基于深度卷积神经网络的宫颈细胞病理智能辅助诊断方法

针对宫颈细胞病理自动筛查问题,提出一种基于深度卷积神经网络的智能辅助诊断方法。首先采用基于改进UNet深度卷积神经网络模型的语义分割方法,检测出宫颈细胞病理涂片扫描图像中的细胞(粘连簇团)区域。接着,利用VGG 16深度卷积神经网络模型,结合迁移学习技术,对检测出的细胞(粘连簇团)区域进行精确识别。为了提高深度卷积神经网络模型的性能,在进行细胞(粘连簇团)区域检测、识别的过程中,采用了数据增强技术。同时,针对该领域相关研究缺乏宫颈细胞病理液基涂片扫描图像数据集的问题,我们收集四川大学华西附二院的典型LCT筛查病例,建立了宫颈细胞病理图像HXLCT数据集,并由资深病理医生完成数据标注。实验表明,本文方法能够较好地完成宫颈细胞病理涂片扫描图像中的细胞(粘连簇团)区域检测(正确率为91.33%),并能对检测出的区域完成正常、疑似病变二分类识别(正确率为91.6%,召回率为92.3%,ROC曲线线下面积为0.914)。本文工作将有助于宫颈细胞病理自动筛查系统的开发,对于宫颈癌早期防治具有重要意义。     

脊髓灰质炎病毒免疫电镜形态发生学研究

在感染细胞内,脊髓灰质炎病毒进行复制是否需要核的参与至今仍存在争议。而有关该病毒形态发生的系统报道仍不多见。本文使用已建立的低温包埋金标记技术结合常规包埋技术对脊髓灰质炎病毒感染Hep-2细胞后的形态发生学进行了定位研究,探讨了该病毒与感染细胞的相互作用关系,细胞核在病毒复制中的作用以及病毒诱导空泡、核突出和复制复合体之间的关系。感染后4到12小时的细胞中,观察到大量核突出物形成及其由细胞核到细胞质的迁移动态。感染初期,核     

宫颈细胞的生物学特征-卷积神经网络联合分类方法

为了改善在宫颈细胞的分类工作中,出现的将异常的病变细胞与正常细胞判断混淆的误诊问题,提出了一种细胞生物学特征-卷积神经网络联合分类方法.首先,使用ResNet分类网络提取出特征向量,然后再将其与手动提取的DNA指数、细胞核/浆比特征一起输入到全连接层,并使用基于MSE损失值的逻辑回归分类,对宫颈细胞进行分类识别.使用5折交叉验证法在Heer数据集上的实验结果表明,这种将卷积神经网络与细胞生物学特征相结合的联合分类方法相较于ResNet卷积神经网络,分类结果的整体准确率提高4％,达到了95％;同时优化MSE损失函数的方法在准确率达到瓶颈的情况下,能够将严重错分率由2.10％降为0.248％,且保持了细胞的整体识别准确率.提出的方法进行计算机辅助检测,能够提升宫颈细胞分类工作准确率、降低误诊率.     

基于YOLO模型的宫颈细胞簇团智能识别方法

针对宫颈细胞簇团自动识别问题,本文提出了一种基于YOLO v2模型的智能识别方法。首先,针对宫颈细胞簇团识别任务的特点,采用resnet 50模型作为YOLO v2网络的基础特征提取模块。同时,提出了相应的数据扩增方法与YOLO v2网络的训练方案。同时,我们收集宫颈细胞液基涂片扫描图像,建立了宫颈细胞簇团图像数据集,并由细胞病理专家对其中的细胞簇团进行了标注。实验表明,本文方法能够有效完成宫颈细胞病变簇团的自动识别,在测试图像集中,针对细胞簇团识别的准确率为75.9%,召回率为86.3%;针对宫颈细胞图像识别的准确率为87.0%,召回率为86.7%。本文将深度学习技术引入到宫颈细胞辅助筛查领域,对于促进宫颈癌早期自动筛查系统的研究,具有重要意义。     

人生殖道疣和子宫颈癌中病毒基因检测及超微结构研究

本实验用透射电镜观察了32例子宫颈癌,6例宫颈湿疣和2例外阴疣的超微结构,同时以～(32)P-dCTP标记的HPV 16 DNA为探针,检测了宫颈癌活检组织中的病毒DNA相关序列。电镜下,生殖道疣各例均可见典型凹空细胞病变,细胞核内的核小体,染色质之间颗粒,染色质周围颗粒显著增多。核酸杂交显示23例宫颈癌组织中存在病毒DNA相关序列。杂交呈强阳性及中强阳性组织中的癌细胞核中的核小体明显多于弱阳性及阴性标本。。     

