Deep learning-based cervical cancer TCT slice negative exclusion method and system

Abstract

The invention provides a deep learning-based cervical cancer TCT slice negative exclusion method, and the method comprises the following steps: carrying out the segmentation preprocessing of a cervical cancer TCT image, and segmenting a target image into a plurality of image blocks with partial overlapping regions; detecting each image block by using a target detection model, and outputting a cervical cancer positive target detection frame under the image block and a corresponding detection frame confidence coefficient; setting a detection frame confidence threshold, and screening and reserving the detection frames with the detection frame confidence greater than the detection frame confidence threshold based on the detection frame confidence threshold; detecting the detected target images in the reserved detection frame by using a negative discharge classification model, outputting positive probabilities of all the detected target images, and calculating an overall average positive probability; and setting a negative exclusion probability threshold value, if the overall average positive probability is smaller than or equal to the preset negative exclusion probability threshold value, determining that the target image is overall negative, otherwise, determining that the target image is an image to be manually checked, and manually checking the image by a doctor.

Description

technical field [0001] The present invention relates to the field of medical image analysis, in particular to a method and system for cervical cancer TCT slice vaginal discharge based on deep learning. Background technique [0002] At present, cervical cancer is one of the most common malignant tumors in women. There are about 130,000 new cases in China every year. The incidence rate is second only to breast cancer among female malignant tumors. Thinprep Cytologic Test (TCT) can achieve the purpose of early diagnosis and early treatment of patients. Doctors need to conduct cytological examination on TCT slides of each patient, which is the most important cervical cancer screening at present. means. According to statistics, the population of cervical cancer screening in my country is as high as 350 million, but there are currently less than 20,000 registered pathologists in China, and it takes 5 to 10 years to train an experienced pathologist. The huge shortage of doctors le...

Images