Low Dose CT Denoising via Photon-Count Projection Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computed tomography (CT) scanners face challenges in reducing radiation dose without increasing image noise, which degrades image quality, especially in low-dose scans where noise is strongly correlated and iterative reconstructions are computationally expensive.
Innovation Solution
A projection data de-noiser system that estimates the number of detected photons and applies a de-noising algorithm preferentially to projections with lower photon counts, using a total variation-minimization approach to reduce streaks and bias while preserving edges, thereby improving image quality without the heavy computational burden of iterative methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If radiation dose is reduced by decreasing tube current, tube voltage and/or number of scans, then patient radiation exposure is reduced, but image noise increases which degrades image quality
Solution Approach 1:
The patent applies de-noising algorithms to projection data before the reconstruction process. By pre-processing the projection data to reduce noise and streak artifacts, the system enables lower radiation dose scans to produce diagnostic-quality images without requiring iterative reconstruction methods that would be computationally prohibitive.
Solution Approach 2:
The patent replaces iterative reconstruction methods with a direct de-noising approach applied to projection data. Instead of using computationally expensive iterative algorithms to reconstruct and then clean images, the system directly processes projection data to remove noise and streak artifacts, achieving the same goal with minimal computational burden.
2Manufacturing precision
If iterative reconstruction methods are used to reduce noise, then image quality improves, but computational cost increases significantly
Solution Approach 1:
The patent replaces iterative reconstruction methods with a direct de-noising approach applied to projection data. Instead of using computationally expensive iterative algorithms to reconstruct and then clean images, the system directly processes projection data to remove noise and streak artifacts, achieving the same goal with minimal computational burden.
Solution Approach 2:
The patent applies de-noising algorithms to projection data before the reconstruction process. By pre-processing the projection data to reduce noise and streak artifacts, the system enables lower radiation dose scans to produce diagnostic-quality images without requiring iterative reconstruction methods that would be computationally prohibitive.
3Manufacturing precision
If de-noising algorithms are applied to low-dose CT data, then image noise is reduced, but streak artifacts and bias remain due to strongly correlated noise between neighboring voxels
Solution Approach 1:
The patent applies different de-noising strategies to different regions of the projection data based on their photon count characteristics. By analyzing the local photon count distribution and applying adaptive de-noising parameters to regions with low photon counts versus high photon counts, the system effectively reduces streak artifacts and bias while preserving image quality in regions with sufficient photon statistics.
Data Source
Figure 1
Figure 2~3
AI summary
A system includes a source that rotates about an examination region and emits radiation that traverses the examination region, a radiation sensitive detector array that detects radiation traversing the examination region and generates projection data indicative of the detected radiation, and a projection data de-noiser that de-noises the projection data, wherein the de-noiser de-noises a projection based on a number of detected photons for the projection.