Method, system and medium for generating registered multimodal MRI with lesion segmentation labels

Abstract

The invention discloses a method, system and medium for generating a registered multi-modal MRI with a lesion segmentation label. The method of the invention includes obtaining a random matrix with a normal distribution and inputting the trained structural feature decoding in the generating confrontation network The structural feature map is generated by decoding the device; the structural feature map is fused with the randomly selected lesion segmentation label map through random input to obtain the fusion result, and input into the trained random encoder in the generative adversarial network to obtain encoding; the encoded input is trained in the generative adversarial network A good decoder for each modality generates the registered multimodal MRI respectively. The invention modularizes the generator in the generative confrontation network into an encoder and a decoder, and through the combined training of multiple sets of encoders, decoders and discriminators, it can receive a random input that meets the design specifications and then generate a group of segmentations with lesions The multi-modal MRI image of the label registration can be widely used in the field of medical imaging.

Description

technical field [0001] The present invention relates to image generation technology in the medical field, in particular to a method, system and medium for generating a registered multimodal MRI with a lesion segmentation label, which is used to obtain a registered image according to a given random input that meets the specification Multimodal MRI images with lesion segmentation labels. Background technique [0002] With the development of deep learning, more and more fields have begun to use deep neural networks for image processing tasks. However, the training of deep neural networks requires a large amount of data, but in many fields, the construction of data sets is very difficult. Therefore, image generation technology has important uses in image intelligent processing scenarios in many fields, such as medical imaging and biological cell imaging. In the intelligent processing of medical images, medical images have many modalities, such as magnetic resonance imaging (MR...

AI Technical Summary

Problems solved by technology

However, medical image collection is very difficult, especially for rare diseases, making medical image datasets scarce and small, which makes many training tasks impossible

Naturally, registered multimodal image data is even scarcer

Images