Search

Passive acoustic monitoring combined with deep learning-based bird sound classifiers is an effective tool, particularly in remote areas. While self-supervised learning has recently excelled in natural language processing and image recognition, its application to bird sound recognition remains limited. This study proposes an innovative self-supervised learning approach, which leverages vast amounts of passive acoustic recordings for pre-training, followed by fine-tuning of target species. Compared to the three state-of-the-art models based on transfer learning from ImageNet, the proposed method demonstrated improvements in overall recognition performance, with even more significant gains for tail-end species. These results confirm that domain-specific pre-training in self-supervised learning enhances downstream recognition tasks and provides greater robustness, benefiting tail-end species in imbalanced ecological datasets. Our experiments further demonstrate that integrating open-source datasets and data augmentation techniques is the most effective strategy for mitigating data imbalances and cross-domain issues. In addition, introducing a ‘catch-all’ category into training datasets has been shown to improve model robustness in open set recognition scenarios. We also identified the minimum viable sample size requirements for our proposed model and explored the impact of overlapping bird vocalizations during dawn choruses on model performance. Targeting 31 bird species in the montane regions of subtropical Taiwan, the model achieved a class-wise mean average precision of 0.782 and an overall precision of 85.6 % at the F0.5 threshold in dawn chorus soundscape recordings. This study confirms the effectiveness and advantages of self-supervised learning in bird sound recognition, supporting long-term monitoring of bird distribution and vocal activity in remote montane areas.

In the field of automatic optical inspection (AOI), this study presents innovative strategies to enhance object detection accuracy while minimizing dependence on large annotated datasets. We initially developed a defect detection model using a dataset of 3579 images across 32 categories, created in collaboration with a major Taiwanese panel manufacturer. This model was evaluated using 12,000 ambiguously labeled images, with improvements achieved through data augmentation and annotation refinement. To address the challenges of limited labeled data, we proposed the Adaptive Fused Semi-Supervised Self-Learning (AFSL) method. This approach, designed for anchor-based object detection models, leverages a small set of labeled data alongside a larger pool of unlabeled data to enable continuous model optimization. Key components of AFSL include the Bounding Box Assigner, Adaptive Training Scheduler, and Data Allocator, which together facilitate dynamic threshold adjustments and balanced training, significantly enhancing the model’s performance on AOI datasets. The AFSL method improved the mean average precision (mAP) from 43.5% to 57.1% on the COCO dataset and by 2.6% on the AOI dataset, demonstrating its effectiveness in achieving high levels of precision and efficiency in AOI with minimal labeled data.

Abstract This study evaluated the aerosol exposure risks while using common noninvasive oxygenation devices. A simulated mannequin was designed to breathe at a minute ventilation of 20 L/min and used the following oxygen-therapy devices: nasal cannula oxygenation (NCO) at 4 and 15 L/min, nonrebreathing mask (NRM) at 15 L/min, simple mask at 6 L/min, combination of NCO at 15 L/min and NRM at 15 L/min, high-flow nasal cannula (HFNC) at 50 L/min, and flush rate NRM. Two-dimension of the dispersion distance and the aerosol concentrations were measured at head, trunk, and foot around the mannequin for over 10 min. HFNC and flush-rate NRM yielded the longest dispersion distance and highest aerosol concentrations over the three sites of the mannequin than the other oxygenation devices and should use with caution. For flow rates of

Abstract To evaluate aerosol exposure risk and prevention strategies during bystander, pre-hospital, and inpatient cardiopulmonary resuscitation (CPR). This study compared hands-only CPR, CPR with a surgical or N95 mask, and CPR with a non-rebreather mask at 15 L/min. 30:2 compression–ventilation ratio CPR was tested with face-mask ventilation (FMV), FMV with a high efficiency particulate air (HEPA) filter; supraglottic airway (SGA), SGA with a surgical mask, SGA with a HEPA filter, or SGA with both. Continuous CPR was tested with an endotracheal tube (ET), ET with a surgical mask, a HEPA filter, or both. Aerosol concentration at the head, trunk, and feet of the mannequin were measured to evaluate exposure to CPR personnel. Hands-only CPR with a surgical or N95 face mask coverings and ET tube ventilation CPR with filters showed the lowest aerosol exposure among all study groups, including CPR with NRM oxygenation, FMV, and SGA ventilation. NRM had a mask effect and reduced aerosol exposure at the head, trunk, and feet of the mannequin. FMV with filters during 30:2 CPR reduced aerosol exposure at the head and trunk, but increased at the feet of the mannequin. A tightly-sealed SGA when used with a HEPA filter, reduced aerosol exposure by 21.00–63.14% compared with a loose-fitting one. Hands-only CPR with a proper fit surgical or N95 face mask coverings is as safe as ET tube ventilation CPR with filters, compared with CPR with NRM, FMV, and SGA. FMV or tight-sealed SGA ventilation with filters prolonged the duration to achieve estimated infective dose of SARS-CoV-2 2.4–2.5 times longer than hands-on CPR only. However, a loose-fitting SGA is not protective at all to chest compressor or health workers standing at the foot side of the victim, so should be used with caution even when using with HEPA filters.

