Cross-modal MRI synthesis method based on morphological feature GAN

Abstract

The invention discloses a cross-modal MRI synthesis method based on morphological feature GAN, including establishing an MRFE-GAN model, including a residual network module and a modal representative feature extraction module; the source mode obtains a pseudo-target mode through the residual network module; The representative features of the pseudo-target modality are extracted by the modality representative feature extraction module and combined with the basic information of the source modality, and the synthetic target modality is generated by fusion. The invention can obtain a more real and effective target mode; can effectively overcome the information difference between cross-domain modes, can effectively extract data of different levels; effectively reduce the difference between the synthetic mode and the real target mode, so that the synthetic mode Images are more realistic and reliable.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to a cross-modal MRI synthesis method based on morphological feature GAN. Background technique [0002] Medical imaging techniques such as computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET) are an important part of modern healthcare. MRI, which can capture contrast differences in soft tissue, has become the primary imaging modality for studying neuroanatomy. By applying different pulse sequences and parameters, a wide variety of tissue contrasts can be generated while imaging the same anatomical structure, resulting in images of different contrasts, ie MRI modalities. For example, by selecting pulse sequences such as magnetization-prepared gradient echo (MPRAGE) and recalled gradient of variation (SPGR), T1-weighted (T1) images can be generated that clearly delineate gray and white matter tissue. In contrast, T2-weighted (T2...

AI Technical Summary

Problems solved by technology

However, obtaining multiple different contrast images (or modalities) for the same detector is time-consuming and expensive

So in practice, the number of contrast agent modalities acquired for the same patient is always limited due to related factors such as limited scan time and expensive cost

[0003] Although there are currently some synthesis methods for modality synthesis through deep networks, due to the large difference in feature distribution between source and target modalities, even existing methods can understand the mapping between different modalities to a certain extent, but the input There may still be some differences between the generated image and the difference cannot be effectively reduced, resulting in a large difference between the synthetic image and the real image

Images