Automatic detection method of abnormal cervical cells based on depth convolution neural network

Abstract

The invention discloses an automatic detection method of abnormal cervical cells based on a depth convolution neural network, which mainly comprises the following steps: (1) manual labeling of a TCT cervical cell picture; (2) segmentation of labeled cell images; (3) multi-classification of cells after segmentation; (4) training of cell images in migration model; (5) segmentation, recognition and localization of the pictures of the cells to be detected; (6) re-segmentation and recognition of abnormal cell clusters. The automatic detection method for abnormal cervical cells of the invention classifies the divided cells into nine types, trains the cells by migration learning mode, and obtains a better fitting multi-classification model. The model can identify, screen and accurately locate theunlabeled source cervical cell images, and re-segment and identify the abnormal cells, which improves the detection accuracy. The invention has good auxiliary effect in the field of cervical cell pathological diagnosis.

Description

technical field [0001] The invention belongs to the field of pathological diagnosis of cervical cells, and in particular relates to an automatic detection method for abnormal cervical cells based on a deep convolutional neural network. Background technique [0002] In recent years, the high incidence of cervical cancer has become a social problem that threatens women's lives. The current effective diagnostic method for cervical cancer is cervical smear pathological examination, which requires a pathologist to make a diagnosis after finding diseased cells through microscope observation. On the one hand, this requires a lot of manpower and material resources. On the other hand, the accuracy of diagnosis is easily affected by the doctor's subjective factors or visual fatigue. Therefore, automatic cervical cytopathological diagnosis technology becomes more and more important. [0003] Existing cervical cell recognition technology usually focuses on fine segmentation and featur...

AI Technical Summary

Problems solved by technology

[0003] Existing cervical cell recognition technology usually focuses on fine segmentation and feature extraction. However, due to the existence of problems such as overlapping, irregularity, and uneven staining of cell images collected in the microscope, fine cell segmentation has brought great difficulties. , at the same time, in the process of feature extraction, there will be problems that effective features cannot be extracted or too many invalid features are introduced, and good results cannot be achieved.

[0004] For the existing automatic detection technology of abnormal cervical cells, most of them use negative and positive binary classification for identification and detection. Due to the similarity between abnormal cervical cells and basal cells and the indistinguishability of cell clusters, the classification effect is not ideal

Moreover, the current cell classification is mainly aimed at single cells, and the cell clusters have not been processed too much, resulting in unsatisfactory detection results of cell clusters.

At the same time, the existing automatic identification methods for abnormal cervical cells are mainly based on classification, and the input is required to be a segmented image. The initial collection of cervical cells cannot be directly processed, and the direct purpose of identifying and locating abnormal cervical cells cannot be achieved.

Images