RESUMO
BACKGROUND: When treating patients with coronary artery disease and concurrent renal concerns, we often encounter a conundrum: how to achieve a clearer view of vascular details while minimizing the contrast and radiation doses during percutaneous coronary intervention (PCI). Our goal is to use deep learning (DL) to create a real-time roadmap for guiding PCI. To this end, segmentation, a critical first step, paves the way for detailed vascular analysis. Unlike traditional supervised learning, which demands extensive labeling time and manpower, our strategy leans toward semi-supervised learning. This method not only economizes on labeling efforts but also aims at reducing contrast and radiation exposure. METHODS AND RESULTS: CAG data sourced from eight tertiary centers in Taiwan, comprising 500 labeled and 8952 unlabeled images. Employing 400 labels for training and reserving 100 for validation, we built a U-Net based network within a teacher-student architecture. The initial teacher model was updated with 8952 unlabeled images inputted, employing a quality control strategy involving consistency regularization and RandAugment. The optimized teacher model produced pseudo-labels for label expansion, which were then utilized to train the final student model. We attained an average dice similarity coefficient of 0.9003 for segmentation, outperforming supervised learning methods with the same label count. Even with only 5 % labels for semi-supervised training, the results surpassed a supervised method with 100 % labels inputted. This semi-supervised approach's advantage extends beyond single-frame prediction, yielding consistently superior results in continuous angiography films. CONCLUSIONS: High labeling cost hinders DL training. Semi-supervised learning, quality control, and pseudo-label expansion can overcome this. DL-assisted segmentation potentially provides a real-time PCI roadmap and further diminishes radiation and contrast doses.
Assuntos
Vasos Coronários , Aprendizado Profundo , Aprendizado de Máquina Supervisionado , Humanos , Vasos Coronários/diagnóstico por imagem , Intervenção Coronária Percutânea/métodos , Doença da Artéria Coronariana/diagnóstico por imagem , Angiografia Coronária/métodos , Processamento de Imagem Assistida por Computador/métodosRESUMO
Levels of persistent organic pollutants (POPs), including PAHs, PCBs, DDTs, and PBDEs, were measured in sediment collected from along the Taiwan coast and compared to previous studies. The dominant POPs were PAHs, followed by PCBs, PBDEs, and DDTs. The highest levels of PAHs and PCBs were found in sediment from harbors in southern Taiwan, which are surrounded by densely populated areas and affected by multiple industrial activities. In contrast, significantly higher levels of PBDEs were found at the northern coastline, which has a higher population and includes the metropolitan Taipei area. Using diagnostic PAH ratios, the predominant sources of PAHs in coastal Taiwan was determined to be pyrolytic-related activities. The main component of each POP was low- to moderately-chlorinated congeners, p,p'-DDE and BDE209, respectively. Further studies are required to assess the impact of these POPs on marine and coastal ecosystem.
Assuntos
Bifenilos Policlorados , Poluentes Químicos da Água , Ecossistema , Monitoramento Ambiental , Sedimentos Geológicos , Bifenilos Policlorados/análise , Taiwan , Poluentes Químicos da Água/análiseRESUMO
Visual acuity (VA) measurement is utilized to test a subject's acuteness of vision. Conventional VA measurement requires a physician's assistance to ask a subject to speak out or wave a hand in response to the direction of an optotype. To avoid this repetitive testing procedure, different types of automatic VA tests have been developed in recent years by adopting contact-based responses, such as pushing buttons or keyboards on a device. However, contact-based testing is not as intuitive as speaking or waving hands, and it may distract the subjects from concentrating on the VA test. Moreover, problems related to hygiene may arise if all the subjects operate on the same testing device. To overcome these problems, we propose an intelligent VA estimation (iVAE) system for automatic VA measurements that assists the subject to respond in an intuitive, noncontact manner. VA estimation algorithms using maximum likelihood (VAML) are developed to automatically estimate the subject's vision by compromising between a prespecified logistic function and a machine-learning technique. The neural-network model adapts human learning behavior to consider the accuracy of recognizing the optotype as well as the reaction time of the subject. Furthermore, a velocity-based hand motion recognition algorithm is adopted to classify hand motion data, collected by a sensing device, into one of the four optotype directions. Realistic experiments show that the proposed iVAE system outperforms the conventional line-by-line testing method as it is approximately ten times faster in testing trials while achieving a logarithm of the minimum angle of resolution error of less than 0.2. We believe that our proposed system provides a method for accurate and fast noncontact automatic VA testing.
Assuntos
Mãos , Testes Visuais , Algoritmos , Humanos , Aprendizado de Máquina , Acuidade VisualRESUMO
Motion related human activity recognition using wearable sensors can potentially enable various useful daily applications. So far, most studies view it as a stand-alone mathematical classification problem without considering the physical nature and temporal information of human motions. Consequently, they suffer from data dependencies and encounter the curse of dimension and the overfitting issue. Their models are hard to be intuitively understood. Given a specific motion set, if structured domain knowledge could be manually obtained, it could be used for better recognizing certain motions. In this study, we start from a deep analysis on natural physical properties and temporal recurrent transformation possibilities of human motions and then propose a useful Recurrent Transformation Prior Knowledge-based Decision Tree (RT-PKDT) model for recognition of specific human motions. RT-PKDT utilizes temporal information and hierarchical classification method, making the most of sensor streaming data and human knowledge to compensate the possible data inadequacy. The experiment results indicate that the proposed method performs superior to those adopted in related works, such as SVM, BP neural networks, and Bayesian Network, obtaining an accuracy of 96.68%.