Search

KR-20260062778-A - APPARATUS AND METHOD FOR RESTORING A 3D MAGNETIC RESONANCE IMAGING INCLUDING BLURRING INTO A CLEAN MAGNETIC RESONANCE IMAGE WITH THE BLURRING REMOVED

KR20260062778AKR 20260062778 AKR20260062778 AKR 20260062778AKR-20260062778-A

Abstract

The present invention relates to a clean MRI restoration device for restoring a 3D magnetic resonance imaging (MRI) containing deblurring caused by off-resonance effects into a clean MRI with the deblurring removed, comprising: a preprocessing unit that converts the 3D MRI into a 2D image to generate a preprocessed image; and an image restoration unit that inputs the preprocessed image into a pre-prepared model to extract and emphasize frequency features in the frequency domain and image features in the spatial domain, and removes the deblurring of the preprocessed image based on the emphasized frequency and image features to output the clean MRI.

Inventors

  • 정희철
  • 안재신

Assignees

  • 경북대학교 산학협력단

Dates

Publication Date
20260507
Application Date
20250123
Priority Date
20241029

Claims (10)

  1. A clean MRI restoration device that restores a 3D magnetic resonance imaging (MRI) containing deblurring caused by off-resonance effects into a clean MRI with the blurring removed. A preprocessing unit that converts the above 3D MRI into a 2D image to generate a preprocessed image; A clean MRI restoration device comprising: an image restoration unit that inputs the above-mentioned preprocessed image into a pre-prepared model to extract and emphasize frequency features in the frequency domain and image features in the spatial domain, and removes blurring of the above-mentioned preprocessed image based on the emphasized frequency and image features to output the clean MRI.
  2. In paragraph 1, The above-mentioned pre-prepared model is, At least one GDoT (Gated Dual Domain Transformer) block that extracts and emphasizes frequency and image features from the above-mentioned preprocessed image, and It includes at least one transformer block that removes blurring of the preprocessed image based on the emphasized frequency and image features, A clean MRI reconstruction device comprising a GDoT block composed of a GFP (Gated Frequency Projection) block that extracts and emphasizes frequency features and a GSP (Gated Spatial Projection) block that extracts and emphasizes image features.
  3. In paragraph 2, The above GFP block generates a query using the frequency features, and the above GSP block generates a key and a value using the image features, and A clean MRI restoration device characterized by the above transformer block removing the blurring using an attention map generated through the above query and key.
  4. In paragraph 1, A clean MRI reconstruction device, wherein the above-described pre-prepared model is trained based on a first loss function that corrects the difference between the above-described pre-processed image and a ground truth image that does not include blurring, and a second loss function that corrects the difference between the above-described clean MRI and the above-described ground truth image.
  5. In paragraph 1, A clean MRI restoration device further comprising: an artifact removal unit that inputs a plurality of input images, obtained by pixel-transforming the preprocessed image in a predetermined direction, into a pre-prepared model, and selects a median of a plurality of output images output from the model to remove checkerboard artifacts appearing as a regular and fixed pattern in the preprocessed image.
  6. A clean MRI restoration method for restoring a 3D magnetic resonance imaging (MRI) containing deblurring caused by off-resonance effects into a clean MRI with the blurring removed, A step in which a preprocessing unit converts the 3D MRI into a 2D image to generate a preprocessed image; A clean MRI restoration method comprising the step of: inputting the preprocessed image into a pre-prepared model to extract and emphasize frequency features in the frequency domain and image features in the spatial domain, and removing blurring of the preprocessed image based on the emphasized frequency and image features to output the clean MRI.
  7. In paragraph 6, The above-mentioned pre-prepared model is, At least one GDoT (Gated Dual Domain Transformer) block that extracts and emphasizes frequency and image features from the above-mentioned preprocessed image, and It includes at least one transformer block that removes blurring of the preprocessed image based on the emphasized frequency and image features, A clean MRI reconstruction method comprising a GDoT block composed of a GFP (Gated Frequency Projection) block that extracts and emphasizes frequency features and a GSP (Gated Spatial Projection) block that extracts and emphasizes image features.
  8. In Paragraph 7, The above GFP block generates a query using the frequency features, and the above GSP block generates a key and a value using the image features, and A clean MRI restoration method characterized by removing blurring using an attention map generated through the above-described query and key, wherein the above-described transformer block.
  9. In paragraph 6, A clean MRI restoration method, wherein the above-described pre-prepared model is trained based on a first loss function that corrects the difference between the above-described pre-processed image and a ground truth image that does not include blurring, and a second loss function that corrects the difference between the above-described clean MRI and the above-described ground truth image.
  10. In paragraph 6, A clean MRI restoration method further comprising the step of: inputting a plurality of input images, in which an artifact removal unit has pixel-transformed the preprocessed image in a predetermined direction, into a pre-prepared model, and selecting a median of a plurality of output images output from the model to remove checkerboard artifacts that appear as a regular and fixed pattern in the preprocessed image.

