Heart nuclear magnetic resonance image central ventricle myocardial segmentation model training method, segmentation method and device

Abstract

The invention provides a heart nuclear magnetic resonance image central ventricle myocardial segmentation model training method, segmentation method and device, preprocessing of a heart nuclear magnetic resonance image is carried out through employing contrast-limited adaptive histogram equalization and Gaussian blur, and the image contrast of a training sample is effectively improved, therefore, the recognition and segmentation effects are improved. Meanwhile, in the segmentation model training process, the dependency on training data can be effectively reduced through a transfer learning mode, and the recognition accuracy is improved under the condition of lacking the training data. Furthermore, a mixed loss function combining cross entropy loss, Dice loss and edge loss is adopted in training, the influence of unrelated backgrounds can be reduced, boundary contour features are concerned, the problem of data category imbalance is solved and the training result is more accurate under the condition of ensuring the stability of the model training process, and the demand on the number of training data is reduced to a certain extent.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a training method, a segmentation method and a device for cardiac nuclear magnetic resonance image ventricular myocardium segmentation model. Background technique [0002] Deep learning has made breakthroughs in various subfields of computer vision in recent years. It learns the human visual system to process external information by simulating neurons in human brain regions, automatically extracts multi-level features of images and maps images into high-level abstract feature spaces to achieve specified tasks. Due to its excellent image feature extraction ability, there are also a lot of related research in the field of medical image segmentation. In essence, medical image diagnosis is a problem of computer vision, but due to the special conditions in the process of medical image processing and the special needs of medical analysis, even mature computer vision t...

AI Technical Summary

Problems solved by technology

[0004] Embodiments of the present invention provide a model training method, segmentation method and device for myocardial segmentation of cardiac MRI images, to eliminate or improve one or more defects in the prior art, and to solve the problem caused by the lack of samples in medical images. The problem of low precision and poor generalization ability

Images