Gap-Filler Sinogram Processing for Lower-Dose Dual-Energy CT
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Dual-energy CT methods incur increased dose, exposure time, and equipment costs, limiting their economic viability in many applications, and existing gap-filling methods for incomplete sinograms are inadequate.
Innovation Solution
A processing device uses deep neural networks to generate gap-filler sinogram information, combining first and second sinogram information to fill gaps and enhance resolution, reducing the need for additional scans and hardware.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If dual-energy CT methods are used to obtain additional material information and correct beam hardening artifacts, then image quality and material differentiation capabilities are improved, but radiation dose and exposure time are doubled
Solution Approach 1:
The patent applies partial action by capturing only a subset of projection images at the second spectral parameter rather than all projections. Specifically, it captures projections at selected angular positions and uses deep neural networks to generate the missing projections, thereby reducing the total number of X-ray exposures while maintaining sufficient data for high-quality dual-energy CT reconstruction and material differentiation
Solution Approach 2:
The patent uses deep neural networks to generate synthetic projection images that copy and replicate the characteristics of actual X-ray projections. The neural network learns from captured projections and synthesizes missing projections at the second spectral parameter, creating virtual data that mimics real measurements without requiring additional physical X-ray exposures
2Adaptability or versatility
If dual-energy CT methods are used to obtain additional material information, then material differentiation capabilities are improved, but equipment costs and design complexity increase
Solution Approach 1:
The patent makes the existing single-energy CT system multi-functional by enabling it to perform dual-energy CT capabilities through software-based deep neural network processing. Instead of requiring separate hardware systems for dual-energy imaging, the same detector and X-ray source are used to capture projections at different spectral parameters, with the neural network providing the additional processing functions for material differentiation
Solution Approach 2:
The patent replaces complex mechanical dual-energy CT hardware (such as dual X-ray tubes or spectral filtering mechanisms) with a computational approach using deep neural networks. The system uses a single X-ray source and detector, capturing projections at different energies and using AI-based processing to simulate the functionality of more complex mechanical dual-energy systems
3Measurement precision
If the number of projection images captured is increased to improve CT image quality, then image resolution and accuracy are improved, but exposure time and work cycle duration increase
Solution Approach 1:
The patent captures projections at only selected angular positions rather than all possible angles, using deep neural networks to generate the missing projection data. This partial sampling approach reduces the total number of X-ray exposures and exposure time while the neural network interpolation ensures sufficient image quality for accurate CT reconstruction
Solution Approach 2:
The deep neural network generates synthetic projection images that replicate the appearance and information content of actual X-ray projections. These generated projections fill in the gaps from missing angular views, providing complete sinogram data for high-quality CT reconstruction without requiring additional physical measurements
Data Source
AI summary
A processing device for obtaining gap-filler sinogram information is adapted to obtain first sinogram information associated with a first spectral parameter, and second sinogram information associated with a second spectral parameter, and is adapted to obtain the gap-filler sinogram information which is associated with the second spectral parameter and which fills the one gap in the second sinogram information, on the basis of the first sinogram information and the second sinogram information. A computer program, a method, and a computer implementation are also described.