Virtual Coil Parallel Imaging MRI Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autocalibrating parallel imaging techniques in MRI are computationally expensive due to their exponential scaling with the number of receiver coils, making them inefficient for large coil arrays.
Innovation Solution
A method and system that generate a complete MR data set for a virtual coil using MR data from at least two RF source coils and synthesis weights, reducing the computational burden by eliminating the need for separate coil reconstructions and combining data via linear combination weights.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If per-coil autocalibrating PI reconstructions are performed for each receiver coil, then phase-cancellation artifacts are eliminated and image quality is improved, but computational expense increases exponentially with the number of coils
Solution Approach 1:
The patent combines multiple coil datasets into a single merged dataset before performing the autocalibrating PI reconstruction. Instead of reconstructing each coil separately and then combining images, the method merges the k-space data from multiple coils using calculated weights, then performs a single reconstruction on the merged dataset. This eliminates the exponential computational scaling while maintaining the ability to produce high-quality images without phase-cancellation artifacts.
Solution Approach 2:
The patent introduces a virtual coil as an intermediary construct that represents the combined sensitivity profile of multiple physical coils. The virtual coil serves as a mediator that allows the reconstruction algorithm to process merged data from multiple coils as if it were from a single coil, thereby simplifying the computational process while preserving the spatial encoding information needed to eliminate phase-cancellation artifacts.
2Reliability
If multiple separate coil images are reconstructed and combined via sum-of-squares, then phase cancellation problems are eliminated, but computational time increases significantly
Solution Approach 1:
The patent performs the data merging operation preliminarily, before the reconstruction step. By calculating the merged dataset from multiple coil inputs before reconstruction, and by pre-calculating the weights during a training phase, the method eliminates the need for separate reconstructions of each coil. This preliminary merging action reduces the overall reconstruction time while maintaining image accuracy.
3Measurement precision
If the number of receiver coils is increased to improve spatial coverage and signal-to-noise ratio, then imaging performance is enhanced, but computational complexity scales exponentially
Solution Approach 1:
The patent merges datasets from multiple coils into a single consolidated dataset using calculated weights, thereby linearizing the computational scaling with respect to the number of coils. This merging approach allows the system to utilize multiple coils for improved signal-to-noise ratio and spatial coverage without suffering from exponential computational complexity in the reconstruction process.
Data Source
AI summary
A method for generating a magnetic resonance (MR) image includes acquiring calibration data from each of a plurality of RF source coils. Calibration data for a virtual coil is generated based on the calibration data from the plurality of RF source coils and a set of synthesis weights is generated based on the calibration data from the plurality of RF source coils and the calibration data for the virtual coil. Accelerated MR data is acquired from each of the plurality of RF source coils. An image can be reconstructed based on an application of the set of synthesis weights to the accelerated MR data from the plurality of RF source coils.