Trained Reconstruction Model for Unaliased MRI Slice Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Simultaneous multi-slice (SMS) imaging in MRI faces challenges in reconstructing unaliased images due to the simultaneous acquisition of MR signals from multiple slices, leading to aliased images when directly reconstructed using inverse Fourier transform.
Innovation Solution
A system and method utilizing a trained reconstruction model, including a machine learning model, to process target k-space data, which involves applying multiband excitation RF pulses and phase modulation to generate unaliased images by determining reference data sets and inputting aliased images into the model for outputting unaliased images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple slices are simultaneously excited and MR signals are acquired from multiple slices, then imaging speed and productivity are improved, but image quality deteriorates due to aliasing when directly reconstructed using inverse Fourier transform
Solution Approach 1:
The patent introduces a trained reconstruction model as an intermediary between the acquired k-space data and the final image reconstruction. This model processes the aliased images and reference data to generate unaliased slice images, effectively mediating the transition from simultaneous multi-slice acquisition to individual slice visualization without direct inverse Fourier transform
Solution Approach 2:
The patent performs preliminary actions by acquiring reference data from a reference region before reconstructing the target slice images. This reference data is processed in advance to create a reference image that serves as a basis for the subsequent reconstruction process, enabling the separation of simultaneously acquired slice signals
2Manufacturing precision
If a trained reconstruction model is used to process k-space data and generate unaliased images, then image quality and reconstruction speed are improved, but device complexity increases due to the introduction of machine learning models and additional processing steps
Solution Approach 1:
The trained reconstruction model operates autonomously to process the k-space data and generate unaliased images without requiring complex real-time control systems. The model self-adjusts based on the input data patterns learned during training, reducing the need for complex hardware interventions and manual calibration procedures
Solution Approach 2:
The patent merges multiple functions into the trained reconstruction model, including data processing, image reconstruction, and aliasing removal capabilities. This consolidation reduces overall system complexity by replacing multiple separate processing stages with a single integrated model that handles all reconstruction tasks
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
The approach effectively generates unaliased images for each slice, improving image quality and reconstruction speed by using a non-linear machine learning model to process phase-modulated k-space data, reducing the need for additional reference data and enhancing SMS imaging efficiency.
Implementation Method 1
magnetic resonance imaging (MRI)... simultaneously excite, for one or more times, the plurality of target slices of the ROI... acquire the target k-space data from the plurality of excited target slices
Implementation Method 2
apply phase modulation to at least one of the excited target slices by applying, to each of the at least one of the excited target slices, a transmit phase that varies with a plurality of target acquisition periods and/or a phase encoding direction
Data Source
AI summary
A method for SMS imaging may include obtaining target k-space data related to a region of interest (ROI) of an object. The method may also include generate, based on the target k-space data using a trained reconstruction model, a plurality of target images each of which corresponds to one of a plurality of target slices of the ROI at one of a plurality of target acquisition periods.