Spectral CT Material Decomposition via Landmark Auxiliary Variables

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Projection domain decomposition in dual-energy CT introduces strong negatively correlated noise, making it challenging to reduce low-frequency noise without smearing the image or introducing bias, which degrades spectral image quality.

Innovation Solution

The approach involves constructing auxiliary variables, or landmarks, in the low/high energy image domain to capture beam hardening distortions, estimating changes in pixel values with added materials like soft tissue, bone, or iodine, and generating air values to reconstruct pixel compositions, thereby mitigating beam hardening distortions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Object-affected harmful factors

If projection domain decomposition is used to reduce beam hardening distortions, then beam hardening artifacts are reduced, but strong negatively correlated noise is introduced between Compton scatter and photoelectric effect line integrals

Engineering Contradiction:
Improvebeam hardening artifactsVSAvoidnoise correlation
Core Design Contradiction:
Object-affected harmful factorsVSReliability

Solution Approach 1:

The patent extracts and removes the beam hardening effect from the projection data before decomposition. By modeling and subtracting the beam hardening component, the method eliminates the source of artifacts while preserving the underlying signal, thereby avoiding the introduction of strongly correlated noise that plagues conventional decomposition methods.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies preliminary beam hardening correction to the projection data before performing material decomposition. This preliminary action prepares the data by removing the non-linear beam hardening effects, ensuring that subsequent decomposition operates on corrected data and produces accurate material maps without excessive noise correlation.

Inventive Principle:
Principle #10Preliminary action

2Reliability

If a de-noising algorithm is applied to reduce noise, then noise is reduced, but image smearing or bias is introduced

Engineering Contradiction:
Improvenoise reductionVSAvoidimage accuracy
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The patent introduces material-specific basis images as intermediaries in the decomposition process. These basis images serve as reference templates that guide the decomposition algorithm, enabling effective noise reduction while preserving the accuracy of material quantification. The basis images act as a mediator between the noisy input data and the final material maps, preventing both noise propagation and image smearing.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Adaptability or versatility

If non-linear projection domain decomposition is performed on noisy input data, then material separation is achieved, but noise induced bias is generated that degrades spectral image quality

Engineering Contradiction:
Improvematerial separation capabilityVSAvoidspectral image quality
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent performs preliminary beam hardening correction on the projection data before material decomposition. This preliminary action removes the non-linear effects that would otherwise interact with noise to create bias. By correcting beam hardening first, the subsequent linear decomposition operates on corrected data, achieving accurate material separation without noise-induced bias.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent segments the decomposition process into distinct stages: beam hardening correction, noise reduction, and material decomposition. This segmentation allows each stage to be optimized independently, ensuring that material separation is achieved through linear decomposition on corrected data, thereby avoiding the noise-induced bias that results from applying non-linear decomposition directly to noisy data.

Inventive Principle:
Principle #1Segmentation

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 method effectively reduces beam hardening artifacts and noise-induced bias, improving the quality of spectral images by accurately determining material compositions without smearing the image or introducing bias.

Implementation Method 1

an x-ray tube configured to switch between a first peak emission spectrum and a second peak emission spectrum

Methodology Applied
Scientific EffectX-ray emission: X-Ray

Implementation Method 2

photoelectric effect line integrals

Methodology Applied
Scientific EffectPhotoelectric effect: Photoelectric Effect

Implementation Method 3

Compton scatter line integrals

Methodology Applied
Scientific EffectCompton scattering: Compton Scattering

Data Source

PatentEP3268934B1Multi-energy (spectral) image data processing
Publication Date: 2023.08.09 KONINKLIJKE PHILIPS NV
  • EP3268934B1 patent drawingFigure 1
  • EP3268934B1 patent drawingFigure 2
  • EP3268934B1 patent drawingFigure 3~4

AI summary

A method includes generating a material landmark images in a low and high energy image domain. The material landmark image estimates a change of a value of an image pixel caused by adding a small amount of a known material to the pixel. The method further includes generating an air values image in the low and high energy image domain. The air values image estimates a value for each image pixel where a value of a pixel is replaced by a value representing air. The method further includes extracting from de- noised low and high images generated from the low and high line integrals, a material composition of each image pixel based on the material landmark images and air values image. The method further includes generating a signal indicative the extracted material composition.