A novel pasteurization method and an abnormal cervical cell automatic identification method

Abstract

The invention discloses a novel pasteurization method and an abnormal cervical cell automatic identification method. The novel pasteurization method comprises two modules, the module 1 is used for training a cervical cell classification model based on a mass cervical cell data set, and the module 2 is used for identifying abnormal cervical cells by using the trained classification model. The invention provides the novel pasteurization method, so that the problems existing in the traditional pasteurization method are well solved, and the possibility is provided for realizing automatic and high-precision positioning of cervical cell nuclei by a computer. The invention also provides the abnormal cervical cell automatic identification method, and experiments show that the method realizes the ultrahigh-precision automatic identification of abnormal cervical cells, so that the diagnosis burden of pathologists is greatly reduced, the diagnosis efficiency and the precision of the cervical diseases are improved, and the method has very high practical value and huge social benefits.

Description

technical field [0001] The invention relates to the technical field of medical image diagnosis, in particular to a novel Pap staining method and an automatic identification method for abnormal cervical cells. Background technique [0002] In recent years, the development of deep learning is in full swing, especially since the proposal of the convolutional neural network model represented by ResNet and DenseNet, the convolutional neural network has become the earliest and most widely used deep learning model. The reason why ResNet and DenseNet can achieve such a large performance improvement is actually due to the idea of ​​​​"crossing connections" used in the network structure. Although both use "spanning connections", their design ideas are different. Among them, ResNet's crossing connection is mainly to solve the problem that the deep network is not easy to fit the identity mapping. ResNet is the output of the residual block through the crossing connection. Provide a reas...

AI Technical Summary

Problems solved by technology

[0004] At present, the staining method used in cervical cytology is mainly Papanicolaou staining. There are two main disadvantages in the Papanicolaou staining method: 1. The contrast of nuclear and cytoplasmic staining of basal cervical cells is low;

In addition, the traditional Pap manual reading technology relies on manpower to find some diseased cells from a large number of cells under the microscope. The work intensity is huge, which is easy to cause fatigue, and requires operators to have high pathological knowledge and clinical experience. The diagnosis result is affected by many aspects such as the operator's subjective factors, and human error is inevitable

Images