Search

To enable machines to understand human-centric images and videos, they need the capability to detect human–object interactions. This capability has been studied using various approaches, but previous research has mainly focused only on recognition accuracy using widely used open datasets. Given the need for advanced machine-learning systems that provide spatial analysis and services, the recognition model should be robust to various changes, have high extensibility, and provide sufficient recognition speed even with minimal computational overhead. Therefore, we propose a novel method that combines the skeletal method with object detection to accurately predict a set of $ \langle $ human, verb, object $ \rangle $ triplets in a video frame considering the robustness, extensibility, and lightweight of the model. Training a model with similar perceptual elements to those of humans produces sufficient accuracy for advanced social systems, even with only a small training dataset. The proposed model is trained using only the coordinates of the object and human landmarks, making it robust to various situations and lightweight compared with deep-learning methods. In the experiment, a scenario in which a human is working on a desk is simulated and an algorithm is trained on object-specific interactions. The accuracy of the proposed model was evaluated using various types of datasets.

In recent years, deep learning methodologies have been increasingly applied to the intricate challenges of visual-inertial odometry (VIO), especially in scenarios with rapid movements and scenes lacking clear structure. This paper introduces a novel hybrid approach that leverages the inherent strengths of traditional VIO techniques, while harnessing the potential of advanced machine learning technologies. By seamlessly integrating an iterated extended Kalman filter with deep learning techniques, our approach systematically takes into account uncertainties, thereby enhancing the overall reliability and robustness of the system. The proposed algorithm has been rigorously evaluated on the KITTI and EuroC datasets, outperforming other deep learning VIO methods. It achieved a translation error of 2.28% and a rotation error of 0.226 degrees per 100 meters on the KITTI odometry dataset.

Purpose To investigate the effect of patient positioning on tendinosis grade, visible range, and infraspinatus tendon (IST) thickness, and to determine the feasibility of internal rotation (IR) position to assess IST on ultrasound (US). Materials and Methods This study included 52 shoulders of 48 subjects who were evaluated for IST in three different positions: neutral position (N), IR, and position with the ipsilateral hand on the contralateral shoulder (HC). Two radiologists retrospectively graded IST tendinosis from grade 0 to grade 3 and the visible range from grade 1 to grade 4. The thickness of the IST was measured by another radiologist with a short-axis view. A generalized estimating equation was used for statistical analysis. Results The tendinosis grades were higher in the HC position than in the IR position, with a cumulative odds ratio of 2.087 (p = 0.004, 95% confidence interval [CI]: 1.268–3.433). The tendinosis grades in the HC position (p = 0.370) and IR position (p = 0.146) were not significantly different from those in the N position. The overall difference in IST thickness was significant (p < 0.001), but the visible range (p = 0.530) was not significantly different according to position. Conclusion Patient positioning significantly affected the grade of tendinosis and thickness but not the visible range of the IST. The IR position is a feasible position for assessing the IST on US.

Graph Convolutional Networks (GCNs) have gained widespread adoption in modeling human skeleton sequences for two-person interaction recognition. Most GCN-based models achieve state-of-the-art results by leveraging either intra-body or inter-body connections. However, using only intra-body relations may ignore important interactive features between two individuals, whereas relying on inter-body relations may weaken the specific motion dynamics of each skeleton. To address these shortcomings, we propose a Distinct Motion-Preserving GCN (DMP-GCN) that utilizes intra-body and inter-body graphs to extract interactive features from two human bodies while preserving the distinct motion characteristics of each skeleton. Specifically, two motion-specific streams are adopted to capture specific motion features of each human skeleton and an interactive stream is applied to model the interactive dynamics of two bodies. In addition, we introduce a new graph labeling strategy, namely Distance Variation Labeling, which is a data-driven approach for defining the edge strength in the skeleton graph. Extensive experiments show that our proposed approaches outperform state-of-the-art methods on two large-scale human interaction datasets, NTU RGB+D (mutual) and NTU RGB+D 120 (mutual).

