Search

Insulin resistance (IR) is associated with cardiovascular disease in non-diabetic patients. The triglyceride-glucose (TyG) index, incorporating serum glucose and insulin concentrations, is a surrogate insulin resistance marker. We investigated its association with obstructive coronary artery disease (CAD) and sex differences therein. Patients with stable angina pectoris requiring invasive coronary angiography between January 2010 and December 2018 were enrolled. They were divided into two groups according to TyG index. Two interventional cardiologists diagnosed obstructive CAD by angiography review. Demographic characteristics and clinical outcomes were compared between groups. Relative to lower index, patients with higher (≥ 8.60) TyG index had higher BMIs and more prevalent hypertension, diabetes, and elevated lipid profiles [total cholesterol, low-density lipoprotein (LDL), high-density lipoprotein (HDL), triglycerides (TG), fasting plasma glucose (FPG)]. Higher TyG index increased women's obstructive CAD risk after multivariate adjustment (adjusted odds ratio (aOR) 2.15, 95% confidence interval (95% CI) 1.08–4.26, p = 0.02) in non-diabetic populations compared with men. No sex difference was found for diabetic patients. Higher TyG index significantly increased the obstructive CAD risk, overall and for non-diabetic women. Larger-scale studies are needed to confirm our findings. [ABSTRACT FROM AUTHOR]

In this paper, we propose a brain–computer interface (BCI)-based Human-in-the-Loop (Hit-Loop) agent for human and artificial intelligence (AI) colearning in music listening and appreciation with an Artificial Intelligence of Things (AIoT) application. The novel BCI-based Hit-Loop agent contains human intelligence with BCI-based AIoT-Fuzzy Markup Language (FML) and BCI-FML agents, as well as machine intelligence with AI-FML Hit-Loop and AIoT-FML agents. We used FML to facilitate communication between humans and the AI-FML robots through an AIoT-FML Learning Tool (AIoT-FML-LT), which was the core technology of the AI-FML Hit-Loop agent for the BCI-based music listening and appreciation application. Furthermore, the novel AIoT-FML-LT in conjunction with the BCI-FML and AIoT-FML agents was developed and presented for music listening and student learning. Moreover, the BCI-based AIoT-FML and AI-FML Hit-Loop agents were applied in English language learning, and the AIoT-FML-LT assisted in measuring student learning performance in Taiwan and Japan. The human-like high-level knowledge base and rule base were constructed by various domain experts, as well as the personalized electroencephalography (EEG) and student English learning data sets collected using the BCI device and AIoT-FML-LT, respectively, and were applied to machine learning models such as deep learning and particle swarm optimization. Additionally, the AIoT-FML-LT was connected to the AIoT-FML Hit-Loop agent for human and robot colearning. The relationship between human perceptions and the AIoT-FML Hit-Loop agent in terms of eyes, ears, nose, tongue, body, and brain corresponding to sights, sounds, smells, tastes, objects of touch, and mind are discussed. Finally, the students learned human language and AI language by using the AI-FML robots and AIoT-FML-LT together with the human English learning and AI-FML machine learning models, respectively. The experimental results reveal that the BCI-based Hit-Loop agent for human and AI-FML robot colearning in conjunction with AIoT applications can effectively facilitate music listening and appreciation application as well as English listening in Taiwan and Japan. The learning behavior and performance of the students also improved after incorporation of the human and robot colearning model. [ABSTRACT FROM AUTHOR]

