Image feature extraction method based on deep learning

Abstract

The invention provides an image feature extraction method based on deep learning. The method comprises the following steps: receiving first input data representing an unannotated image of heart nuclear magnetic resonance from a heart nuclear magnetic resonance imaging system by a deep learning network; pre-processing the unannotated image by a deep learning network to generate second input data representing a saliency image of cardiac nuclear magnetic resonance and corresponding annotation data representing a saliency region of the saliency image; processing the first and second input data to perform training of a deep learning network by target feature detection in an unannotated image in a saliency region identified in a saliency image; and processing the third input data by using the trained deep learning network to identify a lesion structure in a new unannotated image. According to the image feature extraction method based on deep learning, a complex structure is not needed, training of a needed model is completed at a time, and heart disease variable structure positioning is carried out more quickly.

Description

technical field [0001] The invention relates to medical image processing, in particular to an image feature extraction method based on deep learning. Background technique [0002] Machine learning-based methods are able to automatically classify individuals. Corresponding to cardiac MRI images, there are characteristics such as small sample size and high feature dimension. Traditional machine learning methods only use simple shallow classification models, which cannot fully mine the advanced features of cardiac functional data. At the same time, these methods cannot effectively exploit the topological information in fNMR data. Therefore, the classification effect of fMRI data based on traditional machine learning methods needs to be further improved. For the deep learning method, although it has been well applied in the medical image classification session, there are many parameters, so it is easy to cause the problem of overfitting. It takes a lot of time to train the dee...

AI Technical Summary

Problems solved by technology

For the deep learning method, although it has been well applied in the medical image classification session, there are many parameters, so it is easy to cause the problem of overfitting. It takes a lot of time to train the deep learning model, and it is difficult to guarantee the current large-scale Time Efficiency in a Data Environment

Images