3D object detection and localization based on only a monocular camera always faces its fundamental ill-posed issue to estimate 3D information. In combination with deep neural networks, recent researches have shown encouraging results to tackle this issue. However, most of them are only applied to street-view cameras based on several available small-size datasets and the 3D prediction accuracy of these methods is still low in comparison to traditional estimation methods using stereo-cameras. With the development of drone delivery applications in city spaces, it is also necessary to have a similar method to detect objects and estimate their 3D position from an aerial view. We proposed a novel Simulation to Reality approach to predict the object’s 3D position from an aerial view. An instance segmentation of an object is used as an intermediate representation not only to create a very large dataset for training by simulation but also to minimize the gap between simulation and reality. We designed a feed-forward neural network to predict the 3D position from instance segmentation and integrated it with a range-attention classification to improve accuracy, especially for 3D object detection at far distances. To evaluate our methods, we created two simulation datasets: one for cross-validation with other state-of-the-art methods and the other one for practical experiments on a real drone with a monocular camera. The experiment’s results demonstrate that we not only achieve better accuracy than the state-of-the-art methods using the monocular camera by testing on the same KITTI-3D dataset but also reach close to the accuracy of a stereo-based technique. Since our model is lightweight, we successfully deployed it on a companion computer on the real drone and the results of practical experiences are promising.

This paper explores the distribution and features of residual stresses formed by super duplex stainless steel pipe welding. Experimental investigations, which encompass an elevated temperature tensile test and metallographic observation along with a hardness test and residual stress measurement, were first conducted to obtain the mechanical properties at high base metal temperatures and to confirm whether or not the duplex stainless steel undergoes martensitic phase development during the welding process. Finally, experiments were performed to scrutinize the residual stress evolution through the metallurgical phase transformation in the weld region and its vicinity. A sequentially coupled 3D thermal, mechanical and metallurgical finite element (FE) model capable of incorporating the experimental consequences was established next. A 3D FE simulation of the girth welding process was conducted, and the axial and hoop residual stress profiles along the girth were evaluated. The results substantiate that martensitic phase evolution occurs in the process of cooling during the welding of super duplex stainless steel, and they also highlight the significance of taking the metallurgical phase transformation into account in the numerical reproduction of the girth welding process for the accurate expression of weld-induced residual stresses, which is especially important for precisely predicting hoop residual stresses.

Traditional models for various machine learning problems such as image classification perform well only under the assumption of a closed set. This implies that inputs must belong to the classes for which the models were trained. Data collected in the real world may not belong to any finite set of classes, and training a model with an infinite number of classes would obviously be impossible. Rather than incorrectly classifying outlier inputs that belong to unknown classes as members of one of the classes on which the model was trained, learning models must recognize and reject such data samples, or request human assistance in labeling them. For example, when a self-driving vehicle detects unfamiliar scenes or objects, it must notify the driver and hand over control. Various models have been proposed to address the open-set problem. However, the existing models are generally complex or use complicated techniques such as generative adversarial networks and auto-encoders, and their efficiency is not proportional to their complexity. In this study, we propose Triple-Sigmoid as a simple activation function comprised of three Sigmoid functions. Using Triple-Sigmoid in the last activation layer of any deep neural network model enables the model to recognize outliers. Although Triple-Sigmoid can be applied to a variety of machine learning problems, including semi-supervised learning, active learning, and incremental learning, we only investigated the open-set recognition problem in this work. The results of numerous experiments are presented to validate that substituting Triple-Sigmoid for conventional activations such as Softmax and Sigmoid not only retained performance in closed-set settings, but also significantly improved performance in open-set configurations. Furthermore, these results demonstrate that Triple-Sigmoid significantly outperformed existing state-of-the-art methods. Source code for the proposed model is available at https://github.com/dinhtuantran/triple-sigmoid.

