MRI Susceptibility Map Generation via Edge-Aware L1 Regularization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for generating susceptibility maps from magnetic resonance imaging (MRI) data face challenges such as ill-posed problems due to zero denominators in Fourier transforms, leading to inaccurate susceptibility calculations, especially at tissue boundaries and regions with shading artifacts.
Innovation Solution
The MRI apparatus employs a combination of sequence control and processing circuitry to generate high-precision susceptibility maps by performing phase-wrapping removal, background magnetic field correction, and using L1 norm regularization with edge detection in the R2* map to minimize gradient smoothing at tissue boundaries, thereby reducing shading artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If L1 norm regularization with gradient of susceptibility is used to solve the ill-posed problem, then the susceptibility calculation becomes stable, but the susceptibility at tissue boundaries is smoothed out and becomes inaccurate
Solution Approach 1:
The patent applies different processing strategies to different regions: in non-edge regions, L1 norm regularization is applied to ensure stability, while in edge regions detected from magnitude images, the regularization is modified or reduced to preserve sharp susceptibility transitions and avoid smoothing artifacts at tissue boundaries
2Measurement precision
If edge detection is performed in magnitude image to prevent smoothing at tissue boundary, then susceptibility accuracy at boundary is improved, but pseudo edges are generated in regions with shading artifacts
Solution Approach 1:
The patent introduces an intermediary step where edges are detected not directly from the magnitude image alone, but from a processed version that combines magnitude image information with phase image information or uses a different processing approach to distinguish true tissue boundaries from shading artifacts, thereby avoiding pseudo-edge generation
3Reliability
If conventional L1 norm regularization is applied, then the ill-posed problem is solved, but susceptibility cannot be accurately calculated in regions with low signal reliability
Solution Approach 1:
The patent applies region-specific processing where susceptibility calculation is performed differently in high-signal-reliability regions versus low-signal-reliability regions, using the detected edges to identify and separately process problematic regions to maintain accuracy where possible while ensuring stability overall
Data Source
AI summary
In general, according to the present embodiment, a magnetic resonance imaging apparatus includes sequence control circuitry and processing circuitry. The sequence control circuitry collects MR data corresponding to each of a plurality of echo times. The processing circuitry generates a plurality of magnitude images corresponding to the plurality of echo times based on the MR data. The processing circuitry generates a relaxation time map of tissue based on the plurality of magnitude images. The processing circuitry generates a susceptibility map quantitatively indicating susceptibility values in a subject based on a magnetic field distribution that is generated based on a plurality of phase images corresponding to the plurality of echo times and the relaxation time map.