Colposcopy image-based cervical cancer detection method, device and equipment and medium

Abstract

The invention discloses a colposcopy image-based cervical cancer detection method, device and terminal equipment and a computer readable storage medium. The method comprises the following steps of: ina bidirectional convolutional neural network-based cervical cancer detection model, positioning and extracting a cervical opening position of an obtained colposcopy image so as to generate an ROI image comprising a cervical cancer disease area; extracting a cutting edge of the ROI image through a bidirectional convolutional neural network so as to generate a cut image; and carrying out cancer grade classification on the cut image through a classification convolutional neural network so as to output a cervical cancer lesion grade and then rapidly and correctly obtain a cervical cancer detection result. According to the method, doctors that are lack of experiences can be helped to rapidly judge diseased regions, discover atypical diseased regions and judge the disease degrees and sampling regions, so that help and auxo-action are provided for discovering cervical cancer and precancerous lesions.

Description

technical field [0001] The present invention relates to the field of medical image processing, in particular to a colposcope image-based cervical cancer detection method, device, terminal equipment and computer-readable storage medium. Background technique [0002] Image segmentation algorithm already exists in the field of traditional digital image processing, mainly using the threshold segmentation method and its various variants. On the basis of the image segmentation algorithm, the image detection algorithm is extended, but they have different levels of image processing. The detection is mainly aimed at the image area level, while the segmentation needs to process each pixel, and the image processing is more delicate. In the field of deep learning, there are more and more segmentation algorithms based on convolutional networks, and the most widely used is the fully convolutional network (FCN). [0003] Existing cervical cancer pre-cancer screening is based on the "three...

AI Technical Summary

Problems solved by technology

However, the effect on segmenting small objects is poor. As the network level deepens, although a deeper network can better fit the sample distribution, the lack of information on small objects is more serious.

Images