ANN-Based MRI Image Reconstruction Using Precomputed Signal Trajectories
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image reconstruction methods in medical imaging, particularly in MRI, are time-consuming and computationally intensive due to the need for solving complex signal evolution equations, and existing ANN-based methods lack iterative corrections for improving image quality.
Innovation Solution
A method and system utilizing an Artificial Neural Network (ANN) for accelerated forward transformation to generate predicted signal data, which is then used to adapt and refine image data iteratively, speeding up the reconstruction process while maintaining high quality, by calculating cost values and adjusting data in both signal and image domains.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If model-based iterative image reconstruction methods are used to account for non-ideal physical effects, then image quality and accuracy are improved, but reconstruction time and computational complexity increase significantly
Solution Approach 1:
The patent pre-calculates and stores signal evolution trajectories for various tissue parameter combinations in a dictionary before actual image reconstruction. This preliminary action allows the iterative reconstruction to quickly match measured signals against pre-computed trajectories, avoiding repeated complex simulations during reconstruction while maintaining accuracy in accounting for non-ideal physical effects
Solution Approach 2:
The patent creates a dictionary of synthetic signal trajectories that copy the complex physics-based signal evolution patterns. Instead of repeatedly solving the full Bloch equations during iterative reconstruction, the method uses these pre-computed copies to efficiently compare and match against actual measured signals, significantly reducing computational time while preserving image quality
2Manufacturing precision
If complex signal evolution equations are solved iteratively to account for non-linear and non-ideal parameters, then manufacturing precision of image data is improved, but device complexity and computational requirements worsen
Solution Approach 1:
The patent segments the complex signal evolution problem into discrete tissue parameter combinations (T1, T2, proton density values) and pre-computes trajectories for each segment. This segmentation allows the system to handle complexity by breaking it into manageable pre-computed pieces stored in a dictionary, rather than solving the full complex equations iteratively during reconstruction
Solution Approach 2:
The system performs preliminary computation of signal trajectories for all possible tissue parameter combinations before actual image reconstruction. This pre-computation stores the complex physics-based signal patterns in advance, allowing the reconstruction phase to use simple matching operations instead of complex iterative equation solving, thereby reducing computational complexity while maintaining precision
Data Source
AI summary
The present disclosure is related to methods and systems for image reconstruction including accelerated forward transformation with an Artificial Neural Network (ANN).