Sinogram-Based Scatter Correction in CT Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computer tomography systems face challenges in reducing image artifacts caused by scattered radiation, which are not fully addressed by existing methods such as scattered ray gratings, as they reduce detective quantum efficiency and are costly, and require precise alignment, making them ineffective for systems with multiple x-ray sources or focal points.
Innovation Solution
A method for scattered ray correction of projection measurement data using a trained identification algorithm that localizes and corrects scatter distributions in sinograms, reducing the need for resource-intensive calculations and scattered ray gratings, and enabling effective correction in systems with minimal inverse geometry or multiple focal points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If scattered ray gratings are used to suppress scattered radiation, then scattered radiation is reduced, but detective quantum efficiency is reduced and manufacturing costs increase
Solution Approach 1:
The patent replaces the mechanical scattered ray grating system with a computational correction method. Instead of using physical gratings to suppress scattered radiation, the system uses trained identification algorithms (neural networks) to identify and correct scatter distributions in projection measurement data, thereby eliminating the need for scattered ray gratings while maintaining scattered radiation reduction effectiveness
Solution Approach 2:
The patent uses trained identification algorithms that have been trained on simulated or measured scatter distributions to create a model that can identify and correct scattered radiation patterns. This computational copy of the scatter distribution allows for correction without physical intervention, preserving detective quantum efficiency while reducing scattered radiation effects
2Object-affected harmful factors
If scattered ray gratings are used to suppress scattered radiation, then scattered radiation is reduced, but manufacturing costs increase significantly
Solution Approach 1:
The patent replaces expensive scattered ray grating manufacturing and alignment systems with software-based identification algorithms. The computational approach eliminates the need for precision-manufactured optical components, significantly reducing manufacturing costs while maintaining the ability to suppress scattered radiation effects in the reconstructed images
Solution Approach 2:
The patent uses computationally inexpensive identification algorithms that can be trained once and applied repeatedly without degradation. The software-based solution replaces expensive, complex scattered ray grating systems with affordable computational methods that have no physical wear or degradation
3Object-affected harmful factors
If scattered ray gratings are used, then scattered radiation is reduced, but precise alignment with x-ray source is required, reducing adaptability to different system configurations
Solution Approach 1:
The patent replaces the mechanically complex scattered ray grating alignment system with a software-based identification algorithm. The algorithm processes projection measurement data from any x-ray source configuration without requiring physical alignment, enabling the system to work with multiple x-ray sources, focal points, and geometric arrangements including minimal inverse geometry and fourth-generation systems
Solution Approach 2:
The patent creates a universal identification algorithm that can handle various x-ray source configurations, detector arrangements, and scan geometries. The trained neural network model adapts to different system configurations without requiring reconfiguration of physical components, providing versatile scattered radiation correction across multiple CT system types
4Measurement precision
If Monte Carlo methods are used to calculate scattered radiation, then exact scattered ray distribution is obtained, but computing effort is considerable
Solution Approach 1:
The patent performs preliminary training of identification algorithms using Monte Carlo methods or measured data to create a trained model. Once trained, the algorithm can quickly identify scatter distributions in clinical scans without requiring extensive computing effort during actual operation. The computationally intensive work is done once during training, not repeatedly during scanning
Solution Approach 2:
The patent creates a computational model (copy) of the scatter distribution through training on simulated or measured data. This trained model can then rapidly identify and correct scatter in clinical scans without requiring repeated Monte Carlo simulations, significantly reducing computing effort while maintaining accuracy
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 improves image quality by accurately identifying and correcting scatter distributions, reducing computing effort and manufacturing costs, and allowing for the use of computer tomography systems with multiple x-ray sources or focal points without the need for scattered ray gratings.
Implementation Method 1
the energy of the photons emitted by the x-ray source is changed by the scattering process, e.g. Compton scattering
Data Source
AI summary
A method is for scattered ray correction of projection measurement data recorded by a computer tomography system. In an embodiment, the method includes recording, localizing, identifying and correcting. In the recording, projection measurement data is recorded from a plurality of projection angles and the projection measurement data is captured in a sinogram. In the localizing, features in the projection measurement data of the sinogram are localized in a defined angle range about a projection angle. In the identifying, a scatter distribution for the projection angle is identified from the localized features by way of a trained identification algorithm. In the correcting, the projection measurement data of the projection angle is corrected on the basis of the scatter distribution.