Deep learning model training method and device, image processing method and device

Abstract

The invention discloses a deep learning model training method and device. The method comprises the following steps: the deep learning model is trained through the chest medical image marked with the frosted glass nodule, the trained deep learning model is obtained, the frosted glass nodule segmentation efficiency is improved through the trained deep learning model, in addition, the segmentation accuracy is effectively improved, and the segmentation process of the frosted glass nodule is simplified. Furthermore, a multi-stage feature fusion monitoring module is added to the deep learning model,so that the response of the deep learning model to the focus area in the image is enhanced, the whole image is processed, and the operation process is simplified; moreover, an autonomous grading supervision module is also added, so that the deep learning model can autonomously adjust the influence of the features of different levels on the segmentation result. In addition, the decoded feature mapis subjected to up-sampling operation, so that the ground glass nodules can be segmented more accurately by the deep learning model.

Description

technical field [0001] The present invention relates to the field of image processing, in particular to a training method for a deep learning model, an image processing method and a device. Background technique [0002] During CT examination, there may be cases of increased density, or cloud-like thin shadows, or round nodules, which look like ground glass in shape, so they are called ground glass nodules. Festival. Growth prediction of ground glass nodules can help doctors predict cancer. [0003] The current ground glass junction segmentation methods usually use traditional segmentation methods, such as threshold processing or region growing methods, but traditional segmentation methods are easily affected by noise, and the accuracy of segmentation results is not high. Contents of the invention [0004] In view of this, embodiments of the present invention provide a deep learning model training method, image processing method and device, which realize automatic segment...

AI Technical Summary

Problems solved by technology

[0003] The current ground glass junction segmentation methods usually use traditional segmentation methods, such as threshold processing or region growing methods, but traditional segmentation methods are easily affected by noise, and the accuracy of segmentation results is not high

Images