CT Detector Bad Pixel Correction with Deep Learning Sinograms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Bad pixels in computed tomography (CT) detectors result in missing data and undesirable artifacts in reconstructed images, rendering the CT system unusable or requiring costly replacements.
Innovation Solution
A deep learning-based method using convolutional neural networks processes sinogram data to estimate missing pixel values by leveraging complementary information within the sinogram domain, correcting artifacts through supervised learning in both the raw sinogram and reconstruction domains.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a CT detector contains bad pixels, then image quality deteriorates with artifacts and missing data, but the system becomes unusable or requires costly replacements
Solution Approach 1:
The patent converts the harmful effect of bad pixels by using the surrounding good pixel data to predict and reconstruct the missing bad pixel values. The deep learning network learns from the statistical relationships in the sinogram data to transform the defective data into useful reconstructed information, effectively converting the harm of bad pixels into a solvable prediction problem that restores system usability without hardware replacement
Solution Approach 2:
The patent introduces a deep learning-based prediction model as an intermediary between the defective sinogram data and the final reconstructed image. This intermediary network processes the incomplete data, predicts missing values, and outputs corrected sinogram data that can be used for normal image reconstruction, thereby mediating the impact of bad pixels on system usability
2Reliability
If traditional interpolation methods are used to estimate missing pixel values, then the system remains operational, but image quality and diagnostic accuracy deteriorate
Solution Approach 1:
The patent replaces traditional mechanical interpolation methods (such as nearest-neighbor, linear, or polynomial interpolation) with a deep learning-based neural network system. This substitution enables the system to learn complex non-linear relationships in the sinogram data and perform much more accurate predictions of missing pixel values, significantly improving measurement precision and image quality while maintaining system usability
Solution Approach 2:
The patent changes the fundamental parameter of how missing values are estimated - from simple mathematical interpolation formulas to sophisticated deep learning model predictions. By training the network on large datasets of sinogram patterns, the system adapts its prediction parameters to match the specific characteristics of the CT scanner and imaging protocols, achieving superior reconstruction accuracy compared to fixed interpolation algorithms
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer-implemented method for correcting artifacts in computed tomography data is provided. The method includes inputting a sinogram into a trained sinogram correction network, wherein the sinogram is missing a pixel value for at least one pixel. The method also includes processing the sinogram via one or more layers of the trained sinogram correction network, wherein processing the sinogram includes deriving complementary information from the sinogram and estimating the pixel value for the at least one pixel based on the complementary information. The method further includes outputting from the trained sinogram correction network a corrected sinogram having the estimated pixel value.