PurposeThe discrepancy between preoperative and final pathological staging has been a long-standing challenge for the application of clinical trials or appropriate treatment options. This study aimed to demonstrate the accuracy of preoperative staging of locally advanced gastric cancer using data from a large-scale randomized clinical trial.Materials and methodsOf the 1050 patients enrolled in the clinical trial, 26 were excluded due to withdrawal of consent (n = 20) or non-surgery (n = 6). The clinical and pathological staging was compared. Risk factor analysis for underestimation was performed using univariate and multivariate analyses.ResultsRegarding T staging by computed tomography, accuracy rates were 74.48, 61.62, 58.56, and 85.16% for T1, T2, T3 and T4a, respectively. Multivariate analysis for underestimation of T staging revealed that younger age, ulcerative gross type, circular location, larger tumor size, and undifferentiated histology were independent risk factors. Regarding nodal status estimation, 54.9% of patients with clinical N0 disease were pathologic N0, and 36.4% of patients were revealed to have pathologic N0 among clinical node-positive patients. The percentage of metastasis involvement at the D1, D1+, and D2 lymph node stations significantly increased with the advanced clinical N stage. Among all patients, 29 (2.8%), including 26 with peritoneal seeding, exhibited distant metastases.ConclusionsEstimating the exact pathologic staging remains challenging. A thorough evaluation is mandatory before treatment selection or trial enrollment. Moreover, we need to set a sufficient case number when we design the clinical trial considering the stage migration.

Night security is known for its long hours and heavy tasks. In Japan, a labor shortage of security guards has become an issue in recent years. To solve these problems, an increasing number of robotic security methods are being used. However, several problems exist with existing security robots. For example, wheeled robots traveling on the ground have difficulty in dealing with obstacles such as steps, while most drones are only for monitoring and do not have a function to help people. In this study, an aerial ubiquitous display (AUD) night security drone has been developed to solve the problems of existing security robots. The AUD is equipped with an infrared camera and a projector to detect human behavior at night and present information to people in need. In this paper, an experiment was conducted with the AUD to evaluate whether it can provide adequate nighttime security. In the experiment, real-time monitoring and information projection from the air were achieved. In addition, new security methods using the AUD were shown to be effective. Replacing security guards with the AUD to provide security at night will improve labor shortages in the future, and better security methods will be developed.

Abstract As the geriatric population expands, caregivers require more accurate training to handle and care for the elderly. However, students lack methods for acquiring the necessary skill and experience, as well as sufficient opportunities to practice on real human beings. To investigate the necessity and feasibility of care training assistant robots in care education, we developed a simulated robot as a shoulder complex joint with multi-DOF. In this study, five experts with years of experience in elderly care participated in the data-acquisition process, to acquire information on aspects such as the glenohumeral angle, as well as the sterno-clavicular joint and its torque. The experts performed three types of range-of-motion exercises: (i) elevation–depression of sterno-clavicular joint; (ii) extension and flexion of glenohumeral joint; (iii) lateral and medial rotation of glenohumeral joint. The experimental results showed that the quantitative results for all the exercises were significantly different between the experts. Moreover, we observed that even experienced professionals need consistent care education based on quantitative data and feedback. Thus, we confirmed the necessity and feasibility of the care training assistant robot for improving the skills required for elderly care.

Background/Aims Endoscopic ultrasonography (EUS) is the most efficient imaging modality for gastric subepithelial tumors (SETs). However, abdominopelvic computed tomography (APCT) has other advantages in evaluating the characteristics, local extension, or invasion of SETs to adjacent organs. This study aimed to compare the diagnostic ability of EUS and APCT based on surgical histopathology results. Methods We retrospectively reviewed data from 53 patients who underwent both EUS and APCT before laparoscopic wedge resection for gastric SETs from January 2010 to December 2017 at a single institution. On the basis of histopathology results, we assessed the diagnostic ability of the 2 tests. Results The overall accuracy of EUS and APCT was 64.2% and 50.9%, respectively. In particular, the accuracy of EUS vs. APCT for the diagnosis of gastrointestinal stromal tumors (GISTs), leiomyomas, and ectopic pancreas was 83.9% vs. 74.2%, 37.5% vs. 0.0%, and 57.1% vs. 14.3%, respectively. Most of the incorrect diagnoses with EUS involved hypoechoic lesions originating in the fourth echolayer, with the most common misdiagnosed lesions being GISTs mistaken for leiomyomas and vice versa. Conclusions APCT showed a lower overall accuracy than EUS; however, APCT remains a useful modality for malignant/potentially malignant gastric SETs.