We disclosed HfO2-based dielectric of superb electrical properties on p-type Si0.8Ge0.2 substrate using an interfacial layer (IL) formed by trimethylaluminum (TMA) pre-treatment and Al scavenger. Our results revealed that the interface trap density (Dit) value and the gate leakage current (JG) could be improved about 60 times and 100 times by tuning the gate electrode composition without sacrificing equivalent oxide thickness (EOT) performance. The mechanism underlying the ${\mathrm{ D}}_{\mathrm{ it}}$ improvement of the SiGe metal-oxide-semiconductor capacitors (MOSCAPs) might be owing to the Al metal scavenger and the minimization of the oxygen atoms diffusing to the high- $\kappa $ /SiGe IL, verified by x-ray photoelectron spectroscopy (XPS) analyses. In addition, the hysteresis levels of SiGe capacitors with various gate electrodes were measured to find out the optimized configuration of metal electrodes. This work demonstrated the Al scavenger effect from the aspects of both material and electrical properties and achieved an impressive EOT value of $\sim 0.8$ nm for the capacitors fabricated on the SiGe substrate.

Abstract Bioactivity-guided phytochemical investigation of Podocarpus neriifolius D. Don. (Podocarpaceae) has led to the isolation of one new (2) and three known (1, 3, and 4) B-type podolactones, along with three totarane-type diterpenes (5-7). Their structures were determined by interpretation of High Resolution ElectroSpray Ionization Mass Spectrometry (HRESIMS) and 1D and 2D NMR data, and comparison with the values reported in the literature. The structure of compound 1, previously identified as 3-deoxy-2α-hydroxynagilactone E (8), was revised as its 2β-epimer, which has been reported recently as a new compound. All of the isolates were evaluated for their antiproliferative activity against a panel of four human cancer cell lines, namely, ovarian (OVCAR3), breast (MDA-MB-231), colon (HT-29), and melanoma (MDA-MB-435), and compounds 1 and 3 were found to be cytotoxic with IC50 values in the low micromolar range for most of the cell lines used. The major compound, inumakilactone A (3), was further tested in vivo using the HT-29, MDA-MB-435, and OVCAR3 cells in a murine hollow fiber model, for the first time. Graphical Abstract

Developing a safer and more effective intersection-control system is essential given the trends of rising populations and vehicle numbers. Additionally, as vehicle communication and self-driving technologies evolve, we may create a more intelligent control system to reduce traffic accidents. We recommend deep reinforcement learning-inspired autonomous intersection management (DRLAIM) to improve traffic environment efficiency and safety. The three primary models used in this methodology are the priority assignment model, the intersection-control model learning, and safe brake control. The brake-safe control module is utilized to make sure that each vehicle travels safely, and we train the system to acquire an effective model by using reinforcement learning. We have simulated our proposed method by using a simulation of urban mobility tools. Experimental results show that our approach outperforms the traditional method.

PurposeTo investigate the effectiveness of aerosol clearance using an aerosol box, aerosol bag, wall suction, and a high-efficiency particulate air (HEPA) filter evacuator to prevent aerosol transmission.MethodsThe flow field was visualized using three protective device settings (an aerosol box, and an aerosol bag with and without sealed working channels) and four suction settings (no suction, wall suction, and a HEPA filter evacuator at flow rates of 415 liters per minute [LPM] and 530 LPM). All 12 subgroups were compared with a no intervention group. The primary outcome, aerosol concentration, was measured at the head, trunk, and foot of a mannequin.ResultsThe mean aerosol concentration was reduced at the head (p < 0.001) but increased at the feet (p = 0.005) with an aerosol box compared with no intervention. Non-sealed aerosol bags increased exposure at the head and trunk (both, p < 0.001). Sealed aerosol bags reduced aerosol concentration at the head, trunk, and foot of the mannequin (p < 0.001). A sealed aerosol bag alone, with wall suction, or with a HEPA filter evacuator reduced the aerosol concentration at the head by 7.15%, 36.61%, and 84.70%, respectively (99.9% confidence interval [CI]: -4.51-18.81, 27.48-45.73, and 78.99-90.40); trunk by 70.95%, 73.99%, and 91.59%, respectively (99.9% CI: 59.83-82.07, 52.64-95.33, and 87.51-95.66); and feet by 69.16%, 75.57%, and 92.30%, respectively (99.9% CI: 63.18-75.15, 69.76-81.37, and 88.18-96.42), compared with an aerosol box alone.ConclusionsAs aerosols spread, an airtight container with sealed working channels is effective when combined with suction devices.

