Multi-modal MRI conversion method and system based on conditional generative adversarial network, and medium

Abstract

The invention discloses a multi-modal MRI conversion method and system based on a conditional generative adversarial network, and a medium. The method comprises the steps of inputting an original MRIimage, inputting the original MRI image into an encoder of a conditional generative adversarial network to obtain a semantic feature map, and identifying a modal category of the original MRI image through a discriminator of the conditional generative adversarial network; and, for other modes except for the mode type of each original MRI image, generating a condition vector of the mode, connectingthe semantic feature map with the condition vector of the mode, and inputting a connected result into a decoder of the conditional generative adversarial network to obtain an MRI conversion map of themode, thereby finally obtaining MRI conversion maps of all other modes. Training can be carried out without supervision and registration of multi-modal images, it can be guaranteed that the multi-modal MRI generated through conversion is registered, it can be guaranteed that the MRI generated through conversion completely reserves key focus information, and inspection can be further carried out according to needs.

Description

technical field [0001] The invention belongs to the field of medical image processing, and in particular relates to a multimodal MRI conversion method, system and medium based on a conditional generative adversarial network, which is used to generate an MRI image through a conditional generative adversarial network according to an MRI image of a given modality and a target modality. Other registered multimodal MRI images. Background technique [0002] Magnetic resonance imaging (MRI) is a common medical image, and it can have multiple modalities according to different imaging parameters, such as T1, T2, T1c, etc. Different modalities have different reference values ​​for doctors, and doctors often need to compare images of multiple modalities to make a ready judgment. In the training and learning of intelligent processing tasks of medical images, we often expect to obtain more modal images, such as medical image processing tasks using convolutional neural networks (CNN) or ...

AI Technical Summary

Problems solved by technology

For doctors, it takes longer to obtain images of different modalities and requires the patient cooperation of patients. For researchers of intelligent processing tasks of medical images, multi-modal MRI datasets are very scarce and difficult to collect. Large, while the registered data is even rarer

Images