Neural Network Motion Correction in MRI K-Space Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Patient motion during MRI scans leads to inefficiencies, requiring re-scans and increasing costs due to blurriness and artifacts in MR images, as existing methods for motion correction are either costly, time-consuming, or limit data acquisition.
Innovation Solution
Divide k-space data into portions based on motion timing and use a neural network to transform motion-corrupted sub-images into motion-corrected images by identifying a dominant pose and aggregating data from other poses, employing a deep-learning neural network for image reconstruction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If hardware for monitoring motion is added, then motion detection capability is improved, but cost and patient setup time increase
Solution Approach 1:
The system uses the MRI scanner's own acquisition data and neural network to automatically detect and correct motion, eliminating the need for external monitoring hardware. The scanner corrects motion artifacts using its inherent capabilities and processed k-space data, making the system self-sufficient without additional equipment.
Solution Approach 2:
The patent replaces physical motion monitoring hardware with a computational approach using neural networks and image processing algorithms. Instead of mechanical sensors and hardware monitors, the system uses software-based detection and correction methods to handle motion artifacts.
2Reliability
If navigator sequences are used, then motion information is obtained, but imaging sequence time is reduced
Solution Approach 1:
The patent combines motion detection and image reconstruction into a single integrated process. The neural network processes k-space data while the imaging sequence continues, merging what would traditionally be separate steps (navigator sequences for motion detection and main imaging sequences) into one simultaneous operation.
Solution Approach 2:
The system continuously processes motion correction throughout the imaging sequence rather than performing separate pre-scan navigator sequences. The neural network operates concurrently with data acquisition, maintaining continuous useful action without interrupting the imaging timeline.
3Reliability
If particular types of acquisitions are used, then motion correction is enabled, but type of information that can be collected is limited
Solution Approach 1:
The neural network is trained to handle various motion scenarios and acquisition types by processing k-space data with different motion characteristics. The system adapts to different acquisition parameters and motion patterns through the learned representations, maintaining versatility across different imaging protocols without requiring protocol-specific correction methods.
4Measurement precision
If k-space data is divided into portions based on timing, then motion correction accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent segments k-space data into different portions based on timing information to separate data from different motion states. This segmentation allows the neural network to process and correct motion artifacts more accurately by working with discrete time-based data portions rather than attempting to correct all data simultaneously.
Solution Approach 2:
The neural network acts as an intermediary that processes the segmented k-space data and produces motion-corrected images. The network mediates between the raw segmented data and the final corrected output, handling the computational complexity through learned transformations rather than explicit complex algorithms.
Data Source
AI summary
K-space data obtained from a magnetic resonance imaging scan where motion was detected is split into two parts in accordance with the timing of the motion to produce first and second sets of k-space data corresponding to different poses. Sub-images are reconstructed from the k first and second sets of k-space data, which are used as inputs to a deep neural network which transforms them into a motion-corrected image.