基于自适应阈值分割的宫颈细胞图像分类算法

本文以宫颈癌细胞图像的自动筛查为应用背景,研究了一种新的宫颈细胞图像分类算法。算法首先采用形态学滤波与自适应直方图均衡的预处理方法进行图像增强;根据对图像内容与直方图分布关系的深入分析,提出采用经验因子加权Otsu自适应阈值分割算法进行细胞核分割,有效地解决了细胞重叠所引起的自适应分割阈值的选取问题;然后,通过提取面积、周长、区域面积与外接凸多边形面积比以及长宽比四种参数,对分割出的细胞核区域进行杂质剔除;最后以最能体现癌细胞特征的面积、平均灰度作为特征参数采用K-means算法对样本图像进行分类实验。实验样本为233幅宫颈细胞图像,其中49幅癌细胞图像,184幅正常细胞图像,实验结果证明了该算法的有效性。      

鸡红血细胞用于激光流式细胞计技术中作为生物内插标准测定细胞核DNA含量分布的意义

鸡红血细胞作为生物内插标准可以监视荧光染色及仪器操作中的一切变化。本文报导丁所测得的几种细胞核 DNA 含量值,显示了检测过程中可靠的内插标准的重要性。     

宫颈上皮脱落细胞的微尺度力学性质及形貌学研究

目前临床中对宫颈脱落细胞的检查多局限于大体细胞形态学的观察.本文应用原子力显微镜及环境扫描电子显微镜对5例临床宫颈炎患者的宫颈脱落细胞进行了微区力学性质表征以及细胞表面微观形态的成像.结果显示,患者正常形态宫颈上皮脱落细胞在针尖压入深度为700 nm时杨氏模量近似正态分布,峰值在20～30kPa.且细胞表面微嵴明显,微...     

基于区域光流法的人体异常行为检测

为了提升智能视频监控系统在保障公共安全方面的能力,提出基于区域光流法对人体异常进行检测的方法.采用改进的背景差分法来准确提取前景目标,然后对提取出的前景目标进行光流特征计算,有效滤除环境噪声的影响.采用方向-幅值直方图来描述行为,通过计算直方图的速度判断是否发生抢劫、追逐行为,计算直方图的方向和幅值的熵以及速度的方差来判断是否发生打斗行为.实验结果表明,该方法在不同视角下均能取得良好效果.     

细胞核体外重建过程中核纤层与核膜组装的关系

核纤层Nuclear lamina是位于内层核膜下的纤维蛋白片层结构,为了研究核纤层与核膜在功能上的关系,我们利用体外核装配这一实验模式,初步探索了核纤层与核膜装配的关系。首先,我们应用噬菌体Lambda DNA和非洲爪蟾卵细胞抽提物建立了体外细胞核组装体系,体外构建的细胞核具有正常细胞核的形态结构(图1),同时证明核纤层参与了细胞核的体外组装。为进一步探讨核纤层在细胞核组装过程中的作用,我们在非细胞体系中加入足量的能与两栖类卵细胞核纤层蛋白Lamin Ⅲ反应的抬体,观察核纤层蛋白抬体对核膜组装的影响,超薄切片电镜观察显示核纤层蛋白抬体的介入导致非细胞体系中膜系组装的紊乱,不能形成正常的双层核膜,而是形成呈同心圆状排列的多层单位膜(图2)。其机制有待进一步探讨,推测核纤层蛋白单体组装成核纤层可能是细胞核组装过程中的一个重要环节,这一环节的缺失导致了膜系统,特别是双层核膜的组装异常。     

Morphological granulometric features of nucleus in automatic bone marrow white blood cell classification.

