3D Image Reconstruction Artifact Reduction via Machine Learning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In medical imaging, particularly computed tomography, reconstructing three-dimensional image data sets from two-dimensional images captured over a restricted angular range often results in severe artifacts, which are difficult to remove even with advanced algorithms, leading to potential misclassification of features as artifacts.

Innovation Solution

A method that generates an artifact-reduced image data set using a machine learning algorithm and incorporates it into the reconstruction process, allowing for the reduction of artifacts while preserving actual image features by using the artifact-reduced data set as additional information alongside the original two-dimensional images, either through iterative reconstruction or filtered back projection.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If images are captured over a restricted angular range, then the imaging process becomes feasible in certain applications, but severe artifacts result in the reconstructed image data set

Engineering Contradiction:

Improveapplicability of imaging methodVSAvoidimage reconstruction quality

Core Design Contradiction:

Adaptability or versatilityVSManufacturing precision

Solution Approach 1:

The patent introduces an intermediary artifact reduction step between image capture and final reconstruction. A machine learning-based artifact reduction algorithm processes the initially reconstructed image to suppress artifacts caused by restricted angular range, thereby mediating between the practical constraint of limited imaging angles and the quality requirement of artifact-free reconstruction

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent applies preliminary artifact reduction processing before final image reconstruction. By pre-processing the reconstructed image to remove artifacts caused by restricted angular sampling, the method prepares improved input data for subsequent reconstruction algorithms, thereby improving final image quality without requiring expanded angular range

Inventive Principle:

Principle #10Preliminary action

2Manufacturing precision

If machine learning methods are used to remove artifacts, then artifact suppression improves, but genuine features may be misclassified and removed

Engineering Contradiction:

Improveartifact suppression qualityVSAvoidfeature preservation accuracy

Core Design Contradiction:

Manufacturing precisionVSReliability

Solution Approach 1:

The patent implements a feedback mechanism where the artifact reduction algorithm's output is evaluated and used to refine the processing. By incorporating feedback loops that monitor for potential feature misclassification and adjust processing parameters accordingly, the system balances artifact suppression with preservation of genuine anatomical features

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent dynamically adjusts processing parameters of the machine learning algorithm based on image characteristics. By changing parameters such as processing intensity, threshold values, and regularization terms adaptively, the system optimizes the balance between artifact removal and feature preservation for different imaging scenarios

Inventive Principle:

Principle #35Parameter changes

3Manufacturing precision

If iterative reconstruction methods are used, then reconstruction quality improves, but processing time and computational resources increase

Engineering Contradiction:

Improvereconstruction qualityVSAvoidprocessing time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary artifact reduction processing before final iterative reconstruction. By pre-processing the data to remove artifacts caused by restricted angular range, the method improves the quality of input data for iterative reconstruction, thereby reducing the number of iterations needed and decreasing overall processing time

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent divides the reconstruction process into separate segments: an initial artifact reduction step using machine learning, followed by a refined reconstruction step. This segmentation allows each algorithm to focus on its strength—artifact suppression and detailed reconstruction respectively—improving overall efficiency and quality

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS11308664B2Method for reconstructing a three-dimensional image data set

Publication Date: 2022.04.19 SIEMENS HEALTHINEERS AG

AI summary

Systems and methods are provided for reconstructing a three-dimensional result image data set from computed tomography from a plurality of two-dimensional images that create an image of an object undergoing examination from a particular imaging angle, The imaging angles of all the images lie within a restricted angular range. A three-dimensional artifact-reduced image data set is provided based on the two-dimensional images using an algorithm for reducing artifacts that are the result of a restriction of the angular range. The result image data set is reconstructed using a reconstruction algorithm that processes both the artifact-reduced image data set and the two-dimensional images as input data.