This paper proposes an Adaptive Fuzzy Neural Agent (AFNA) with a Patch Learning Mechanism and IEEE 1855 Fuzzy Markup Language (FML) for human and machine co-learning. There are three phases of patch learning mechanism embedded in AFNA, including (1) training an initial global model, (2) training a patch model for each identified patch, and (3) updating the global model using the training data that do not fall into any patch. The AFNA can be applied to construct the student and robot co-learning regression model, as well as the regression model for the dataset retrieved from the game of Go. First, students generate human learning data through interactions with handheld devices or robots based on the AFNA in Taiwan and Japan. Then, the AFNA utilizes the student learning data collected in the classroom and the Go game data provided by both Google DeepMind and Facebook AI Research open-source OpenGo to train the Fuzzy Machine-Learning Model. In addition, the trained Fuzzy Machine-Learning Model of AFNA is deployed to the robots to make students and machines co-learn together based on IEEE 1855 FML. The experiments show that the AFNA with Patch Learning Mechanism and Fuzzy Machine-Learning Model can improve the performance of regression model based on the datasets of student learning and Go game. In the future, we hope to apply the AFNA with robots to the other domain areas, embed it with the Artificial Intelligence of Things devices, and introduce it to more teaching fields in various countries. [ABSTRACT FROM AUTHOR]

Activin A, a cytokine belonging to the transforming growth factor-β family, has been shown to play pivotal roles in tissue remodeling after renal injury and is present in elevated levels in diabetic patients. However, the association between activin A and albuminuria remains unclear. We aimed to evaluate their association by using cross-sectional data from community-dwelling middle-aged and older adults in Taiwan. We assessed 466 participants (67% male; mean age 71 ± 13 years) from the I-Lan Longitudinal Aging study for whom data pertaining to serum activin A level and urine albumin-to-creatinine ratio (UACR) were available. Of these, 323 (69%) had normal albuminuria, 123 (26%) had microalbuminuria, and 20 (4%) had overt albuminuria. Patients with overt albuminuria and microalbuminuria had significantly higher activin A concentrations than those in the normal albuminuria group (p < 0.001). Circulating activin A was significantly correlated with multiple risk factors, including higher systolic blood pressure and higher UACR. Univariate and multivariate results indicated that activin A level was an independent variable for albuminuria. The cutoff value of 602 pg/mL of activin A demonstrated a sensitivity of 70.6% and specificity of 75.7% (AUC 0.774) in diagnosing overt albuminuria. In conclusion, middle-aged and older adults with elevated activin A levels were associated with a higher incidence of albuminuria. [ABSTRACT FROM AUTHOR]

Emergency evacuation is concerning to both industry and academics. This work applied the technology of a wireless sensor network to design and propose a nonlinear mathematical model and heuristic approach to fire emergency evacuation problem. The proposed model first considers basic factors including the distribution of people in each area, possible safe exit locations, possible hazard locations, rooms, open spaces, doors, and junctions. Based on these common factors and the proposed heavy smoke diffusion, the objective of the proposed model is to evacuate all people in the shortest time through the concept of load balance coming from the relationship between crowd density and moving speed. The algorithm of simulated annealing is finally applied to solve the evacuation planning problem derived from the proposed model. Experimental results verify the efficiency of the proposed model and the resulting solution. [ABSTRACT FROM AUTHOR]

Inherently, piezoelectric actuator is one of the devices equipped with the micrometer positioning capability and characterized by small size, fast response, high stiffness, and large blocking force. These advantages give piezoelectric actuator the possibility of being high-accuracy industrial machineries. However, factors of nonlinear hysteresis, modeling uncertainties, and environmental disturbances result in unacceptable positioning errors and greatly increase the control difficulties. In this paper, a hybrid nonlinear robust control design that integrates a feedback linearization control method and a robust compensator is proposed. It aims to eliminate above-mentioned impacts and tackle the micrometer (μm) positioning design of piezoelectric actuators. The feedback linearization controller is developed for converging of positioning errors exponentially. The robust compensator is used to mitigate the total impact of hysteresis, modeling uncertainties and environmental disturbances and carry out the fine tuning of positioning errors to zero. Simulation results and practical tests reveal that the controlled piezoelectric actuator reaches 1 μm positioning accuracy, and the proposed robust control law delivers promising positioning performance under impacts of hysteresis, modeling uncertainties and environmental disturbances. [ABSTRACT FROM AUTHOR]

