implementation of safe and beneficial interventions for children and adolescents with cerebral palsy (CP). Although the importance of interdisciplinary interventions that integrate all components of the International Classification of Functioning, Disability and Health (ICF) into family-centered practices is widely recognized, this approach is not yet widely adopted. Instead, many programs remain focused on isolated domains. This study presents the theoretical and methodological foundation of TREINI, an interdisciplinary and family-centered program developed for children and youth with CP and other neurodevelopmental disorders. TREINI incorporates intervention strategies that address all ICF domains. It is grounded in the biopsychosocial model of health and utilizes principles based on the best evidence in pediatric rehabilitation, including intensive training, task-oriented training, and a naturalistic learning environment. Unlike traditional rehabilitation approaches, the care provided by the TREINI program is delivered through an intensive and interdisciplinary approach, by a team working collaboratively in a single location. In addition to including evidence-based interventions, the TREINI program features two innovative components: the "City of Tomorrow", a naturalistic learning environment, and the "TREINI Exoflex" therapeutic suit, specifically designed to address deficiencies in the body functions and structures of children with CP and other neurodevelopmental disorders. This program has been carefully designed to support the process of neurological re-education and rehabilitation for children and adolescents with neuropsychomotor developmental delays.

Enhancing stroke rehabilitation with whole-hand haptic rendering: development and clinical usability evaluation of a novel upper-limb rehabilitation device

IntroductionThere is currently a lack of easy-to-use and effective robotic devices for upper-limb rehabilitation after stroke. Importantly, most current systems lack the provision of somatosensory information that is congruent with the virtual training task. This paper introduces a novel haptic robotic system designed for upper-limb rehabilitation, focusing on enhancing sensorimotor rehabilitation through comprehensive haptic rendering.MethodsWe developed a novel haptic rehabilitation device with a unique combination of degrees of freedom that allows the virtual training of functional reach and grasp tasks, where we use a physics engine-based haptic rendering method to render whole-hand interactions between the patients’ hands and virtual tangible objects. To evaluate the feasibility of our system, we performed a clinical mixed-method usability study with seven patients and seven therapists working in neurorehabilitation. We employed standardized questionnaires to gather quantitative data and performed semi-structured interviews with all participants to gain qualitative insights into the perceived usability and usefulness of our technological solution.ResultsThe device demonstrated ease of use and adaptability to various hand sizes without extensive setup. Therapists and patients reported high satisfaction levels, with the system facilitating engaging and meaningful rehabilitation exercises. Participants provided notably positive feedback, particularly emphasizing the system’s available degrees of freedom and its haptic rendering capabilities. Therapists expressed confidence in the transferability of sensorimotor skills learned with our system to activities of daily living, although further investigation is needed to confirm this.ConclusionThe novel haptic robotic system effectively supports upper-limb rehabilitation post-stroke, offering high-fidelity haptic feedback and engaging training tasks. Its clinical usability, combined with positive feedback from both therapists and patients, underscores its potential to enhance robotic neurorehabilitation.

Objective performance indicators versus GEARS: an opportunity for more accurate assessment of surgical skill.

Surgical skill evaluation that relies on subjective scoring of surgical videos can be time-consuming and inconsistent across raters. We demonstrate differentiated opportunities for objective evaluation to improve surgeon training and performance. Subjective evaluation was performed using the Global evaluative assessment of robotic skills (GEARS) from both expert and crowd raters; whereas, objective evaluation used objective performance indicators (OPIs) derived from da Vinci surgical systems. Classifiers were trained for each evaluation method to distinguish between surgical expertise levels. This study includes one clinical task from a case series of robotic-assisted sleeve gastrectomy procedures performed by a single surgeon, and two training tasks performed by novice and expert surgeons, i.e., surgeons with no experience in robotic-assisted surgery (RAS) and those with more than 500 RAS procedures. When comparing expert and novice skill levels, OPI-based classifier showed significantly higher accuracy than GEARS-based classifier on the more complex dissection task (OPI 0.93 ± 0.08 vs. GEARS 0.67 ± 0.18; 95% CI, 0.16-0.37; p = 0.02), but no significant difference was shown on the simpler suturing task. For the single-surgeon case series, both classifiers performed well when differentiating between early and late group cases with smaller group sizes and larger intervals between groups (OPI 0.9 ± 0.08; GEARS 0.87 ± 0.12; 95% CI, 0.02-0.04; p = 0.67). When increasing the group size to include more cases, thereby having smaller intervals between groups, OPIs demonstrated significantly higher accuracy (OPI 0.97 ± 0.06; GEARS 0.76 ± 0.07; 95% CI, 0.12-0.28; p = 0.004) in differentiating between the early/late cases. Objective methods for skill evaluation in RAS outperform subjective methods when (1) differentiating expertise in a technically challenging training task, and (2) identifying more granular differences along early versus late phases of a surgeon learning curve within a clinical task. Objective methods offer an opportunity for more accessible and scalable skill evaluation in RAS.

DGCL: dual-graph neural networks contrastive learning for molecular property prediction.

In this paper, we propose DGCL, a dual-graph neural networks (GNNs)-based contrastive learning (CL) integrated with mixed molecular fingerprints (MFPs) for molecular property prediction. The DGCL-MFP method contains two stages. In the first pretraining stage, we utilize two different GNNs as encoders to construct CL, rather than using the method of generating enhanced graphs as before. Precisely, DGCL aggregates and enhances features of the same molecule by the Graph Isomorphism Network and the Graph Attention Network, with representations extracted from the same molecule serving as positive samples, and others marked as negative ones. In the downstream tasks training stage, features extracted from the two above pretrained graph networks and the meticulously selected MFPs are concated together to predict molecular properties. Our experiments show that DGCL enhances the performance of existing GNNs by achieving or surpassing the state-of-the-art self-supervised learning models on multiple benchmark datasets. Specifically, DGCL increases the average performance of classification tasks by 3.73$\%$ and improves the performance of regression task Lipo by 0.126. Through ablation studies, we validate the impact of network fusion strategies and MFPs on model performance. In addition, DGCL's predictive performance is further enhanced by weighting different molecular features based on the Extended Connectivity Fingerprint. The code and datasets of DGCL will be made publicly available.

Designing a robot-assisted rehabilitation system for hand function recovery using virtual reality and haptic robot

It is well understood that motor dysfunctions in patients can be restored through a certain amount of repetitive training. Robots and virtual reality offer an excellent solution to incentivize and optimize training intensity while also reducing the workload for therapists. In this paper, leveraging virtual reality technologies and haptic robots, we have developed a cost-effective robot-assisted training system and three training exercises for hand rehabilitation. The Unity game engine was utilized to create training scenarios where hand avatars interact with virtual objects. Off-the-shelf haptic robots were employed to manipulate the hand avatars, enabling completion of training tasks with a realistic tactile experience. Haptic feedback is rendered using impedance control laws. Additionally, to address the needs of both patients and therapists we propose to employ robot-assisted force and feedback force. These mechanisms are designed to provide assistance to patients during tasks if needs and to transmit the patient's force feedback to the therapist, facilitating comprehensive interaction and communication between the two parties. Several experiments were conducted involving six healthy subjects, who used electrode stimulation to simulate impaired limbs, to demonstrate the effectiveness of the proposed training frameworks.