As the elderly population increases, the importance of the caregiver’s role in the quality of life of the elderly has increased. To achieve effective feedback in terms of care and nursing education, it is important to design a robot that can express emotions or feel pain like an actual human through visual-based feedback. This study proposes a care training assistant robot (CaTARo) system with 3D facial pain expression that simulates an elderly person for improving the skills of workers in elderly care. First, in order to develop an accurate and efficient system for elderly care training, this study introduces a fuzzy logic–based care training evaluation method that can calculate the pain level of a robot for giving the feedback. Elderly caregivers and trainees performed the range of motion exercise using the proposed CaTARo. We obtained quantitative data from CaTARo, and the pain level was calculated by combining four key parameters using the fuzzy logic method. Second, we developed a 3D facial avatar for use in CaTARo that is capable of expressing pain based on the UNBC-McMaster Pain Shoulder Archive, and we then generated four pain groups with respect to the pain level. To mimic the conditions for care training with actual humans, we designed the system to provide pain feedback based on the opinions of experts. The pain feedback was expressed in real time by using a projector and a 3D facial mask during care training. The results of the study confirmed the feasibility of utilizing a care training robot with pain expression for elderly care training, and it is concluded that the proposed approach may be used to improve caregiving and nursing skills upon further research.

In a wearable robot arm, the minimum joint configuration and link length must be considered to avoid increasing the burden on the user. This work investigated how the joint configuration, length of arm links, and mounting position of a wearable robot arm affect the cooperative and invasive workspaces of the overall workspace. We considered the joint configurations and link lengths of passive and active joints in our proposed wearable robot arm, which is called the Assist Oriented Arm (AOA). In addition, we comprehensively studied the position of the arm on the user. As a result, three locations around the shoulders and two around the waist were chosen as potential mounting sites. Furthermore, we evaluated the weight burden when the user mounted the wearable robot arm at those positions.

In this paper, we present robust methods for automatically segmenting phases in a specified surgical workflow by using latent Dirichlet allocation (LDA) and hidden Markov model (HMM) approaches. More specifically, our goal is to output an appropriate phase label for each given time point of a surgical workflow in an operating room. The fundamental idea behind our work lies in constructing an HMM based on observed values obtained via an LDA topic model covering optical flow motion features of general working contexts, including medical staff, equipment, and materials. We have an awareness of such working contexts by using multiple synchronized cameras to capture the surgical workflow. Further, we validate the robustness of our methods by conducting experiments involving up to 12 phases of surgical workflows with the average length of each surgical workflow being 12.8 minutes. The maximum average accuracy achieved after applying leave-one-out cross-validation was 84.4%, which we found to be a very promising result.

Laparoscopic gastrectomy is evolving. With the increasing expertise and experience of oncologic surgeons in the minimally invasive surgery for gastric cancer, the indication for laparoscopic gastrectomy is expanding to advanced cases. Many studies have demonstrated the benefits of minimally invasive surgery, including reduced risk of surgery-related injury, reduced blood loss, less pain, and earlier recovery. In order to establish concrete evidence for the suitability of minimal invasive surgery for gastric cancer, many multicenter RCTs, comparing the short- and long-term outcomes of laparoscopic versus open surgery, are in progress. Advances in laparoscopic gastrectomy are moving toward increasingly minimally invasive approaches that enable the improvement of the quality of life of patients, without compromising on oncologic safety.

In this paper we propose a new method to evaluate the effects of nationwide conservation voltage reduction (CVR) on peak-load shaving, using substation operating results management system (SOMAS) data. Its evaluation is based on a national CVR factor, which is defined as the weighted average of CVR factors associated with all transformer banks and weighting coefficients are determined by the reconstructed loads corresponding to each transformer bank. To make use of the data resulting from nationwide CVR without installing additional measuring devices, we adopt a linearized static-load model with a linearizing parameter. SOMAS data are used to evaluate the effects of nationwide CVR on peak-load shaving in the Korean power system. Evaluation results show that the national CVR factor of the Korean power system has small values in the summer season and large values in the winter season. This means that the effect of nationwide CVR on peak-load shaving in the Korean power system presents stronger benefits during winter months.