Dynamic assessment with an intelligent agent can differentiate the capabilities and proficiency of students. It can therefore be advocated as an interactive approach to conduct assessments on students in learning systems. Facebook AI Research proposed ELF OpenGo, an open-source reimplementation of the AlphaZero algorithm. They also developed Darkforest, which displays the competence and skills of high-level amateur Go players. To enable these open-source AI bots to assist humans at different levels in learning Go, this paper proposes an intelligent agent for real-world applications in robotic edutainment and humanized co-learning. To achieve this, we successfully constructed an OpenGo Darkforest (OGD) cloud platform using these AI bots and further combined the brain computer interface with the OGD cloud platform to observe the relationship between the brainwaves and win rates of human Go players. The intelligent agent also converted human brainwaves into physiological indices and reflected these in the robot to express human feelings or emotions in real-time. For future educational applications, this paper also presents intelligent robot teachers learning together with students in Taiwan and Japan. More than 200 students have been co-learning with intelligent robot teachers in Tainan, Kaohsiung, Taipei, and Tokyo from 2018 to 2019. The learning performance and feedback from students and teachers has been extremely positive, especially from remedial students. [ABSTRACT FROM AUTHOR]

Prevention for contrast-induced nephropathy (CIN) is limited by the lack of a single predictor. As activin A is upregulated in heart failure and chronic kidney disease, we aimed to clarify the association between activin A levels and renal outcomes after coronary angiography (CAG). This prospective observational study included 267 patients who received CAG between 2009 and 2015. CIN was defined as elevation of serum creatinine to >0.5 mg/dL or to >25% above baseline within 48 hours after CAG. During follow-up, laboratory parameters were measured every 3–6 months. Renal decline was defined as>2-fold increase in serum creatinine or initiation of dialysis. The patients were stratified into tertiles according to serum activin A levels at baseline. High activin A tertile was significantly associated more CIN and renal function decline compared to low activin A tertile (all p < 0.001). After adjusting potential confounding factors, high serum activin A tertiles was associated to CIN (Odds ratio 4.49, 95% CI 1.07–18.86, p = 0.040) and renal function decline (Hazard ratio 4.49, 95% CI 1.27–11.41, p = 0.017) after CAG. [ABSTRACT FROM AUTHOR]

Galectin-1, a β-galactoside-binding lectin mediating inflammation and neovascularization, is reported to attenuate ventricular remodeling after myocardial infarction. But its role in stable coronary artery disease (CAD) has not been fully elucidated. This study aimed to identify the relationship between the circulating galectin-1 level and the severity of CAD in patients with suspected CAD. Pre-procedure galectin-1 and high-sensitivity C-reactive protein (hs-CRP) concentrations were measured in 834 subjects who underwent scheduled coronary angiography. Subjects were grouped into tertiles of the galectin-1 levels. SYNTAX scores were calculated to evaluate the severity of CAD. All patients were followed until January 2019 or the occurrence of major adverse cardiovascular events (MACE). Patients with higher galectin-1 concentrations were older; had greater prevalence of hypertension, diabetes, chronic kidney disease, and heart failure; and were more likely to present with higher hs-CRP levels and SYNTAX scores. During the follow-up period of 1.3 ± 1.1 years, patients in the highest tertile of galectin-1 were associated with a greater risk of MACE after adjustment for age, sex, comorbidities, co-medications, serum levels of hemoglobin, creatinine, hs-CRP, ejection fraction, SYNTAX scores, and revascularization modalities (adjusted hazard ratio 10.95, 95% confidence interval 2.29–52.47, p = 0.003). Galectin-1 showed better discriminatory performance than hs-CRP, and non-inferior performance to SYNTAX scores, in predicting the incidence of MACE. [ABSTRACT FROM AUTHOR]

An amendment to this paper has been published and can be accessed via a link at the top of the paper. [ABSTRACT FROM AUTHOR]