Training method and device for neural network model, storage medium and electronic device

Abstract

The invention provides a training method and device for a neural network model, a storage medium and an electronic device. The training method of the neural network model includes: extracting features from multiple modal images of the same body part to obtain corresponding multiple modal features; fusing the multiple modal features to obtain a first fusion vector; using a cyclic neural network method Perform feature aggregation and reconstruction on the first fusion vector to obtain a first feature vector; calculate a first loss function according to the first feature vector; train a neural network model according to the first loss function. The present invention obtains the first fusion vector by fusing multiple modal features, and uses a cyclic neural network to perform feature aggregation and reconstruction on the first fusion vector to obtain the first feature vector, so as to effectively mine multiple modal images. Correlation of modal features.

Description

technical field [0001] The invention relates to the technical field of medical imaging, in particular to a method and device for training a neural network model, a storage medium and an electronic device. Background technique [0002] In recent years, deep learning algorithms have been well applied in the field of image processing, which has inspired many researchers to apply this technology to medical image analysis. However, in medical image analysis, it is usually necessary to comprehensively analyze images of multiple modalities of the same body part, which increases the difficulty of applying deep learning algorithms to medical image analysis. [0003] In addition, although some researchers have used deep neural networks to train images of multiple modalities to assist doctors in making correct judgments, they usually use simple merging of the features of images of multiple modalities without considering Differences and commonalities between features in images of multi...

AI Technical Summary

Problems solved by technology

However, in medical image analysis, it is usually necessary to comprehensively analyze images of multiple modalities of the same body part, thus increasing the difficulty of applying deep learning algorithms to medical image analysis

[0003] In addition, although some researchers have used deep neural networks to train images of multiple modalities to assist doctors in making correct judgments, they usually simply combine the features of images of multiple modalities without considering the The differences and commonalities between features in images of multiple modalities, so the correlation between multiple modal images cannot be effectively mined

Images