MRI Neural Network Reconstruction for Fast k-Space Synthesis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Magnetic Resonance Imaging (MRI) systems face challenges with slow imaging speed, which limits temporal and spatial resolution due to the time-consuming nature of MRI scanning.
Innovation Solution
A system and method for MRI that acquires first k-space data, synthesizes second k-space data for unfilled locations using various techniques such as partial Fourier imaging, parallel imaging, or regridding, and applies a neural network-based reconstruction algorithm to reconstruct images efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional MRI scanning is used, then comprehensive k-space data is acquired, but scanning time is excessive and temporal resolution is limited
Solution Approach 1:
The patent applies partial sampling in k-space domain by acquiring only a subset of k-space locations (first k-space data) instead of complete sampling. This partial action reduces scanning time while the neural network synthesizes the missing second k-space data to maintain image quality, resolving the contradiction between comprehensive data acquisition and time efficiency
Solution Approach 2:
The system performs preliminary acquisition of first k-space data at selected locations before complete scanning would be required. By pre-acquiring this subset and using neural network-based synthesis to generate the remaining second k-space data, the system prepares sufficient information for high-quality reconstruction without the time cost of full sampling
2Productivity
If scanning time is reduced, then temporal resolution improves, but k-space data becomes insufficient for high-quality reconstruction
Solution Approach 1:
The neural network acts as an intermediary that bridges the gap between incomplete first k-space data and the complete information needed for high-quality reconstruction. It synthesizes the missing second k-space data by learning from the acquired data and applying learned patterns, enabling fast imaging without information loss
Solution Approach 2:
The system creates a copy or synthesis of the missing second k-space data based on the acquired first k-space data. Instead of physically acquiring all k-space locations, the neural network generates synthetic copies of the missing information, maintaining data completeness while reducing scanning time
3Productivity
If neural network-based reconstruction is applied, then reconstruction efficiency improves, but computational complexity increases
Solution Approach 1:
The neural network is pre-trained on large datasets of k-space and image pairs before deployment. This preliminary training phase captures the complex relationships between k-space data and image features, allowing the network to perform efficient reconstruction during actual imaging without requiring complex computational algorithms at runtime
Data Source
AI summary
A method for magnetic resonance imaging may include acquiring first k-space data that is generated by entering acquired magnetic resonance (MR) data into a plurality of first k-space locations. The method may further include synthesizing second k-space data for a plurality of second k-space locations that are not filled with the acquired MR data. The method may further include reconstructing an image from the first k-space data and the second k-space data by applying a reconstruction algorithm. The reconstruction algorithm is based at least in part on a neural network technique.