Non-Cartesian MRI Artifact Correction with Tissue-Mixing Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Non-Cartesian MRI is susceptible to chemical shift artifacts leading to blurring and destructive interference at fat-water tissue interfaces, which cannot be effectively mitigated at lower bandwidths without increasing noise and scan time.
Innovation Solution
A deep learning-based approach using a trained neural network with a tissue mixing model to simulate fat-water interactions, correcting chemical shift artifacts in MRI data by modeling interactions between different tissue types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If scanning is performed at high readout bandwidth (e.g., 400 Hz/pixel) to mitigate chemical shift artifacts, then image quality is improved, but noise (acoustic and thermal) increases and gradient hardware requirements increase
Solution Approach 1:
A deep learning model is introduced as an intermediary between the acquired MRI data and the final reconstructed image. The model learns to predict and remove chemical shift artifacts from images acquired at lower bandwidths, effectively mediating between the low-bandwidth input and high-quality output without requiring the system to operate at high bandwidth during acquisition.
Solution Approach 2:
The patent replaces the traditional mechanical/hardware-based approach of using high readout bandwidth and strong gradient fields to mitigate chemical shift artifacts with a software-based deep learning approach. Instead of relying on physical system parameters (bandwidth, gradient amplitude), the solution uses an artificial intelligence model to computationally correct the artifacts.
2Manufacturing precision
If scanning is performed at high readout bandwidth to mitigate chemical shift artifacts, then image quality is improved, but scan time increases due to gradient heating limitations
Solution Approach 1:
The deep learning model serves as a post-processing intermediary that corrects chemical shift artifacts in images acquired at lower bandwidths. This allows the system to maintain shorter scan times by operating at lower bandwidth during acquisition while still achieving high image quality through the AI-based correction step.
Solution Approach 2:
The deep learning model is trained in advance on a large dataset of MRI images with and without chemical shift artifacts. This preliminary training enables the model to quickly and accurately correct artifacts during clinical use without requiring time-consuming processing during the actual scan, thus preserving short scan times.
3Productivity
If non-Cartesian MRI scanning is used, then scanning efficiency is improved, but chemical shift artifacts increase leading to blurring and destructive interference
Solution Approach 1:
The deep learning model acts as a post-processing intermediary specifically designed to correct chemical shift artifacts in non-Cartesian MRI images. It takes the efficiently acquired but artifact-contaminated images as input and outputs corrected images with restored quality, enabling the system to benefit from both the efficiency of non-Cartesian scanning and the quality of artifact-free images.
Solution Approach 2:
The patent changes the approach from trying to prevent chemical shift artifacts during acquisition (which would require high bandwidth and strong gradients) to correcting them after acquisition using deep learning. This parameter change in the processing workflow allows non-Cartesian scanning to proceed at efficient low bandwidths while achieving high image quality through the AI correction step.
Data Source
AI summary
A computer-implemented method for generating a chemical shift artifact corrected reconstructed image from magnetic resonance imaging (MRI) data includes inputting into a trained deep neural network an image generated from the MRI data acquired during a non-Cartesian MRI scan of a subject. The method also includes utilizing the trained deep neural network to generate the chemical shift artifact corrected reconstructed image from the image, wherein the trained deep neural network was trained utilizing a tissue mixing model that models interactions between different tissue types to mitigate chemical shift artifacts. The method further includes outputting from the trained deep neural network the chemical shift artifact corrected reconstructed image.