摘要：【中文】本发明的一种基于神经网络的组织癌变的检测方法及系统，可解决现有的针对组织癌变检测的方法判断过程繁琐且误差较大的技术问题。包括以下步骤：获取图像数据，确定数据集；在数据集中确定先验知识注意力蒙版的设定；基于先验知识注意力蒙版，构建神经网络模型；对图像进行正则化处理；基于神经网络模型，使用训练数据为学习素材，手工标签作为学习目标，利用反向传播的方式对网络中的可训练参数进行训练，得到最终训练好的模型；以经过正则化处理的图像作为输入，经过训练好的神经网络模型输出，得到最终的检测结果。本发明通过注意力机制，更好的在组织切片的图像中，利用周围组织作为参考，提高对判断一小块组织切片是否患有癌症的准确率。 【EN】The invention discloses a method and a system for detecting tissue canceration based on a neural network, and can solve the technical problems of complicated judgment process and larger error of the existing method for detecting tissue canceration. The method comprises the following steps: acquiring image data and determining a data set; determining a setting of a priori knowledge attention mask in a data set; constructing a neural network model based on the prior knowledge attention mask; regularizing the image; based on a neural network model, training data are used as learning materials, manual labels are used as learning targets, and trainable parameters in the network are trained in a back propagation mode to obtain a finally trained model; and taking the image subjected to regularization processing as input, and outputting the image through the trained neural network model to obtain a final detection result. According to the invention, through an attention mechanism, surrounding tissues are better utilized as references in the images of the tissue slices, and the accuracy rate of judging whether a small tissue slice has cancer is improved.

摘要：【中文】本发明的一种用于细胞迁移实验的芯片装置、制备方法及实验方法，可解决使用细胞布置器过程繁琐，且对细胞污染几率大影响实验精度的技术问题。所述芯片表面包含多个重复功能单元；每个功能单元至少包含接种区、隔离区和迁移区三个功能区域；接种区是细胞起始接种黏附的区域，接种区表面具有疏水性；隔离区是位于接种区外围设置隔离柱体阵列，隔离柱体阵列将接种区包围形成一个环形结构，相邻隔离柱体之间保留设定的间隙，隔离柱体外表面具有疏水性；迁移区位于隔离区外围。本发明实现细胞起始接种范围的限定，并能控制细胞向指定区域范围和方向迁移，同时能模拟体内不同迁移界面的结构，无需对细胞进行损伤操作，也不需要对芯片进行拆卸和组装。 【EN】The chip device, the preparation method and the experimental method for the cell migration experiment can solve the technical problems that the process of using the cell arranging device is complicated, and the probability of cell pollution is high, so that the experimental precision is influenced. The chip surface comprises a plurality of repeating functional units; each functional unit at least comprises three functional areas, namely a seeding area, an isolation area and a migration area; the inoculation area is an area where cells are initially inoculated and adhered, and the surface of the inoculation area has hydrophobicity; the isolation area is formed by arranging an isolation column array at the periphery of the inoculation area, the isolation column array surrounds the inoculation area to form an annular structure, a set gap is reserved between adjacent isolation columns, and the outer surface of each isolation column has hydrophobicity; the migration region is located at the periphery of the isolation region. The invention realizes the limitation of the initial inoculation range of the cells, can control the migration of the cells to the range and the direction of the designated area, can simulate the structures of different migration interfaces in vivo, does not need to carry out damage operation on the cells, and does not need to disassemble and assemble the chip.

