Multi-Slice MRI Reconstruction via Hankel Tensor Completion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autocalibrating parallel imaging methods for multi-slice MRI face challenges in prolonging acquisition windows, leading to blurring and artifacts due to off-resonance effects, and multi-contrast MRI requires time-consuming independent scans prone to motion artifacts.
Innovation Solution
The proposed method employs a block-wise Hankel tensor completion framework for simultaneous reconstruction of multiple adjacent slices, leveraging similarities in coil sensitivity and image content across slices to reduce artifacts and accelerate data acquisition, using high-order SVD and rank truncation for low-rank tensor completion, and complementary sampling patterns to suppress aliasing artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional autocalibrating parallel imaging methods are used to acquire sufficient ACS data in multi-slice spiral MRI, then coil sensitivity estimation is improved, but the acquisition window is prolonged leading to blurring and artifacts due to off-resonance effects
Solution Approach 1:
The patent combines central k-space lines from multiple adjacent slices to form a composite ACS dataset. By merging data from multiple slices with interleaved sampling patterns, sufficient calibration information is obtained without extending the acquisition window of any single slice, thereby avoiding off-resonance blurring while achieving accurate coil sensitivity estimation.
Solution Approach 2:
The patent uses the similarity of coil sensitivity maps across adjacent slices to create copies of calibration information. The ACS data from one slice can be used to estimate coil sensitivities for adjacent slices, eliminating the need to acquire separate ACS data for each slice and reducing total acquisition time.
2Loss of information
If multiple independent scans are performed for multi-contrast MRI at the same slice location with various pulse sequences, then differential diagnostic information is obtained, but the acquisition time is prolonged and susceptibility to motion artifacts increases
Solution Approach 1:
The patent merges multiple contrast datasets acquired from the same slice location into a unified reconstruction framework. By combining T1-weighted, T2-weighted, and other contrast data in a joint low-rank tensor completion approach, the system recovers all contrast information simultaneously from a single scan session, eliminating the need for multiple separate scans and reducing motion susceptibility.
Solution Approach 2:
The patent extends the reconstruction problem from 2D spatial domain to 4D tensor space by incorporating multiple contrasts as an additional dimension. This dimensional extension allows the system to exploit correlations across different contrast types and recover all contrasts jointly from undersampled data, achieving multi-contrast imaging in a single acquisition.
3Productivity
If 1D random sampling patterns are used to accelerate multi-slice Cartesian MRI acquisition, then productivity is improved, but aliasing artifacts occur in single direction
Solution Approach 1:
The patent introduces asymmetry in the sampling pattern by alternating the phase-encoding direction among adjacent slices. While one slice is sampled with vertical phase encoding, the next slice uses horizontal phase encoding, creating complementary aliasing patterns that can be effectively separated and suppressed through low-rank tensor completion.
Solution Approach 2:
The patent transforms the sampling problem from 1D undersampling to 2D complementary sampling by utilizing the slice dimension. By alternating phase-encoding directions across slices, the system creates incoherent aliasing artifacts in orthogonal directions that can be suppressed through the low-rank tensor completion framework, effectively converting a 1D artifact problem into a solvable 2D problem.
Data Source
AI summary
Image reconstruction methods for multi-slice and multi-contrast magnetic resonance imaging with complementary sampling schemes are provided, comprising: data acquisition using complementary sampling schemes between slices or/and contrasts) in spiral imaging or Cartesian acquisition; joint calibrationless reconstruction of multi-slice and multi-contrast data via block-wise Hankel tensor completion.