MRI Pulse Sequence Adjustment for Compressed Sensing Undersampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current MRI systems face challenges in selecting an optimal undersampling factor for compressed sensing, leading to inefficient data acquisition times and potential image corruption due to manual adjustments based on operator experience, which lack consistency and accuracy.
Innovation Solution
A neural network is trained to predict the undersampling factor using magnetic resonance scan parameters, adjusting pulse sequence commands to optimize k-space sampling based on historical data and machine learning algorithms, reducing reliance on human expertise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If a low sampling rate is used for bandwidth efficiency, then energy consumption is reduced, but image quality deteriorates due to aliasing artifacts
Solution Approach 1:
The system performs preliminary actions by capturing metadata about the scene (motion characteristics, depth information, texture complexity) before the actual image capture. This metadata is used to pre-determine the optimal undersampling pattern, allowing the system to prepare the sampling strategy in advance based on scene analysis, thus enabling low sampling rates without quality loss
Solution Approach 2:
The system dynamically changes sampling parameters (sampling rate, pattern, orientation) based on scene characteristics. By adapting the sampling rate and pattern to match the actual scene content rather than using a fixed rate, the system achieves high image quality at lower average sampling rates, resolving the contradiction between energy efficiency and image quality
2Manufacturing precision
If traditional anti-aliasing filters are used to prevent aliasing, then image quality is maintained, but temporal resolution and frame rate are reduced
Solution Approach 1:
The system extracts only the essential frequency information needed to represent the scene accurately, separating the critical signal components from the redundant or less important ones. By taking out only the necessary information and discarding or coarsely sampling the rest, the system avoids the need for heavy anti-aliasing filtering while maintaining image quality and temporal resolution
Solution Approach 2:
The sampling rate and pattern are made dynamic rather than static. The system adjusts sampling parameters in real-time based on scene motion and content, allowing high frame rates for static or simple scenes while using lower rates for complex scenes, thus eliminating the need for conservative fixed-rate anti-aliasing approaches
3Manufacturing precision
If scene complexity increases, then more sampling points are needed to maintain quality, but energy consumption and processing load increase
Solution Approach 1:
The system applies different sampling densities to different regions of the scene based on local complexity. High-sampling areas are applied only to regions with fine details or motion, while low-sampling or skipped areas are used in uniform or static regions. This local adaptation maintains overall image quality while significantly reducing the total number of samples needed
Solution Approach 2:
The system uses partial sampling strategies where only a subset of pixels are fully sampled while others are interpolated or skipped. By applying sampling at partial density rather than uniform full density, the system achieves acceptable image quality with reduced energy consumption, especially in scenes where not all regions require maximum sampling detail
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Disclosed herein is a method of operating a medical system (100, 300). The method comprises receiving (200) pulse sequence commands (124) configured to control a magnetic resonance imaging system (302) to acquire k-space data (330) according to a Compressed Sensing magnetic resonance imaging protocol. The method further comprises receiving (202) magnetic resonance scan parameters that are descriptive of a configuration of the pulse sequence commands and a configuration of the magnetic resonance imaging system. The method further comprises receiving (204) an predicted undersampling factor (128) in response to inputting the magnetic resonance scan parameters into a neural network, wherein the neural network is configured to output the predicted undersampling factor in response to receiving magnetic resonance scan parameters. The method further comprises adjusting (206) the pulse sequence commands (130) to select or modify sampling of the k- space data based on the predicted undersampling factor.