A method for segmenting clustered nuclei in cervical smear images

Abstract

The invention provides a method for segmenting clustered nuclei in a cervical smear image, comprising the steps of: (1) preparing a data set for segmentation; (2) selecting a data set and dividing it into a test set and a training set; (3) defining DeepHLF Network, the original picture is input into the network, the network progressively extracts features and retains each level of features, and then groups the features; DeepHLF uses a high-low coupling parallel fusion module to fuse the features of each group to generate three types of features; the three types of features are combined in a cross cycle , generate multiple feature maps, and finally each feature map generates a segmentation result map; (4) Propose a mathematical method and a weight loss function to solve the kernel and background category correction. The method of the present invention can not only segment cluster-type nuclei, but also will not miss shallow nuclei or nuclei with similar gray scales to the cytoplasm during the segmentation process.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to a method for cell segmentation in cervical smear images. Background technique [0002] Early detection of cervical cancer is of great significance to reduce the mortality rate of cervical cancer, and now the cervical smear screening technology is generally used for inspection. Today's smear screening is mainly based on manual reading, but this method is very inefficient; with the maturity of cervical cell diagnosis technology and the development of domestic cell automatic production technology, the development of computer-aided The diagnosis system has also become inevitable, and it has great significance for the screening of cervical cancer. [0003] At present, there are many cell image segmentation methods. Here, the information or theories used by these algorithms can be abstracted into cell segmentation algorithms using area information, segmentation algorithms usi...

AI Technical Summary

Problems solved by technology

Related theories or algorithms commonly used in cell segmentation include: wavelet analysis, genetic algorithm, mathematical morphology, neural network, etc., especially neural networks. In recent years, with the in-depth development of deep learning, some networks have also been derived in the field of cell nucleus segmentation. , such as Unet, shallow neural network, FCN, but it can only segment a single nucleus very well like the traditional method, and the cluster nucleus segmentation effect is not good

The cell segmentation algorithm using regional information is sensitive to noise and grayscale inhomogeneity, and it is easy to miss cell nuclei

The segmentation algorithm using cell edge information mainly relies on the grayscale information of the picture, and the segmentation effect on pictures with similar foreground and background is not good; the cell segmentation method combined with related theories, such as the existing neural network framework, has relatively limited segmentation capabilities. Only a single nucleus can be segmented, and the segmentation of clustered nuclei and nuclei with similar gray levels to the nucleus and cytoplasm is very poor

Images