Photon Counting Detector Zero-Count Error Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Photon counting detector (PCD) systems in computed tomography (CT) imaging face challenges with zero counts, which lead to division-by-zero errors and increased noise, particularly in low-dose imaging, due to their design and operation, affecting image quality and radiation dose efficiency.
Innovation Solution
A method and system that identify and replace zero counts in CT data with non-zero numbers, estimate and remove biases, and reconstruct images to maintain spatial resolution and reduce statistical biases, using a probability function and sinogram estimator to correct for pre-log and post-log biases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If photon counting detectors are used to reduce radiation dose and improve noise rejection, then radiation dose efficiency and noise rejection are improved, but zero-count errors and division-by-zero errors increase
Solution Approach 1:
The patent converts the harmful zero-count errors into a manageable statistical problem by modeling them with a Poisson distribution. Instead of treating zero counts as fatal errors, the invention uses probabilistic modeling to predict and correct for them, transforming a harmful artifact into a quantifiable parameter that can be compensated for in the image reconstruction process.
Solution Approach 2:
The patent changes the parameter representation from raw count values to probability distributions. By modeling detector counts as Poisson-distributed random variables and working with their probability distributions rather than deterministic values, the system can handle zero counts gracefully and propagate uncertainty through the reconstruction process, eliminating division-by-zero errors while preserving statistical information.
2Measurement precision
If smaller detector pixels are used to increase spatial resolution, then spatial resolution is improved, but count statistics deteriorate leading to more zero counts
Solution Approach 1:
The patent transforms the problem by changing from deterministic count values to probabilistic parameters. Each detector pixel's count is represented as a Poisson-distributed random variable characterized by its mean and variance. This parameter transformation allows the system to work with sub-unity expected counts without encountering zero-count failures, as the probabilistic framework naturally handles cases where the expected value is less than one.
Solution Approach 2:
The patent introduces probability distributions as an intermediary layer between the raw detector counts and the image reconstruction process. This intermediary representation allows the system to bridge the gap between low photon counts and the requirements of image reconstruction algorithms, smoothing out the discontinuities caused by zero counts while preserving the statistical nature of the measurement process.
3Adaptability or versatility
If multiple energy bins are used for spectral CT imaging, then spectral imaging capability is improved, but zero-count probability increases
Solution Approach 1:
The patent segments the energy spectrum into multiple bins while applying probabilistic modeling to each bin independently. By treating each energy bin's counts as a separate Poisson-distributed random variable, the system can handle the reduced counts in each bin without encountering zero-count errors. The segmentation is performed in the probabilistic domain, allowing the system to manage the increased zero-count probability inherent in multi-bin spectral imaging.
4Reliability
If conventional zero-count replacement methods are used, then division-by-zero errors are avoided, but statistical bias and noise are introduced
Solution Approach 1:
The patent implements feedback by using the known Poisson distribution properties to guide the replacement of zero counts. Instead of arbitrary replacement, the system uses the statistical properties (mean and variance) of the Poisson distribution to determine appropriate replacement values. This feedback mechanism ensures that zero-count replacements are consistent with the underlying physics and do not introduce artificial biases into the data.
Solution Approach 2:
The patent changes the approach from deterministic replacement to probabilistic parameter adjustment. Rather than replacing zero counts with fixed values, the system adjusts the parameters of the Poisson distribution to account for zero counts in a way that preserves statistical accuracy. This parameter-based approach allows the system to maintain computational stability while avoiding the statistical biases introduced by conventional replacement methods.
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
The solution effectively reduces CT number bias and noise variance, preserving image quality and spatial resolution, as demonstrated by improved phantom imaging results, especially in low-dose PCD-CT imaging, without degrading image features or requiring iterative processing.
Implementation Method 1
The intensity of the radiation received by each detector element is dependent upon the attenuation of the x-ray beam by the object
Implementation Method 2
the general one-to-one correspondence between the digital output of the PCD with the number of absorbed x-ray photons
Data Source
AI summary
A system and method for creating computed tomography (CT) images that includes acquiring or accessing CT data of a subject, identifying zero counts in the CT data, and replacing the zero counts in the CT data with at least one non-zero number to create zero-count free CT data. The method also includes estimating a probability function of the zero-count free CT data, removing bias in the zero-count free CT data using the probability function, and reconstructing the zero-count free CT data after removal of the bias to create a corrected image of the subject with preserved conditional independence and spatial resolution.