Identification method for atypical abnormal glandular cell in uterine neck cell smear

Abstract

The invention provides an identification method for an atypical abnormal glandular cell in a uterine neck cell smear. According to the method, glandular cells are divided into glandular cell clustersof clear structure, indivisible glandular cell clusters and single glandular cells; for the glandular cell cluster of clear structure, arrangement information and appearance information of the glandular cells serve as two inputs of a double flow neural network model; the single glandular cell and the indivisible glandular cell cluster are input to a convolutional neural network model; and the twomodels output an identification result. The method can be used to improve the identification effect for the glandular cells with obvious arrangement structures, identification is more complete, and misidentification and identification omission are reduced.

Description

technical field [0001] The invention relates to the field of medical cell image processing, in particular to a method for identifying atypical abnormal gland cells in cervical cell smears. Background technique [0002] Cervical cancer is a preventable and curable disease. Recent studies have shown that the incidence of adenocarcinoma and squamous cell carcinoma has doubled, especially in young women. The incidence of adenocarcinoma has increased by 49.3% compared with the previous proportion of 10-15% of cervical cancer. In the existing clinical diagnosis process, the analysis methods of cervical cell images, mainstream screening imaging technologies such as TCT and SurePath completely rely on the judgment of the film reader based on personal experience, and usually can only draw qualitative conclusions about the presence or absence of lesions . Due to the uneven level of film-reading doctors, the visual fatigue caused by high-intensity film-reading work, and the repeated...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to overcome the current problem of low recognition rate in the detection of atypical glandular epithelial cells, and provide a method for identifying atypical abnormal glandular cells in cervical cell smears. By fully analyzing the appearance characteristics of cervical glandular cells, using Graph theory method, deep learning, image understanding technology, combined data-driven method and cervical cell field knowledge mechanism, finally realized the rapid identification of cervical gland cells, and gave the visualization of the arrangement information of the identification results of gland cells, for the improvement of cervical cell artificial intelligence Intelligent Assisted Screening Provides Important System Framework

Images