This paper proposes a linearized load model to evaluate the effect of conservation voltage reduction at a national level. In this model, the respective active and reactive linearizing parameters for active and reactive loads in a power system are estimated using energy management system (EMS) data resulting from conservation voltage reductions. To verify the validity of the linearized load model, PSS/E simulations were conducted for a test power system. Given that conservation voltage reductions are usually executed in the range of 2.0%–5.0%, the proposed model was found to be sufficient to accurately evaluate the effect of conservation voltage reduction. Additionally, Korean EMS data were used to estimate the linearizing parameters for aggregated loads in an actual power system.

Glucosylation of the 21-hydroxyl group of glucocorticoid changes its solubility into hydrophilicity from hydrophobicity and, as with glucocorticoid glucuronides as a moving object in vivo, it is conceivable that it exhibits the same behavior. Therefore, glucosylation to the 21-hydroxyl group while maintaining the 11β-hydroxyl group is particularly important, and glucosylation of corticosterone was confirmed by high-resolution mass spectrometry and 1D (1H and 13C) and 2D (COSY, ROESY, HSQC-DEPT and HMBC) NMR. Moreover, the difference in bioactivity between corticosterone and corticosterone 21-glucoside was investigated in vitro. Corticosterone 21-glucoside showed greater neuroprotective effects against H2O2-induced cell death and reactive oxygen species (ROS) compared with corticosterone. These results for the first time demonstrate that bioconversion of corticosterone through the region-selective glucosylation of a novel compound can present structural potential for developing new neuroprotective agents.

In this paper, we propose a novel interactive multi-resolution display system. The proposed system is based on a projector-mounted mobile robot in an Intelligent Space. The Intelligent Space can calculate the location of the region of interest (ROI) by recognizing the user's pointing gesture. The steerable projector mounted on the mobile robot can improve the brightness and resolution of the ROI of a large image projected by a stationary projector installed in the Intelligent Space. In the proposed system, the user is not required to hold any apparatuses for interacting with the display. Additionally, the proposed system is easy to use because it is designed with the natural and intuitive hand movement of user in mind. In the experiments, we demonstrate the feasibility of the proposed system.

Purpose: This study aimed to evaluate the impact of facilitating target delineation of continuous positive airway pressure (CPAP) in patients undergoing stereotactic ablative radiation therapy (SABR) for lung tumors by lung expansion and respiratory motion management. Materials and Methods: We performed a prospective single-institutional trial of patients who were diagnosed with either primary lung cancer or lung metastases and received SABR with a dose of 40 to 60 Gy in 4 fractions. Four-dimensional computed tomography simulations were conducted for each patient: once without CPAP and again with CPAP. Results: Thirty-two patients with 39 tumors were analyzed, after the withdrawal of five patients due to discomfort. For 26 tumors separated from the diaphragm, CPAP significantly increased the superoinferior distance between the tumor and the diaphragm (5.96 cm vs. 8.06 cm; p < 0.001). For 13 tumors located adjacent to the diaphragm, CPAP decreased the overlap of planning target volume (PTV) with the diaphragm significantly (6.32 cm³ vs. 4.09 cm³; p = 0.002). PTV showed a significant reduction with CPAP (25.06 cm³ vs. 22.52 cm³, p = 0.017). In dosimetric analyses, CPAP expanded lung volume by 58.4% with a significant reduction in mean dose and V5 to V40. No more than grade 2 adverse events were reported. Conclusion: This trial demonstrated significant improvement of CPAP in target delineation uncertainties for lung SABR, with dosimetric benefits, a favorable safety profile and tolerability. Further investigation is warranted to explore the role of CPAP as a novel strategy for respiratory motion management. [ABSTRACT FROM AUTHOR]