MRI Reconstruction Kernel Using Temporal Correlations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing MRI reconstruction methods require extensive reference data for each frame, leading to increased measurement time and potential artifacts in dynamic imaging.
Innovation Solution
A method that uses information from neighboring frames to calibrate a reconstruction kernel, reducing measurement time by using preliminary reconstruction frames as reference data for calculating a larger n+x-dimensional reconstruction kernel.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If reference data are used for each frame to calculate reconstruction kernel, then reconstruction quality is improved, but measurement time is increased
Solution Approach 1:
The patent applies preliminary action by calculating a preliminary reconstruction kernel from reference data before the actual dynamic imaging acquisition. This preliminary kernel enables initial reconstruction of undersampled frames, allowing the system to proceed with reduced measurement time while maintaining acceptable reconstruction quality. The preliminary action prepares the reconstruction framework in advance, avoiding the need to collect extensive reference data for each dynamic frame.
Solution Approach 2:
The patent extends the reconstruction kernel from traditional 2D or 3D space to n+x dimensions by incorporating temporal information from multiple time frames. This dimensional extension allows the kernel to leverage correlations across time, improving reconstruction quality without requiring proportionally more reference data at each individual frame, thus resolving the time-quality tradeoff.
2Measurement precision
If extensive reference data are collected for each frame, then reconstruction accuracy is improved, but data processing complexity is increased
Solution Approach 1:
The preliminary reconstruction kernel is calculated once from reference data before dynamic imaging, avoiding the need to process extensive reference data for each frame during the actual acquisition. This preliminary calculation simplifies the real-time processing complexity while maintaining reconstruction accuracy through the use of the pre-computed kernel.
Solution Approach 2:
By extending the reconstruction kernel to n+x dimensions that incorporate temporal correlations across multiple frames, the patent improves reconstruction accuracy by utilizing additional temporal information without proportionally increasing processing complexity. The dimensional extension allows efficient use of existing data across time rather than requiring separate extensive reference data for each frame.
3Productivity
If undersampled data are used to reduce measurement time, then acquisition speed is improved, but reconstruction quality deteriorates
Solution Approach 1:
The preliminary reconstruction kernel is calculated from reference data before the undersampled dynamic imaging acquisition. This preliminary preparation enables the system to reconstruct undersampled frames effectively without requiring extensive reference data during the fast acquisition, thus maintaining reconstruction quality while achieving high acquisition speed through undersampling.
Solution Approach 2:
The n+x dimensional reconstruction kernel incorporates temporal information from multiple time frames, allowing the system to reconstruct undersampled data with improved quality by leveraging correlations across time. This dimensional extension recovers information that would otherwise be lost due to undersampling, maintaining reconstruction quality while enabling faster acquisition.
4Loss of time
If traditional reconstruction methods are used with undersampled data, then measurement time is reduced, but artifacts in dynamic imaging are increased
Solution Approach 1:
The preliminary reconstruction kernel is calculated from reference data to establish a robust reconstruction framework before dynamic imaging. This preliminary preparation reduces artifacts in the final dynamic images by providing a well-conditioned kernel that accounts for the sampling pattern, enabling fast undersampled acquisition without the severe artifacts that would otherwise result.
Solution Approach 2:
The n+x dimensional reconstruction kernel incorporates temporal correlations across multiple frames, which helps suppress artifacts in dynamic imaging by leveraging information from neighboring time points. This temporal dimension provides additional constraints that reduce reconstruction artifacts while maintaining fast measurement times through undersampling.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach allows for the reconstruction of undersampled MRI data with reduced measurement time and fewer artifacts, improving the efficiency and quality of dynamic MRI imaging.
Implementation Method 1
An MR image is generated in image space by Fourier transformation
Implementation Method 2
nuclear spins oriented in the direction of a main magnetic field (z-direction) are stimulated by the radiation of electromagnetic RF pulses
Implementation Method 3
nuclear spins oriented in the direction of a main magnetic field (z-direction)
Implementation Method 4
By time-varying superpositions of additional location-dependent magnetic fields for all three spatial directions, a spatial encoding is generated which can be described as the traversal of a trajectory in the k-space
Data Source
AI summary
The invention relates to a method for generating a series of magnetic resonance images from an MR bin series obtained by means of an MRI measurement, wherein the MR bin series comprises a plurality of bin frames (f1, f2, f3) with determined MR data, wherein at least one bin frame is undersampled, i.e., has missing MR data in addition to the determined MR data, wherein the bin frames differ by the value of x prespecified parameters (x=1 or greater) under which the MR data were determined, comprising: a) calculating an n-dimensional preliminary reconstruction kernel from preliminary reference data of a reference frame; b) determining preliminary reconstruction frames by reconstruction of the MR data missing in the MR bin series by means of the preliminary reconstruction kernel; c) calculating an n+x-dimensional reconstruction kernel by calculation from the MR data of the preliminary reconstruction frames, wherein MR data of different preliminary reconstruction frames are used for the calculation of the n+x-dimensional reconstruction kernel; d) determining further reconstruction frames by reconstructing the data missing in the bin frames by means of the n+x-dimensional reconstruction kernel; e) generating n-dimensional magnetic resonance images from the reconstruction frames determined in step (d).