A method and system for classifying images of irregular cervical cell clusters

Abstract

The invention discloses a method for classifying images of irregular cervical cell clusters, which is characterized in that it comprises the following steps: establishing a training sample set of suspicious cell clusters off-line, and using a multi-resolution input three-channel neural network model to train suspicious diseased cell clusters to determine Model; extract a single cell cluster area on the cervical pathological slice image, and apply the trained suspicious cell cluster judgment model to judge the abnormality of each cell cluster; mine the cell clusters that cannot be correctly judged, and input them as training data for key training of the model . The invention takes the irregular cell group as the processing and identification unit, uses a multi-resolution input three-channel neural network model to quickly identify suspicious diseased cell groups, and simultaneously improves the recognition accuracy and recognition efficiency.

Description

technical field [0001] The invention belongs to the field of medical cytopathological image processing, and more particularly relates to a method and system for classifying images of irregular cervical cell clusters. Background technique [0002] Cervical cancer is a malignant tumor with high incidence in women. Cervical liquid-based cytopathology is currently the most important means of preventing and screening cervical cancer. Accurate interpretation of diseased cells in cytopathological slice images is an important basis for doctors to determine the patient's condition and formulate a treatment plan. Since a cytopathological slide contains many fields of view, manual interpretation of cytopathological images is very time-consuming. Therefore, automatically and quickly identifying suspicious areas in slices for further interpretation by doctors can greatly improve the efficiency of doctors' diagnosis. This intelligent assisted film reading technology is of great signifi...

AI Technical Summary

Problems solved by technology

However, this block fusion scheme has the following disadvantages: the cells at the boundaries of the sub-blocks are artificially cut, reducing the recognition accuracy; too large a block is still difficult to deal with directly, and too small a block brings more boundary problems and increases processing overhead ;Inconsistency in processing results of adjacent blocks at subblock boundaries

Although traditional image processing and feature extraction methods can be used to identify suspicious lesion areas in slices, traditional image features or custom image features are difficult to completely characterize cell morphology, and usually only part of the cell morphology can be described, such as nuclear size, depth, etc. dyed

Images