CT Noise Deletion via Projection Space Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for de-noising in CT imaging systems, such as weighted averages or iterative cost-optimization, often blur the original signal while reducing noise, leading to a loss of contrast in low SNR conditions.
Innovation Solution
A method that estimates noise in image space, forward projects this estimate, modifies it using statistical properties in projection space, and subtracts it from the original data to generate noise-removed projection data, preserving signal contrast through a non-linear operator like hard thresholding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If weighted average or iterative cost-optimization methods are used to reduce noise, then noise is averaged out, but contrast among neighboring pixels is also averaged out resulting in blurred signal
Solution Approach 1:
The patent segments the noise reduction process into two distinct phases: (1) estimating noise statistics from the noisy data, and (2) using these statistics to guide a non-linear denoising operation. This segmentation allows the system to treat noise estimation and signal preservation as separate tasks, avoiding the blurring effect of traditional single-step filtering methods.
Solution Approach 2:
The patent changes the parameter space by moving from direct pixel-value manipulation to noise-statistic-based processing. By estimating noise standard deviation and using it to adaptively control the denoising strength, the system dynamically adjusts processing parameters based on local noise characteristics rather than applying fixed filtering kernels that blur edges.
2Productivity
If lower amounts of photons are used to reduce x-ray dose and increase gantry speed, then image acquisition time is reduced and patient exposure is lowered, but signal-to-noise ratio decreases resulting in increased statistical noise
Solution Approach 1:
The patent performs preliminary noise estimation before the actual denoising operation. By first analyzing the noisy projection data to estimate noise statistics (standard deviation, mean), the system prepares the necessary information to guide subsequent denoising, ensuring that noise reduction can be effectively applied even to low-photon-count data without requiring re-acquisition.
Solution Approach 2:
The patent implements a feedback mechanism where noise statistics estimated from the noisy data are used to control the denoising process. The estimated noise standard deviation feeds into the denoising algorithm to adaptively determine filtering strength, creating a closed-loop system that responds to the actual noise level in the acquired data.
Data Source
AI summary
An imaging system includes a computer programmed to reconstruct original CT projection data, estimate noise in image space, forward project the image noise estimate to generate an initial projection noise estimate, modify the initial projection noise estimate using a statistical property of noise in projection space, remove noise in the original CT projection data by subtracting the modified noise estimate therefrom to generate noise-removed projection data, and reconstruct a final image based on the noise-removed projection data.