An automatic identification method for abnormal cervical cells based on a novel Pap staining method

Abstract

The invention discloses a novel Pap staining method and an automatic identification method for abnormal cervical cells, including two modules: module 1: training a cervical cell classification model based on a massive cervical cell data set; module 2: using the trained classification model to identify abnormalities cervical cells. The present invention proposes a novel Papanicolaou staining method, thereby well solving the problems existing in the traditional Papanicolaou staining method, and providing the possibility for automatic high-precision localization of cervical cell nuclei by computer. A method for automatic identification of abnormal cervical cells is also proposed. Experiments show that this method realizes ultra-high-precision automatic identification of abnormal cervical cells. Therefore, the present invention greatly reduces the diagnostic burden of pathologists and improves the diagnosis of cervical diseases. Diagnosis efficiency and accuracy have great practical value and huge social benefits.

Description

technical field [0001] The invention relates to the technical field of medical image diagnosis, in particular to an automatic identification method for abnormal cervical cells based on a novel Pap staining method. 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...

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