Active learning-based cervical cancer identification model training method

Abstract

The invention discloses an active learning-based cervical cancer identification model training method. The method comprises the following steps of preparing multiple cervical cancer section digital images, constructing a sample library, and training an identification model by adopting the cervical cancer section digital images in the sample library; for the cervical cancer section digital images in the sample library, outputting corresponding identification results through identification of the identification model, wherein the identification results comprise cells difficultly identified by the identification model; and in the identification results, tagging the difficultly identified cells in a manual tagging mode, updating the tagged cervical cancer digital images to the sample library,and re-training the identification model for performing identification, until the difficultly identified cells do not exist in all the identification results. Compared with the prior art, the method has the beneficial effects of high identification speed, high identification precision and capability of automatically filtering a large amount of cervical cancer cells with remarkable features.

Description

technical field [0001] The invention relates to the field of cancer recognition, in particular to a method for training a cervical cancer recognition model based on active learning. Background technique [0002] Cervical cancer is a malignant tumor that seriously endangers women's reproductive health. According to incomplete statistics, there are nearly 500,000 new cases of cervical cancer in the world every year, and about 132,000 new cases of cervical cancer in my country each year, accounting for about 100,000 people in the world. 28% of the total. In order to improve the accuracy of cervical cancer diagnosis, in recent years, many medical research teams at home and abroad have begun to study the identification methods of various cervical cancer cells, but most of the identification algorithms at home and abroad still use supervised identification models, that is to say However, this recognition model still needs a large number of manual annotations by annotating doctors ...

AI Technical Summary

Problems solved by technology

In order to improve the accuracy of cervical cancer diagnosis, in recent years, many medical research teams at home and abroad have begun to study the identification methods of various cervical cancer cells, but most of the identification algorithms at home and abroad still use supervised identification models, that is to say , this recognition model still needs a large number of manual annotations by annotating doctors to obtain a high recognition rate of cervical cancer cells

[0003] However, due to the high resolution of digital images of cervical...

Images