Parallel Imaging Reconstruction Using Correlation Values
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current parallel imaging methods for magnetic resonance (MR) image reconstruction face significant computation time challenges due to large matrices required for calculating linear combination coefficient weights, especially in 3D image reconstructions, which hinder efficient image processing and increase computation time.
Innovation Solution
The method generates linear combination coefficient weights by solving systems of linear equations formulated with correlation values instead of calibration data, reducing the size of the matrices and computation requirements, and uses correlation values to synthesize unacquired data for efficient image reconstruction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If calibration data is used to calculate linear combination weights in parallel imaging, then accurate image reconstruction is achieved, but computation time increases significantly due to large matrix sizes
Solution Approach 1:
The patent extracts only the essential correlation information from the calibration data by computing correlation values between different coil signals. Instead of using the entire calibration data matrix, only the correlation values are retained and used to form a reduced system of linear equations, thereby eliminating redundant data while preserving the essential information needed for accurate weight calculation.
Solution Approach 2:
The patent segments the large calibration data matrix into smaller correlation value computations. By breaking down the problem into pairwise correlation calculations between coils and k-space locations, the method transforms one large matrix operation into multiple smaller, more manageable computations that can be performed more efficiently.
2Measurement precision
If fully sampled calibration data is acquired for accurate weight calculation, then reconstruction accuracy improves, but data acquisition time and storage requirements increase
Solution Approach 1:
The patent applies partial action by acquiring only a reduced set of calibration data that is sufficient for computing correlation values, rather than acquiring complete fully sampled calibration data. The correlation values can be computed from undersampled calibration data, which reduces the acquisition time and storage requirements while still providing accurate weight calculations for parallel imaging reconstruction.
3Loss of information
If large matrices are used for parallel imaging reconstruction, then complete information is processed, but computational complexity and memory requirements increase exponentially
Solution Approach 1:
The patent extracts the essential relationships between coils and k-space locations by computing correlation values, which capture the spatial encoding information needed for reconstruction. This extraction process removes the redundant dimensional information from the full calibration matrix, resulting in a compressed representation that maintains all necessary information for accurate reconstruction with reduced computational complexity.
Solution Approach 2:
The patent changes the parameters of the reconstruction problem by transforming the large calibration matrix into a set of correlation values. This parameter transformation converts a high-dimensional matrix problem into a lower-dimensional problem involving correlation coefficients, thereby reducing computational complexity and memory requirements while preserving the essential reconstruction information.
Data Source
AI summary
A system and method for parallel imaging is disclosed that generates linear combination coefficient weights by solving systems of linear equations formulated with correlation values. An MRI apparatus includes a computer programmed to acquire MR data from an imaging volume for a plurality of encoding locations using an array of RF receiver coils. Correlation values are calculated from the MR data. From these calculated correlation values, synthesis weights are generated. An image is then reconstructed based on an application of the synthesis weights to at least a portion of the MR data acquired from the array of RF receiver coils.