K-space trajectory infidelity correction in MRI

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

K-space trajectory infidelity in MR imaging leads to artifacts in reconstructed images, with conventional methods facing trade-offs between model complexity and computation cost, and deep learning approaches risking alteration of the patient's structure representation.

Innovation Solution

A machine-learned model, such as a deep learned autoencoder or encoder-decoder network, is trained to correct k-space measurements for trajectory infidelity by estimating trajectory shifts and correcting k-space data, using losses for trajectory shift, k-space correction, and corrupted data estimation, allowing for direct correction in the k-space domain.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional model-based methods are used to correct k-space trajectory, then trajectory infidelity can be addressed, but model complexity and computation cost increase

Engineering Contradiction:

Improvetrajectory correction accuracyVSAvoidmodel complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent replaces conventional model-based correction methods with a deep learning-based approach. The neural network model learns to correct trajectory infidelity directly from data, substituting complex mathematical models with a data-driven system that captures non-linear relationships between trajectory deviations and image artifacts.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent changes the approach from explicit mathematical modeling to implicit learning through training data. By training the neural network on examples of trajectory infidelity and corresponding corrections, the system adapts its internal parameters to capture the correction mapping without requiring explicit mathematical formulation.

Inventive Principle:

Principle #35Parameter changes

2Reliability

If deep learning is used to remove artifacts in image space, then artifact removal capability improves, but risk of altering patient structure representation increases

Engineering Contradiction:

Improveartifact removal capabilityVSAvoidpatient structure representation accuracy

Core Design Contradiction:

ReliabilityVSManufacturing precision

Solution Approach 1:

The patent applies correction in the k-space domain before image reconstruction, rather than correcting artifacts after image formation. This preliminary correction of trajectory infidelity in the frequency domain prevents artifact generation at the source, avoiding the risk of altering patient structure representation that occurs when correcting artifacts in the already-reconstructed image space.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent uses k-space as an intermediary domain between the raw trajectory data and the final image reconstruction. By correcting trajectory infidelity in this intermediate k-space representation before reconstruction, the system separates the correction function from the reconstruction function, preventing direct manipulation of patient structure information.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11354833B2K-space trajectory infidelity correction in magnetic resonance imaging

Publication Date: 2022.06.07 SIEMENS HEALTHINEERS AG

AI summary

For k-space trajectory infidelity correction, a model is machine trained to correct k-space measurements in k-space. K-space trajectory infidelity correction uses deep learning. Trajectory infidelity is corrected from a k-space point of view. Since the image artifacts arise from k-space acquisition distortion, a machine learning model is trained to correct in k-space, either changing values of k-space measurements or estimating the trajectory shifts in k-space.