X-ray Cone-Beam CT Scatter Correction via Neural Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current x-ray cone-beam CT image reconstruction methods suffer from reduced image contrast due to scattered x-rays, making accurate tumor positioning and soft tissue contouring challenging, especially in radiotherapy applications.
Innovation Solution
A method involving the calculation of scattering projection images using a Monte Carlo method or Boltzmann's transport equation, followed by end-to-end training of a multi-layered neural network to estimate and subtract scattering components from measured projection images, resulting in scatter-free cone-beam CT images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If scattered x-rays are detected along with direct x-rays by the flat panel detector, then the detector captures complete projection data, but the image contrast is reduced making tumor positioning and soft tissue contouring challenging
Solution Approach 1:
The patent segments the projection data by separating scattered x-rays from direct x-rays using Monte Carlo simulation. The simulation calculates scattering components that are then subtracted from measured projection images, effectively segmenting the mixed signal into useful direct beam data and harmful scattered radiation data, thereby restoring image contrast while preserving complete projection data
Solution Approach 2:
The patent introduces Monte Carlo simulation as an intermediary computational model that predicts scattered radiation patterns. This virtual mediator allows the system to estimate and remove scattering effects without physically blocking any x-rays, thus maintaining data completeness while eliminating the contrast-reducing scattered components through computational subtraction
2Object-affected harmful factors
If a grid is used to reduce scattered x-rays, then some scatter reduction is achieved (less than 50%), but the grid cannot sufficiently improve image contrast and reduces less than half of scattered x-rays
Solution Approach 1:
The patent replaces the mechanical grid system with a computational approach using Monte Carlo simulation and neural networks. Instead of physically filtering scattered x-rays with a mechanical grid that achieves less than 50% reduction, the system uses simulation-based prediction and digital subtraction to achieve superior scatter removal and image contrast improvement
Solution Approach 2:
The patent changes the approach from physical parameter modification (grid density, material) to computational parameter optimization. By using Monte Carlo simulation parameters and neural network training, the system achieves adaptive scatter correction that dynamically adjusts to different anatomical regions and imaging conditions, surpassing the fixed performance of physical grids
3Productivity
If cone-beam CT images are reconstructed from projection images containing scattered x-rays, then the reconstruction process is straightforward, but the resulting images have reduced contrast making soft tissue contouring more challenging
Solution Approach 1:
The patent performs preliminary scatter correction by using Monte Carlo simulation to calculate scattering components before the final image reconstruction. The scattered radiation patterns are predicted and subtracted from projection data in advance, so that the subsequent reconstruction process works with already-corrected data, maintaining reconstruction speed while improving soft tissue contouring 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 significantly enhances image contrast, enabling more accurate tumor positioning and facilitating online adaptive radiotherapy planning by improving visibility of cancer tissues and nearby organs at risk.
Implementation Method 1
calculating scattering component images within projection images (hereinafter, referred to as 'scattering projection images') of a predetermined number of patients from predetermined x-ray beam angles
Implementation Method 2
the calculation by a Monte Carlo method or a Boltzmann's transport equation is calibrated by a phantom experiment before subtracting each scattering projection image from each measured projection image
Implementation Method 3
the calculation by a Monte Carlo method or a Boltzmann's transport equation is calibrated by a phantom experiment before subtracting each scattering projection image from each measured projection image
Implementation Method 4
The flat panel detector detects not only direct x-rays but also scattered x-rays
Data Source
AI summary
An improved x-ray cone-beam CT image reconstruction by end-to-end training of a multi-layered neural network is proposed, which employs cone-beam CT images of many patients as input training data, and precalculated scattering projection images of the same patients as output training data. After the training is completed, scattering projection images for a new patient are estimated by inputting a cone-beam CT image of the new patient into the trained multi-layered neural network. Subsequently, scatter-free projection images for the new patient are obtained by subtracting the estimated scattering projection images from measured projection images, beam angle by beam angle. A scatter-free cone-beam CT image is reconstructed from the scatter-free projection images.