Search

Abstract. Background:. The goal of the assisted reproductive treatment is to transfer one euploid blastocyst and to help infertile women giving birth one healthy neonate. Some algorithms have been used to assess the ploidy status of embryos derived from couples with normal chromosome, who subjected to preimplantation genetic testing for aneuploidy (PGT-A) treatment. However, it is currently unknown whether artificial intelligence model can be used to assess the euploidy status of blastocyst derived from populations with chromosomal rearrangement. Methods:. From February 2020 to May 2021, we collected the whole raw time-lapse videos at multiple focal planes from in vitro cultured embryos, the clinical information of couples, and the comprehensive chromosome screening results of those blastocysts that had received PGT treatment. Initially, we developed a novel deep learning model called the Attentive Multi-Focus Selection Network (AMSNet) to analyze time-lapse videos in real time and predict blastocyst formation. Building upon AMSNet, we integrated additional clinically predictive variables and created a second deep learning model, the Attentive Multi-Focus Video and Clinical Information Fusion Network (AMCFNet), to assess the euploidy status of embryos. The efficacy of the AMCFNet was further tested in embryos with parental chromosomal rearrangements. The receiver operating characteristic curve (ROC) was used to evaluate the superiority of the model. Results:. A total of 4112 embryos with complete time-lapse videos were enrolled for the blastocyst formation prediction task, and 1422 qualified blastocysts received PGT-A (n = 589) or PGT for chromosomal structural rearrangement (PGT-SR, n = 833) were enrolled for the euploidy assessment task in this study. The AMSNet model using seven focal raw time-lapse videos has the best real-time accuracy. The real-time accuracy for AMSNet to predict blastocyst formation reached above 70% on the day 2 of embryo culture, and then increased to 80% on the day 4 of embryo culture. Combing with 4 clinical features of couples, the AUC of AMCFNet with 7 focal points increased to 0.729 in blastocysts derived from couples with chromosomal rearrangement. Conclusion:. Integrating seven focal raw time-lapse images of embryos and parental clinical information, AMCFNet model have the capability of assessing euploidy status in blastocysts derived from couples with chromosomal rearrangement.

As a major infectious disease, tuberculosis (TB) still poses a threat to people’s health in China. As a triage test for TB, reading chest radiography with traditional approach ends up with high inter-radiologist and intra-radiologist variability, moderate specificity and a waste of time and medical resources. Thus, this study established a deep convolutional neural network (DCNN) based artificial intelligence (AI) algorithm, aiming at diagnosing TB on posteroanterior chest X-ray photographs in an effective and accurate way. Altogether, 5,000 patients with TB and 4,628 patients without TB were included in the study, totaling to 9,628 chest X-ray photographs analyzed. Splitting the radiographs into a training set (80.4%) and a testing set (19.6%), three different DCNN algorithms, including ResNet, VGG, and AlexNet, were trained to classify the chest radiographs as images of pulmonary TB or without TB. Both the diagnostic accuracy and the area under the receiver operating characteristic curve were used to evaluate the performance of the three AI diagnosis models. Reaching an accuracy of 96.73% and marking the precise TB regions on the radiographs, ResNet algorithm-based AI outperformed the rest models and showed excellent diagnostic ability in different clinical subgroups in the stratification analysis. In summary, the ResNet algorithm-based AI diagnosis system provided accurate TB diagnosis, which could have broad prospects in clinical application for TB diagnosis, especially in poor regions with high TB incidence.

Digital twin technology aims to build a map of physical entities in virtual space, and simulate the real-time state and dynamic characteristics of physical devices through bi-directional interactive data flow. In order to guarantee the safe and stable operation of protection system equipment in intelligent substations and improve the efficiency of protection system operation and condition maintenance, this paper proposes a protection system state evaluation and fault prediction method based on digital twin technology. The architecture, application and operation control mode of digital twin technology in a real-time state analysis of a protection system are studied. The state evaluation model based on matter-element extension and the fault prediction model based on clustering algorithm are constructed. By analyzing the historical data of an intelligent station protection system in actual operation, a database that can be updated and corrected in real time is constructed. The effectiveness and accuracy of the state evaluation and fault prediction method are verified with actual cases, which can provide technical support for the operation and maintenance of the protection system.

The main challenge in video salient object detection is how to model object motion and dramatic changes in appearance contrast. In this work, we propose an attention embedded spatio-temporal network (ASTN) to adaptively exploit diverse factors that influence dynamic saliency prediction within a unified framework. To compensate for object movement, we introduce a flow-guided spatial learning (FGSL) module to directly capture effective motion information in the form of attention based on optical flows. However, optical flow represents the motion information of all moving objects, including movements of non-salient objects caused by large camera motion and subtle changes in background. Therefore, using the flow-guided attention map alone causes the spatial saliency to be influenced by all moving objects rather than just the salient objects, resulting in unstable and temporally inconsistent saliency maps. To further enhance the temporal coherence, we develop an attentive bidirectional gated recurrent unit (AB-GRU) module to adaptively exploit sequential feature evolution. With this AB-GRU, we can further refine the spatiotemporal feature representation by incorporating an accommodative attention mechanism. Experimental results demonstrate that our model achieves superior empirical performance on video salient object detection. Moreover, an experiment on the extended application to unsupervised video object segmentation further demonstrates the generalization ability and stability of our proposed method.

This paper proposes a novel faulted section location method for power distribution networks based on the status information collected by fault-indicating equipments. It has an advantage that it does not need to analyze and calculate any electrical parameters, and it is not affected by fault types and system parameters, so this method can be easily applied to power distribution networks. Firstly, a novel method for automatically constructing a line list is proposed in this paper—it can represent the topology structure of the power distribution line and the status information of fault-indicating equipment. Based on the line list, a topology search algorithm is proposed in the method to locate the faulted section. Considering that information loss will cause errors in the line list, an information loss detection algorithm is proposed to detect and correct the wrong status of fault-indicating equipments. Then, different fault conditions such as single fault, multiple faults, distributed generations in system and information loss condition are tested in power distribution networks with different topologies, and the simulation results indicate that the proposed method can deal with each case well. Moreover, the proposed fault location method has an advantage that the running time does not necessarily increase when the node scale is expanded.