Neural Network EPI Artifact Correction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Echo planar imaging (EPI) is prone to various artifacts such as B0 susceptibility, chemical shift, Rician noise, Nyquist, and Gibbs ringing artifacts due to its high-speed image acquisition, which hinders the diagnostic quality of MRI neuroimaging applications.

Innovation Solution

A method using a trained neural network to correct EPI artifacts by generating a synthetic dataset of artifacted images, which includes simulated artifacts, and employing deep learning models like GANs and UNet architectures to remove these artifacts from images.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If echo planar imaging is used for high-speed image acquisition, then imaging speed is improved, but image quality deteriorates due to artifacts

Engineering Contradiction:

Improveimaging speedVSAvoidimage quality

Core Design Contradiction:

SpeedVSManufacturing precision

Solution Approach 1:

The patent applies deep learning models to convert the harmful effect of EPI artifacts into a beneficial correction process. The neural networks are trained specifically on artifacted EPI images to learn the patterns of distortion and automatically correct them, transforming what was previously a limiting factor into an opportunity for enhanced image quality through intelligent post-processing.

Inventive Principle:

Principle #22Blessing in disguise (Convert harm into benefit)

Solution Approach 2:

The patent introduces deep learning models as an intermediary between the artifacted EPI images and the final corrected images. These models act as a mediator that processes the distorted images, separates the actual anatomical information from the artifacts, and produces corrected output, thereby resolving the contradiction between speed and quality.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Manufacturing precision

If deep learning models are used to correct artifacts, then image quality is improved, but computational complexity increases

Engineering Contradiction:

Improveimage qualityVSAvoidcomputational complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by pre-training the deep learning models on large datasets of artifacted and corrected images before actual use. This offline training phase prepares the models in advance, so that during actual EPI correction, the computational burden is reduced to inference only, making the process more efficient and manageable despite the inherent complexity of deep learning.

Inventive Principle:

Principle #10Preliminary action

3Reliability

If artifact correction is applied to improve diagnostic quality, then reliability is improved, but processing time increases

Engineering Contradiction:

Improvediagnostic qualityVSAvoidprocessing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces traditional mechanical or algorithmic artifact correction methods with deep learning-based intelligent systems. Unlike conventional correction techniques that rely on hand-crafted algorithms and iterative optimization, the trained neural networks perform correction in a single pass, dramatically reducing processing time while maintaining or improving diagnostic quality.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS20250005715A1Methods, apparatuses, systems and computer-readable mediums for correcting echo planar imaging artifacts

Publication Date: 2025.01.02 SIEMENS HEALTHINEERS AG

AI summary

A method for correcting echo planar imaging artifacts includes correcting at least one echo planar imaging artifact in an image to obtain a correct image. A trained neural network is used to correct the at least one echo planar imaging artifact. The image is obtained through echo planar imaging.