摘要：【中文】本发明公开了一种整体旋转式筛分装置，包括前部和后部分别设有进浆管和排浆管的旋转筒体和支撑旋转筒体转动的支撑机构，所述旋转筒体内靠近进浆管的一侧设有搅拌机构，旋转筒体内靠近排浆管的一侧设有筛分机构，且旋转筒体通过传动机构与驱动单元连接。本发明可对渣石进行高效筛分，提升泥水环流系统的工作效率，解决现有筛分装置存在的筛分效果差、容易滞排或堵塞等问题，进而提高隧道施工进度。 【EN】The invention discloses an integral rotary screening device which comprises a rotary cylinder body and a supporting mechanism, wherein the front part and the rear part of the rotary cylinder body are respectively provided with a pulp inlet pipe and a pulp outlet pipe, the supporting mechanism is used for supporting the rotary cylinder body to rotate, a stirring mechanism is arranged on one side, close to the pulp inlet pipe, in the rotary cylinder body, a screening mechanism is arranged on one side, close to the pulp outlet pipe, in the rotary cylinder body, and the rotary cylinder body is connected with a driving unit through a transmission mechanism. The slurry circulating system can efficiently screen the slag, improves the working efficiency of the slurry circulating system, solves the problems of poor screening effect, easy discharge stagnation or blockage and the like of the conventional screening device, and further improves the tunnel construction progress.

摘要：【中文】本发明的一种基于潜变量特征生成的图像多风格转化方法，可解决现有的图像转换方法在风格表示和图像质量上都不能满足需求的技术问题。包括以下步骤：S100、采集图像数据；S200、对图像进行预处理；S300、构建图像转化模型并优化；S400、基于步骤S300的图像转化模型，以步骤S200的图像数据作为输入，对图像进行转化处理；S500、对转换得到的图像质量进行评估。本发明在MUNIT的基础上对模型进行了扩充和改进，通过设计跳跃连接对图像的风格转换提供了充分的内容特征信息的传递；风格编码生成器对图像风格编码的潜在变量进行学习，可以实现丰富的图像风格转换；同时模型可以根据输入的风格图像为参考实现特定风格的转换，对于特定风格转换任务的开展具有重大指导意义。 【EN】The image multi-format conversion method based on latent variable feature generation can solve the technical problem that the conventional image conversion method cannot meet the requirements in terms of style representation and image quality. The method comprises the following steps: s100, collecting image data; s200, preprocessing the image; s300, constructing an image conversion model and optimizing; s400, based on the image conversion model in the step S300, taking the image data in the step S200 as input, and performing conversion processing on the image; and S500, evaluating the quality of the image obtained by conversion. The invention expands and improves the model on the basis of MUNIT, and provides sufficient transmission of content characteristic information for the style conversion of the image by designing jump connection; the style code generator learns the latent variables of the image style codes, so that rich image style conversion can be realized; meanwhile, the model can realize the conversion of a specific style by taking the input style image as a reference, and has great guiding significance for the development of a specific style conversion task.

摘要：【中文】本发明的一种基于细胞明场图片对细胞内部结构的分析方法及系统，可解决现有方案操作繁琐，效率较低的技术问题；包括获取细胞明场图片，通过卷积层做图像基本内容特征的提取，再通过残差层对图像进一步做深层特征提取，得到明场图像特征；对荧光图像通过基础的卷积层提取得到风格特征；先经过感知层得到风格信息下的均值和方差表示，在自适应实例归一化下将荧光图像风格特征嵌入到明场图像内容特征中；融合后经过残差层做特征学习，在反卷积层中与明场图像特征提取过程中的卷积中间层进行跳跃连接重构得到荧光图像。本发明利用深度学习下图像到图像转换方法去替代物理试剂染色过程，在保证荧光信息颜色的前提下实现了不同荧光图像特定蛋白成分的凸显。 【EN】The method and the system for analyzing the internal structure of the cell based on the cell bright field picture can solve the technical problems of complex operation and low efficiency of the existing scheme; the method comprises the steps of obtaining a cell bright field picture, extracting basic content features of the image through a convolution layer, and further extracting deep layer features of the image through a residual error layer to obtain bright field image features; extracting the fluorescent image through a basic convolution layer to obtain style characteristics; obtaining mean value and variance representation under style information through a perception layer, and embedding the style characteristics of the fluorescence image into the content characteristics of the bright field image under the self-adaptive example normalization; and after fusion, performing feature learning through a residual error layer, and performing jump connection reconstruction on the deconvolution layer and a convolution intermediate layer in the bright field image feature extraction process to obtain a fluorescence image. The method replaces the physical reagent dyeing process by an image-to-image conversion method under deep learning, and realizes the highlighting of specific protein components of different fluorescence images on the premise of ensuring the fluorescence information color.