Oxyhemoglobin saturation by pulse oximetry (SpO2) has always played an important role in the diagnosis of symptoms. Considering that the traditional SpO2 measurement has a certain error due to the number of wavelengths and the algorithm and the wider application of machine learning and spectrum combination, we propose to use 12-wavelength spectral absorption measurement to improve the accuracy of SpO2 measurement. To investigate the multiple spectral regions for deep learning for SpO2 measurement, three datasets for training and verification were built, which were constructed over the spectra of first region, second region, and full region and their sub-regions, respectively. For each region under the procedures of optimization of our model, a thorough of investigation of hyperparameters is proceeded. Additionally, data augmentation is preformed to expand dataset with added noise randomly, increasing the diversity of data and improving the generalization of the neural network. After that, the established dataset is input to a one dimensional convolution neural network (1D-CNN) to obtain a measurement model of SpO2. In order to enhance the model accuracy, GridSearchCV and Bayesian optimization are applied to optimize the hyperparameters. The optimal accuracies of proposed model optimized by GridSearchCV and Bayesian Optimization is 89.3% and 99.4%, respectively, and trained with the dataset at the spectral region of six wavelengths including 650 nm, 680 nm, 730 nm, 760 nm, 810 nm, 860 nm. The total relative error of the best model is only 0.46%, optimized by Bayesian optimization. Although the spectral measurement with more features can improve the resolution ability of the neural network, the results reveal that the training with the dataset of the shorter six wavelength is redundant. This analysis shows that it is very important to construct an effective 1D-CNN model area for spectral measurement using the appropriate spectral ranges and number of wavelengths. It shows that our proposed 1D-CNN model gives a new and feasible approach to measure SpO2 based on multi-wavelength.

This study investigated the ballistic resistance of a composite target plate fabricated by combining SiC ceramic with the Dyneema fiber. To achieve a light-weight target plate that conforms to the US National Institute of Justice level four (NIJ IV) standards, minimal areal density analysis was conducted to obtain the optimal SiC ceramic-Dyneema fiber thickness combination. This study used energy absorption to analyze the ballistic resistance of the target plates. To drastically reduce experimental costs, most of this work employed ANSYS/LS-DYNA software to conduct finite element numerical simulations. First, ballistic experiments that conformed to NIJ IV standards were conducted to verify the simulation parameter configurations. Subsequently, the correlation function of the relationship between the combined thickness of the composite material and its ballistic resistance was determined through the experimental design, which effectively reduced the simulation analysis time. According to simulation experiments and regression analysis, the equation for the relationship between the combined thickness of the composite material and its ballistic resistance was EAhc,hf=−6276.5+500.6hc+1512.6hf+30.7hchf−8.1hc2−113.6hf2, though there were limitations to its application. From the numerical analysis results, 8.1940 mm SiC ceramic and 6.9637 mm Dyneema fiber were determined to constitute the optimal thickness combination for a composite that features a minimal areal density and which conforms to NIJ IV standards. The combination was verified to be consistent with the numerical simulation analysis results.

To report the success rates of office probing for congenital nasolacrimal duct obstruction (NLDO) among children of different age groups in Taiwan. Methods: In this single-center, retrospective study, 564 eyes of 477 patients under the age of 5 years diagnosed with congenital NLDO were treated in a stepwise manner between 2001 and 2013. For infants aged < 6 months, treatment with massage and observation was suggested, followed by deferred probing under topical anesthesia if symptoms persisted. However, in cases of severe infection, immediate probing was suggested. In children aged > 6 months, office probing was usually highly recommended. Those with probing failures received either a second probing or silicone intubation. Treatment success was defined as anatomic patency by immediate irrigation after probing and absence of epiphora or mucous discharge at the follow-up visit. Results: Primary probing was successful in 457 of 564 eyes (success rate: 81%). The success rate of primary probing was negatively correlated with increasing age: 90.1% (163/181), 79.6% (164/206), 76.8% (73/95), 73.5% (36/49), 75% (18/24), and 33% (3/9) for the age groups of 0 to

