Neural Network Reconstruction for MRI Truncation Artifacts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Magnetic resonance imaging (MRI) often results in images contaminated by truncation artifacts such as blurring and ringing, due to partial k-space sampling, which degrades the diagnostic value of the images.
Innovation Solution
A computer-implemented method using a neural network model trained with pairs of pristine and corrupted images, where the corrupted images are based on partial k-space data truncated at high spatial frequencies. The neural network analyzes the crude images and outputs improved images with reduced truncation artifacts and increased high spatial frequency data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If partial k-space sampling is used to increase acquisition efficiency, then productivity is improved, but manufacturing precision deteriorates due to truncation artifacts
Solution Approach 1:
A neural network model is introduced as an intermediary between the partially-sampled k-space data and the final image reconstruction. The neural network learns to map truncated k-space data to high-quality images by training on pairs of fully-sampled and partially-sampled data, effectively mediating the transition from efficient but artifact-prone acquisition to diagnostic-quality imaging without requiring full k-space sampling
Solution Approach 2:
The patent changes the sampling parameters in k-space by using asymmetric truncation patterns rather than uniform sampling. By selectively sampling specific k-space locations and using neural network-based reconstruction, the system achieves both high acquisition efficiency and diagnostic image quality, resolving the contradiction between sampling efficiency and image quality
2Object-affected harmful factors
If partial k-space sampling is used to suppress artifacts, then object-affected harmful factors are reduced, but object-generated harmful factors increase due to truncation artifacts
Solution Approach 1:
The patent converts the harmful truncation artifacts into beneficial information by training the neural network on pairs of fully-sampled and partially-sampled k-space data. The network learns to recognize and correct the specific patterns of truncation artifacts, transforming what was previously harmful signal loss into an opportunity for artifact suppression and image enhancement through learned reconstruction
3Loss of information
If asymmetric truncation is applied at high spatial frequencies, then loss of information is reduced in low spatial frequencies, but measurement precision deteriorates at high spatial frequencies
Solution Approach 1:
The patent applies preliminary action by pre-training the neural network on comprehensive datasets that include both fully-sampled and partially-sampled k-space data. This preliminary training enables the network to learn the relationships between different spatial frequency components and how to reconstruct high spatial frequency information from truncated data, allowing the system to prioritize low spatial frequency information during asymmetric sampling while still recovering high spatial frequency details through the learned model
Data Source
Figure 1
Figure 2A~2B
Figure 2C
AI summary
A computer-implemented method of removing truncation artifacts in magnetic resonance (MR) images is provided. The method includes receiving a crude image that is based on partial k-space data from a partial k-space that is asymmetrically truncated in at least one k-space dimension. The method also includes analyzing the crude image using a neural network model trained with a pair of pristine images and corrupted images. The corrupted images are based on partial k-space data from partial k-spaces truncated in one or more partial sampling patterns. The pristine images are based on full k-space data corresponding to the partial k-space data of the corrupted images, and target output images of the neural network model are the pristine images. The method further includes deriving an improved image of the crude image based on the analysis, wherein the derived improved image includes reduced truncation artifacts and increased high spatial frequency data.