Training method and device for network model, image processing method and device, and device

Abstract

The present application discloses a training method and device for a network model, an image processing method and device, and equipment. The training method includes: obtaining three-dimensional summed sample features and multiple two-dimensional summed sample features through a feature transformation algorithm according to three-dimensional sample features and multiple two-dimensional sample features corresponding to multiple two-dimensional medical sample images, wherein the multiple Each 2D medical sample image has an overall label with or without lesions, the 3D summed sample features are combined with the 3D sample features transformed from multiple 2D sample features, and the multiple 2D summed sample features are combined with 3D samples Multiple 2D sample features converted from features; according to multiple 2D summed sample features and 3D summed sample features, the trained network model can be obtained, which can reduce the dependence on the accurate labeling of medical image lesions, and can The lesions are accurately and automatically classified, thereby greatly reducing the workload of the doctor's image annotation.

Description

technical field [0001] The present invention relates to the technical field of deep learning, in particular to a method and device for training a network model, a method and device for processing an image, and a device. Background technique [0002] The current lesion classification for medical images mostly relies on accurately labeled lesion images, that is, firstly, the lesion location is automatically detected or manually marked, and then the lesion is classified based on the image features of the lesion area. However, the performance of the current automatic lesion detection algorithm in medical images still does not meet the standard of auxiliary diagnosis, and the results of automatic lesion detection still require further examination and adjustment by experienced radiologists. Therefore, this does not significantly relieve the workload of image annotation. SUMMARY OF THE INVENTION [0003] In view of this, the embodiments of the present application aim to provide ...

AI Technical Summary

Problems solved by technology

However, the performance of the current automatic detection algorithm for medical imaging lesions is still not up to the standard of auxiliary diagnosis, and the results of automatic detection of lesions still need further inspection and adjustment by experienced radiologists

Therefore, this does not significantly ease the workload of image annotation

Images