Breast tumors reprogram their cellular metabolism, nutrient uptake, and utilization-associated biochemical processes. These processes become further transformed as genetically predisposed metastatic breast tumor cells colonize specific organs. Breast tumor cells often metastasize to the brain, bone, lung and liver. Massagué and colleagues isolated organotropic subclones and established organ-specific gene signatures associated with lung-, bone-, and brain-specific metastatic triple-negative breast cancer (TNBC) MDA-MB-231 cells. Using these genetically characterized metastatic subclones specific to lung (LM4175), bone (BoM1833), and brain (BrM-2a), we evaluated marine natural products for the ability to differentially suppress metastatic breast cancer cells in a target organ-dependent manner. Psammaplin-based histone deacetylase (HDAC) inhibitors were found to differentially inhibit HDAC activity, induce activation of hypoxia-inducible factor-1 (HIF-1), and disrupt organotropic metastatic TNBC subclone growth. Further, psammaplins distinctly suppressed the outgrowth of BoM1833 tumor spheroids in 3D-culture systems. Similar results were observed with the prototypical HDAC inhibitor trichostatin A (TSA). These organotropic tumor cell-based studies suggest the potential application of HDAC inhibitors that may yield new directions for anti-metastatic breast tumor research and drug discovery.

In this paper, the design of a quadrotor vehicle having a person-tracking and observation system, which uses human gesture recognition, is described. The system has three operating functions, namely, object tracking, human gesture recognition, and fixed-point cruising. The tracking-learning-detection (TLD) algorithm was used to enable the autonomous tracking of the object from images. An extended Kalman filter (EKF) provides an estimate of the current position of the quadrotor vehicle, and a fuzzy-proportional integral derivative (PID) controller provides position error compensation. The principle of the human gesture recognition system is as follows. A background model is first built from images using a Gaussian mixture model (GMM) to detect the foreground image. A nonlinear support vector machine (SVM) is then employed to recognize changes of gesture and establish interactivity between the vehicle and the user. The coordinates of the vehicle are marked using a GPS for fixed-point cruising. The coordinates and parameters of the points are set so that the quadrotor vehicle can follow them during cruising. Lastly, all of the functions are incorporated into the person-tracking and gesture-recognition system in the quadrotor. The experimental results show the feasibility of the above-mentioned methods, which can help us easily recognize the various gestures in this study. [ABSTRACT FROM AUTHOR]

This study aims to distinguish an array of green activities (ISO 14000, green processes, pollution prevention, and green certifications) and analyze their relationships with firm performance. Employing data from the sampled publicly listed firms in Taiwan and regressions to examine the hypotheses, we find that the degree of a firm's R&D investment fails to affect companies' choice of green activities; however, the degree of firm internationalization can. That is, more internationalized firms are also more likely to employ green certifications among these activities. In terms of the impacts of these green activities on firm performance, a company employing green processes can perform better, followed by ISO 14000, pollution prevention, and lastly, green certifications. [ABSTRACT FROM AUTHOR]

Prenatal examination plays an important role in present medical diagnosis. It provides information on fetal health status as well as the diagnosis of fetal treatment feasibility. The diagnosis can provide peace of mind for the perspective mother. Timely pregnancy termination diagnosis can also be determined if required. Amniocentesis and chorionic villus sampling are two widely used invasive prenatal diagnostic procedures. To obtain complete fetal genetic information and avoid endangering the fetus, noninvasive prenatal diagnosis has become the vital goal of prenatal diagnosis. However, the development of a high-efficiency separation technology is required to obtain the scarce fetal cells from maternal circulation. In recent years, the rapid development of microfluidic systems has provided an effective method for fetal cell separation. Advantages such as rapid analysis of small samples, low cost, and various designs, greatly enhance the efficiency and convenience of using microfluidic systems for cell separation. In addition, microfluidic disks can be fully automated for high throughput of rare cell selection from blood samples. Therefore, the development of microfluidic applications in noninvasive prenatal diagnosis is unlimited.

A 55-year-old male presented with a rapidly expanding mass on the right side of the neck and progressive hoarseness. An electronic laryngoscopy and a computed tomography scan were performed, and the patient was subsequently diagnosed with tumors of the larynx and the thyroid gland. An en bloc near-total thyroidectomy combined with a total laryngectomy was performed. The final pathological analysis revealed a collision tumor that was derived from a laryngeal squamous cell carcinoma and a papillary thyroid carcinoma. Collision tumors of the head and neck are rare. The therapy for a collision tumor should consist of a combination of the treatments that are normally used for each focus. [ABSTRACT FROM AUTHOR]