Transformer Neural Imaging for Cardiac Motion Artifact Reduction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing imaging modalities like CT and MRI struggle with motion artifacts due to patient movement during scanning, particularly in cardiovascular imaging, leading to degraded image quality and limiting the use of high-speed scanners due to high costs.

Innovation Solution

A deep neural network comprising a transformer neural network and a synthesis neural network, trained in a Wasserstein generative adversarial network framework, is used to mitigate temporal and spatial aspects of motion artifacts in MRI and CT images, enhancing image quality and resolution.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If conventional low-speed CT scanner is used, then cost is reduced, but image quality and resolution are degraded due to heart motion artifacts

Engineering Contradiction:

ImprovecostVSAvoidimage quality

Core Design Contradiction:

Ease of manufactureVSManufacturing precision

Solution Approach 1:

A deep learning system acts as an intermediary between low-speed CT scanner and high-quality cardiac imaging. The system includes a motion artifact reduction module that processes images from conventional scanners to remove motion artifacts, and a cardiac image enhancement module that reconstructs high-resolution cardiac images from the cleaned data, enabling low-cost scanners to produce high-quality cardiac images

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the mechanical solution (high-speed CT scanner hardware) with a computational solution (deep learning-based image processing). Instead of using expensive high-speed scanners to physically capture motion-free images, the system uses artificial intelligence to computationally remove motion artifacts and reconstruct high-quality cardiac images from conventional scanner data

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Manufacturing precision

If high-speed CT or ECG-gated scanner is used, then image quality is improved, but cost increases significantly

Engineering Contradiction:

Improveimage qualityVSAvoidcost

Core Design Contradiction:

Manufacturing precisionVSEase of manufacture

Solution Approach 1:

The deep learning system creates a computational copy of the high-speed scanner's imaging capability. By training neural networks on pairs of images from high-speed scanners and corresponding conventional scanner images, the system learns to replicate the high-quality output of expensive equipment using affordable hardware and software processing

Inventive Principle:

Principle #26Copying

3Manufacturing precision

If scan time is extended to reduce motion artifacts, then image quality improves, but patient motion during scan increases artifacts

Engineering Contradiction:

Improveimage qualityVSAvoidscan time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The system converts the harmful effect of patient motion during scanning into a beneficial outcome. The deep learning motion artifact reduction module is specifically designed to detect and remove artifacts caused by patient motion, breathing, and cardiac movement, transforming what was previously a degradation into an opportunity to produce high-quality images from rapid scans

Inventive Principle:

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

Data Source

PatentUS12602754B2Dynamic imaging and motion artifact reduction through deep learning

Publication Date: 2026.04.14 RENESSELAER POLYTECHNIC INST

AI summary

A deep neural network for motion artifact reduction includes a transformer neural network, and a synthesis neural network. The transformer neural network is configured to receive input image data comprising a target image, a prior image, and a subsequent image. The transformer neural network is configured to mitigate a temporal aspect of the motion artifact, and to provide a transformer output related to the input data. The synthesis neural network is configured to receive the transformer output, and to provide at least one output image having a reduced motion artifact relative to the input image data.