Description

Apparatus and method for restoring a 3D magnetic resonance image including blurring into a clean MRI with the blurring removed The present invention relates to an apparatus and method for restoring a 3D magnetic resonance imaging (MRI) containing blurring into a clean MRI with the blurring removed. Magnetic Resonance Imaging (MRI) is widely used in clinical diagnosis due to its ability to visualize internal body structures. However, MRI requires long scan times to fully collect signals in the frequency domain (k-space), and such prolonged scanning causes discomfort to patients and incurs significant costs. Consequently, active research into technologies to accelerate MRI is currently underway. Among the various approaches to accelerate MRI, non-Cartesian locus is the most widely used because it effectively reduces scan time compared to Cartesian locus by collecting sparse k-space data. However, non-Cartesian locus is prone to off-resonance artifacts due to non-ideal conditions, which manifest as complex phase accumulation in the frequency domain. Meanwhile, deep learning has recently demonstrated outstanding performance in medical imaging and computer vision, particularly in low-level vision tasks such as image noise removal, deblurring, and image super-resolution. Following this success, deep learning-based methods for removing non-resonant artifacts have been developed. However, this method relies on the existing CNN (Convolutional Neural Network) architecture, which can lead to a non-optimal solution and has the problem of not being able to utilize cutting-edge deep learning technology. Therefore, to overcome these limitations of existing technology, research is needed on methods to effectively remove non-resonance artifacts. FIG. 1 is a diagram illustrating the internal blocks of a clean MRI restoration device according to an embodiment of the present invention. FIG. 2 is a diagram for explaining the overall operation of the clean MRI restoration device of FIG. 1. FIG. 3 is a drawing showing detailed blocks of a pre-prepared model applied in the image restoration unit of FIG. 1. FIG. 4 is a drawing showing another example of the clean MRI restoration device of FIG. 1. FIG. 5 is a diagram for explaining the operation of the artifact removal unit of FIG. 4. And, FIG. 6 is a flowchart showing the operation of a clean MRI restoration device according to an embodiment of the present invention. The following detailed description of the invention refers to the accompanying drawings, which illustrate specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that various embodiments of the invention are different but need not be mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the invention in relation to one embodiment. It should also be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the following detailed description is not intended to be limiting, and the scope of the invention is limited only by the appended claims, including all equivalents to those claimed therein, provided appropriately described. Similar reference numerals in the drawings refer to the same or similar functions across various aspects. The components according to the present invention are defined by functional distinction rather than physical distinction, and can be defined by the functions each performs. Each component may be implemented as hardware or as program code and processing units that perform each function, and the functions of two or more components may be included and implemented in a single component. Therefore, it should be noted that the names assigned to the components in the following embodiments are not intended to physically distinguish each component but are assigned to imply the representative function performed by each component, and that the technical concept of the present invention is not limited by the names of the components. Preferred embodiments of the present invention will be described in more detail below with reference to the drawings. FIG. 1 is a diagram illustrating the internal blocks of a clean MRI restoration device according to an embodiment of the present invention, FIG. 2 is a diagram explaining the overall operation of the clean MRI restoration device of FIG. 1, and FIG. 3 is a diagram showing detailed blocks of a pre-prepared model applied in the image restoration unit of FIG. 1. The described clean MRI restoration device refers to a device that restores a 3D MRI containing deblurring, which is one of the non-resonance artifacts caused by non-resonance effects, into a clean MRI wi