The proportion of counts of different types of white blood cells in the bone marrow, called differential counts, provides invaluable information to doctors for diagnosis. Due to the tedious nature of the differential white blood cell counting process, an automatic system is preferable. In this paper, we investigate whether information about the nucleus alone is adequate to classify white blood cells. This is important because segmentation of nucleus is much easier than the segmentation of the entire cell, especially in the bone marrow where the white blood cell density is very high. In the experiments, a set of manually segmented images of the nucleus are used to decouple segmentation errors. We analyze a set of white-blood-cell-nucleus-based features using mathematical morphology. Fivefold cross validation is used in the experiments in which Bayes' classifiers and artificial neural networks are applied as classifiers. The classification performances are evaluated by two evaluation measures: traditional and classwise classification rates. Furthermore, we compare our results with other classifiers and previously proposed nucleus-based features. The results show that the features using nucleus alone can be utilized to achieve a classification rate of 77% on the test sets. Moreover, the classification performance is better in the classwise sense when the a priori information is suppressed in both the classifiers.     

Tumor and stromal cell counts in the lung clones of transplantable rhabdomyosarcoma of rats determined by flow cytometry

The phase G1-tumour cells of 76 lung clones obtained from rat transplantable rhabdomyosarcoma had near triploid mean DNA content (the diploid index (DI) = 1.46-2.05) and the phase G1/o-stromal cells was characterized by the diploid DNA content. The percent of tumour cells in lung clones determined by flow cytometry varied from 19.7 to 70.9. A correlation is found between the following items: DI; coefficients of variability of DNA content in tumour and stromal cells; the share of the tumour cells in clones and the coefficient of variability of the DNA content in stromal cells; percentage of the tumour cells in the So+ (G2 + M) phase and percentage of the stromal cells etc. Positive correlation is revealed between the coefficient of variability of the DNA content in tumour cells and coefficients of variability of the DNA content in stromal cells (r = 0.83), the negative one (r = 0.38)--between the coefficient of variability of the DNA content of stromal cells of clones and percentage of tumour cells which produce artificial lung metastasis.     

基于卷积神经网络和胶囊神经网络的宫颈病变图像分类方法研究

为了更好地治疗宫颈癌,准确确定患者的宫颈类型是至关重要的。因此,用于检测和划分宫颈类型的自动化方法在该领域中具有重要的医学应用。虽然深度卷积神经网络和传统的机器学习方法在宫颈病变图像分类方面已经取得了良好的效果,但它们无法充分利用图像和图像标签的某些关键特征之间的长期依赖关系。为了解决这个问题,文章引入了胶囊网络(CapsNet),将CNN和CapsNet结合起来,以提出CNN-CapsNet框架,该框架可以加深对图像内容的理解,学习图像的结构化特征,并开展医学图像分析中大数据的端到端训练。特别是,文章应用迁移学习方法将在ImageNet数据集上预先训练的权重参数传输到CNN部分,并采用自定义损失函数,以便网络能够更快地训练和收敛,并具有更准确的权重参数。实验结果表明,与ResNet和InceptionV3等其他CNN模型相比,文章提出的网络模型在宫颈病变图像分类方面更加准确、有效。     

Co-Occurrence Histogram Based Ensemble of Classifiers for Classification of Cervical Cancer Cells

To explore the potential of conventional image processing techniques in the classification of cervical cancer cells, in this work, a co-occurrence histogram method was employed for image feature extraction and an ensemble classifier was developed by combining the base classifiers, namely, the artificial neural network (ANN), random forest (RF), and support vector machine (SVM), for image classification. The segmented pap-smear cell image dataset was constructed by the k-means clustering technique and used to evaluate the performance of the ensemble classifier which was formed by the combination of above considered base classifiers. The result was also compared with that achieved by the individual base classifiers as well as that trained with color, texture, and shape features. The maximum average classification accuracy of 93.44% was obtained when the ensemble classifier was applied and trained with co-occurrence histogram features, which indicates that the ensemble classifier trained with co-occurrence histogram features is more suitable and advantageous for the classification of cervical cancer cells.     

A fluid dynamics-based deformable model for segmentation of cervical cell images

A novel deformable model for unsupervised segmentation of cervical cells within Pap smear images is presented in this paper. The proposed method is inspired by fluid mechanics and based on the simulation of incompressible fluid flood via grid-based solution of Navier–Stokes equations. In this approach, simulation starts inside the cytoplasmic region and the simulated fluid is attracted toward the cell contours. Unlike most of the other fluid-based methods, gradient magnitude data are not used for extracting topological relief of the image. However, gradient magnitude of the image is still considered as the source for extracting particles. Direction of propagation of the flow is determined by an interaction mechanism based on the permeability rate of these particles. Interaction between fluid and particles guides the advancing fronts of the fluid toward object boundaries. Redefinition of complex topologies with particle groups provides potential of improved segmentation capability and flexibility to the model. We demonstrate the segmentation capability of our model with fully automated and unsupervised experimental setting on Pap smear sample images. Results showed that proposed method may be more adaptive than watershed algorithm and have an improved performance on recovering shape and boundary data of cervical cells.     

