Head three-dimensional MRI super-resolution reconstruction method

Abstract

The invention discloses a head three-dimensional MRI super-resolution reconstruction method, which comprises the following steps: firstly, selecting high-resolution MRI image data, preprocessing the selected MRI data to obtain small data blocks, and taking the small data blocks as training data; then establishing a three-dimensional MRI super-resolution reconstruction model based on a convolutional neural network and a Transform encoder, and training the model by taking the small data blocks as input to obtain the three-dimensional MRI super-resolution reconstruction model; and finally, processing the three axial MRI data to obtain cube cavity data, inputting the cavity data into the trained model to obtain small cubes with filled cavities, and splicing the small cubes with filled cavities to form a complete data space. The method is small in model scale, high in image reconstruction speed and high in reconstructed image quality, and provides more precise data for medical image processing.

Description

technical field [0001] The invention relates to the technical field of medical image processing, in particular to a head three-dimensional MRI super-resolution reconstruction method. Background technique [0002] During magnetic resonance imaging (MRI), the image signal-to-noise ratio and spatial resolution suffer due to limitations of the image acquisition sensor. On the other hand, due to the limitation of image data storage space, some image data that are not closely related to diagnostic information are omitted during image acquisition to reduce the storage pressure of the hospital. For example, in the MRI diagnosis of nasopharyngeal carcinoma, doctors often obtain clinical image information from three views: coronal plane, sagittal plane and transverse plane. In order to save storage space, the on-chip resolution of this type of image data is high to meet the needs of doctors for diagnosis, and the image layer spacing is relatively large to save storage space. The laye...

AI Technical Summary

Problems solved by technology

In the image processing task using only one axial data, for the current medical image machine vision task, in the frequently used 3D convolution operation, due to the influence of the image layer spacing, the real space covered by each convolution kernel It is not a cube, but a cuboid with a long side. The spatial resolution of the images extracted by the convolution kernel in different axes is different, resulting in a decline in model performance.

When using three-axis data for image processing tasks, if the image information of the three azimuth planes cannot be effectively integrated, it will affect subsequent image processing tasks, such as affecting the experimental accuracy of image segmentation and disease classification.

Images