MRI K-Space Neural Reconstruction for Faster High-Quality Imaging

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Magnetic resonance imaging (MRI) is hindered by lengthy imaging times, which can be uncomfortable for patients and limit its application, particularly for those with claustrophobia, and there is a need for improved image quality.

Innovation Solution

A magnetic resonance image processing method using artificial neural networks to process k-space data, involving pre-processing with a linear function and post-processing with inverse Fourier operations to reduce artifacts and enhance image quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of time

If sub-sampling is used to reduce imaging time, then imaging time is reduced, but image quality deteriorates due to multi-domain artifacts

Engineering Contradiction:

Improveimaging timeVSAvoidimage quality

Core Design Contradiction:

Loss of timeVSManufacturing precision

Solution Approach 1:

An artificial neural network model is introduced as an intermediary between sub-sampled k-space data and the final image reconstruction. The neural network processes the incomplete k-space data and predicts missing information, enabling high-quality image reconstruction from sub-sampled data without requiring full sampling, thus resolving the contradiction between reduced imaging time and maintained image quality

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent changes the sampling parameters by using sub-sampling strategies (e.g., variable density sampling, random sampling) instead of uniform full sampling. By combining these parameter changes with neural network-based reconstruction, the system achieves faster imaging while maintaining image quality through intelligent data completion rather than traditional interpolation methods

Inventive Principle:

Principle #35Parameter changes

2Manufacturing precision

If conventional reconstruction methods are used to maintain image quality, then image quality is maintained, but imaging time increases

Engineering Contradiction:

Improveimage qualityVSAvoidimaging time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent replaces conventional mechanical/mathematical reconstruction methods (such as iterative SENSE or GRAPPA algorithms) with an artificial neural network-based reconstruction system. This substitution enables parallel processing of k-space data and artifact suppression through learned patterns, significantly reducing reconstruction time while maintaining or improving image quality compared to traditional methods

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Manufacturing precision

If full sampling is performed to ensure high image quality, then image quality is high, but imaging time becomes excessively long

Engineering Contradiction:

Improveimage qualityVSAvoidimaging efficiency

Core Design Contradiction:

Manufacturing precisionVSProductivity

Solution Approach 1:

The patent applies partial sampling strategies where only a portion of k-space data is acquired using sub-sampling patterns. The neural network then completes the missing data, allowing the system to achieve high image quality with less than full sampling, thereby improving imaging efficiency while maintaining diagnostic image quality

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The neural network model is pre-trained on large datasets of fully sampled images and corresponding sub-sampled images, learning the mapping relationships and artifact patterns in advance. During actual imaging, this preliminary learning enables rapid reconstruction from sub-sampled data without requiring time-consuming iterative optimization, thus improving productivity while maintaining image quality

Inventive Principle:

Principle #10Preliminary action

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

This approach allows for high-quality MRI image reconstruction in reduced time, facilitating accurate diagnosis of lesion regions.

Implementation Method 1

acquiring second k-space data from the first k-space data by using a first artificial neural network model

Methodology Applied

Scientific EffectArtificial neural network processing:

Implementation Method 2

acquiring a first magnetic resonance image from the second k-space data by using an inverse Fourier operation

Methodology Applied

Scientific EffectFourier transformation:

Data Source

PatentUS12488518B2Magnetic resonance image processing apparatus and method using artificial neural network in k-space domain

Publication Date: 2025.12.02 AIRS MEDICAL INC

AI summary

According to an embodiment of the present invention, there is provided a magnetic resonance image processing method that is performed by a magnetic resonance image processing apparatus, the magnetic resonance image processing method including: acquiring first k-space data calculated based on a sub-sampled magnetic resonance signal; acquiring second k-space data from the first k-space data by using a first artificial neural network model; and acquiring a first magnetic resonance image from the second k-space data by using an inverse Fourier operation.