摘要：【中文】本发明的一种基于胎儿四腔心切面图像的多组织分割方法，可解决现有的超声成像技术的局限性使得胎儿先心病的筛查工作负担较大而且准确率较低的技术问题。包括以下步骤：收集并获取胎儿四腔心切面图像数据；对获取的图像数据进行处理并将处理后的数据按设定比例分成训练集和测试集；基于深度学习框架搭建多组织分割模型，基于训练集对模型进行训练，并进行参数优化；基于优化后的多组织分割模型对处理后的测试集数据进行多组织分割。本发明可以实现四腔心切面的多组织自动分割，能实质性地减轻医生的和工作负担对临床经验的依赖性，减少临床上的人力需求，有利于基层医院的医生进行筛查与诊断，减少误诊与漏诊情况的出现。 【EN】The multi-tissue segmentation method based on the fetal four-chamber cardiac section image can solve the technical problems that the screening work burden of fetal congenital heart disease is large and the accuracy is low due to the limitation of the existing ultrasonic imaging technology. The method comprises the following steps: collecting and obtaining image data of a fetal four-cavity heart section; processing the acquired image data and dividing the processed data into a training set and a test set according to a set proportion; building a multi-organization segmentation model based on a deep learning framework, training the model based on a training set, and performing parameter optimization; and performing multi-organization segmentation on the processed test set data based on the optimized multi-organization segmentation model. The invention can realize multi-tissue automatic segmentation of the four-chamber heart section, can substantially reduce the dependence of doctors and workload on clinical experience, reduce the clinical manpower requirement, is beneficial to the screening and diagnosis of doctors in primary hospitals, and reduces the occurrence of misdiagnosis and missed diagnosis.

摘要：【中文】一种用于泛在电力物联网的QPSK调制非相干检测方法，收到的信号接收值包括已知码片和未知码片，所述方法包括如下步骤：S1、对信号接收值进行解析处理，提取出和前导符中的全零比特对应的第一样值序列以及和未知码片对应的第二样值序列；S2、根据第一样值序列和第一伪随机码片序列生成用于估计残余载波频偏的度量值；S3、从度量值中提取频偏影响值并且进行估计，得到频偏影响估计值；S4、利用频偏影响估计值对第二样值序列进行补偿修正得到判决样本序列；S5、根据判决样本序列和第二伪随机码片序列生成判决统计量；S6、对判决统计量进行检测得到检测结果。本发明可靠性高、鲁棒性强，且计算复杂度低，能够降低网络节点的能耗。 【EN】A QPSK modulation incoherent detection method for a ubiquitous power internet of things, a received signal reception value including known chips and unknown chips, the method comprising the steps of: s1, analyzing the signal receiving value, extracting a first sample value sequence corresponding to all zero bits in the preamble and a second sample value sequence corresponding to unknown chips; s2, generating a metric value for estimating residual carrier frequency offset according to the first sample sequence and the first pseudo-random chip sequence; s3, extracting a frequency offset influence value from the metric value and estimating to obtain a frequency offset influence estimation value; s4, compensating and correcting the second sample sequence by using the frequency offset influence estimation value to obtain a judgment sample sequence; s5, generating a decision statistic according to the decision sample sequence and the second pseudo-random chip sequence; and S6, detecting the judgment statistic to obtain a detection result. The invention has high reliability, strong robustness and low computation complexity, and can reduce the energy consumption of network nodes.