3D MR Image Reconstruction Using a Pre-Learned Spatial Subspace
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Generating on-the-fly MR images with sufficient temporal resolution for MR-based therapies and treatments is challenging due to the complexity and slowness of existing systems.
Innovation Solution
A method involving a first pulse sequence to obtain initial k-space data, constructing spatial and temporal factors, determining a transformation, and applying a second pulse sequence to generate real-time images using a pre-learned spatial subspace.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional MR imaging methods are used to generate on-the-fly images, then image quality and diagnostic accuracy are maintained, but temporal resolution is insufficient and image generation speed is slow
Solution Approach 1:
The patent applies preliminary action by pre-learning and storing spatial basis functions and temporal response patterns from training data before actual imaging. During real-time imaging, these pre-computed templates are used to rapidly reconstruct images from k-space data without performing full image reconstruction algorithms at each time point, thereby achieving high temporal resolution while maintaining image quality.
Solution Approach 2:
The patent uses copying by creating a library of pre-computed image templates and temporal response patterns from training data. These copies are then used during actual imaging to quickly generate real-time images through template matching and linear combination, avoiding the need for time-consuming reconstruction algorithms during dynamic imaging procedures.
2Loss of time
If high temporal resolution is achieved through rapid image reconstruction, then therapy monitoring capability is improved, but image quality and contrast consistency may deteriorate
Solution Approach 1:
The patent implements feedback by continuously monitoring and adjusting the reconstruction process using real-time k-space data while comparing against pre-learned temporal patterns. The system feedback-loop ensures that image quality and contrast consistency are maintained by validating reconstructed images against expected temporal responses from training data, allowing rapid reconstruction while preserving diagnostic accuracy.
Solution Approach 2:
The patent applies parameter changes by dynamically adjusting reconstruction parameters such as temporal basis function weights and spatial basis function selections based on real-time imaging conditions. This allows the system to optimize image quality and contrast consistency at each time point while maintaining high temporal resolution through adaptive parameter modification rather than fixed reconstruction protocols.
3Loss of time
If real-time image construction is performed using traditional reconstruction algorithms, then temporal resolution is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent extracts the computationally intensive parts of image reconstruction by pre-computing and storing spatial basis functions and temporal response patterns during a training phase. During actual imaging, only simple linear combinations and matrix multiplications are needed rather than full iterative reconstruction algorithms, dramatically reducing real-time computational complexity while achieving high temporal resolution.
Solution Approach 2:
The patent performs preliminary computational actions by pre-learning and caching spatial and temporal factors from training data before actual imaging procedures. This preliminary processing shifts the heavy computational burden to an offline training phase rather than online imaging phases, reducing real-time computational complexity and enabling rapid image reconstruction during dynamic therapies.
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
Enables rapid construction of real-time MR images with high temporal resolution, allowing for accurate tracking of therapies and treatments by maintaining consistent motion and contrast during MR-guided interventions.
Implementation Method 1
a magnet operable to provide a magnetic field
Implementation Method 2
a transmitter operable to transmit to a region within the magnetic field
Implementation Method 3
a receiver operable to receive a magnetic resonance signal from the region with the magnetic field
Data Source
AI summary
A method for performing real-time magnetic resonance (MR) imaging on a subject is disclosed. A prep pulse sequence is applied to the subject to obtain a high-quality special subspace, and a direct linear mapping from k-space training data to subspace coordinates. A live pulse sequence is then applied to the subject. During the live pulse sequence, real-time images are constructed using a fast matrix multiplication procedure on a single instance of the k-space training readout (e.g., a single k-space line or trajectory), which can be acquired at a high temporal rate.