光谱技术在宫颈上皮内瘤变监测中的研究进展

宫颈上皮内瘤变(CIN)是与宫颈浸润癌密切相关的一组癌前病变,也是宫颈癌临床筛查的重点内容。目前,针对宫颈癌的筛查手段很多,其中光谱技术具有无创、实时、定量、原位、高灵敏性等优点成为该病筛查的研究热点。本文通过对包括漫反射光谱、固有荧光光谱、散射光谱3种光谱原理及其临床应用进行分类说明,并对多光谱成像技术、高光谱成像技术及激光共聚焦显微成像技术在宫颈病变临床诊断中的发展与应用做出综述,介绍了目前光谱技术在宫颈病变临床诊断中的数据收集、诊断方法及特殊优势,并对其未来的可能发展提出了展望。     

Nucleus‐Targeting Gold Nanoclusters for Simultaneous In Vivo Fluorescence Imaging,Gene Delivery,and NIR‐Light Activated Photodynamic Therapy

The nucleus is one of the most important cellular organelles and molecular anticancer drugs, such as cisplatin and doxorubicin, that target DNA inside the nucleus, are proving to be more effective at killing cancer cells than those targeting at cytoplasm. Nucleus‐targeting nanomaterials are very rare. It is a grand challenge to design highly efficient nucleus‐targeting multifunctional nanomaterials that are able to perform simultaneous bioimaging and therapy for the destruction of cancer cells. Here, unique nucleus‐targeting gold nanoclusters (TAT peptide–Au NCs) are designed to perform simultaneous in vitro and in vivo fluorescence imaging, gene delivery, and near‐infrared (NIR) light activated photodynamic therapy for effective cancer cell killing. Confocal laser scanning microscopy observations reveal that TAT peptide–Au NCs are distributed throughout the cytoplasm region with a significant fraction entering into the nucleus. The TAT peptide–Au NCs can also act as DNA nanocargoes to achieve very high gene transfection efficiencies (≈81%) in HeLa cells and in zebrafish. Furthermore, TAT peptide–Au NCs are also able to sensitize formation of singlet oxygen (¹O₂) without the co‐presence of organic photosensitizers for the destruction of cancer cells upon NIR light photoexcitation.     

An optimal transportation approach for nuclear structure-based pathology

Nuclear morphology and structure as visualized from histopathology microscopy images can yield important diagnostic clues in some benign and malignant tissue lesions. Precise quantitative information about nuclear structure and morphology, however, is currently not available for many diagnostic challenges. This is due, in part, to the lack of methods to quantify these differences from image data. We describe a method to characterize and contrast the distribution of nuclear structure in different tissue classes (normal, benign, cancer, etc.). The approach is based on quantifying chromatin morphology in different groups of cells using the optimal transportation (Kantorovich-Wasserstein) metric in combination with the Fisher discriminant analysis and multidimensional scaling techniques. We show that the optimal transportation metric is able to measure relevant biological information as it enables automatic determination of the class (e.g., normal versus cancer) of a set of nuclei. We show that the classification accuracies obtained using this metric are, on average, as good or better than those obtained utilizing a set of previously described numerical features. We apply our methods to two diagnostic challenges for surgical pathology: one in the liver and one in the thyroid. Results automatically computed using this technique show potentially biologically relevant differences in nuclear structure in liver and thyroid cancers.     

基于数学形态学的癌细胞的分割与识别

要判断一幅测得的细胞图像中是否存在癌细胞，如果仅凭经验去判断，不仅工作量大，而且准确率相对较低。文中介绍了一种基于形态学的对一幅细胞图像进行分割和识别的算法。即先对图像进行膨胀或腐蚀预处理，然后通过设置圆度阈值，计算出每一个细胞的圆度来与阈值进行比较，并提取出可疑的癌细胞。实验表明，该算法不仅大幅降低了医务人员的工作量，而且显著提高了癌细胞识别的准确率。最终检测结果的正确率达到了95%以上。