Random Sinogram Variance Reduction in Continuous Bed Motion PET
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Continuous bed motion (CBM) acquisition in PET systems complicates variance reduction of random sinograms due to the summation over all detector pairs in the axial direction, leading to increased noise and artifacts in image reconstruction, as the singles rate is not constant and varies with time as the patient moves through the scanner.
Innovation Solution
Modeling random sinograms as a product of transverse singles efficiencies, treating the plurality of ring detectors as one ring along the axial direction, and decomposing sinogram plane randoms into two-dimensional transverse efficiencies to solve for a mean random sinogram, which is then used to constrain the random contribution and reduce variance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If continuous bed motion acquisition is used to improve axial uniformity, then axial uniformity of PET images is improved, but noise and artifacts increase due to random contribution variance
Solution Approach 1:
The patent applies preliminary action by estimating the mean random sinogram before image reconstruction. The system estimates randoms from the prompt sinogram using the relationship between prompt and random coincidences, then uses this estimated mean random sinogram to correct the data before reconstruction. This preliminary correction reduces the variance of random contributions and prevents noise and artifacts in the final image, while maintaining the axial uniformity benefits of continuous bed motion acquisition.
2Device complexity
If direct use of measured randoms measurements is used for reconstruction, then reconstruction is simplified, but artifacts and increased image noise levels occur
Solution Approach 1:
The patent introduces an intermediary approach by using the prompt sinogram as a mediator to estimate the mean random sinogram. Instead of directly using measured randoms measurements, the system uses the relationship between prompt and random coincidences to derive randoms estimation from prompt data. This intermediary estimation process reduces noise and artifacts while maintaining reconstruction feasibility, bridging the gap between direct measurement and quality correction.
3Measurement precision
If CBM acquisition is used to oversample the image, then super-resolution images are achieved, but effective resolution decreases due to high level noise
Solution Approach 1:
The patent applies preliminary action by performing randoms estimation and variance reduction before the reconstruction process. The system estimates the mean random sinogram from prompt data and uses this to correct the acquired data, reducing the noise level before reconstruction. This preliminary noise reduction preserves the super-resolution benefits of oversampling in continuous bed motion acquisition while improving the effective resolution by reducing high-level noise.
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
This approach reduces noise and artifacts in PET image reconstruction by smoothing the variance of random contributions, resulting in less noisy images and improved axial uniformity, even with low clinical scan durations.
Implementation Method 1
PET systems and corresponding detectors have a limited field of view. Typically, the entire patient cannot be scanned with the patient in one position.
Implementation Method 2
PET systems and corresponding detectors have a limited field of view
Data Source
AI summary
Random sinogram variance is reduced in continuous bed motion acquisition. The randoms are modeled as a product of transverse singles efficiencies. The random sinogram is assumed to be a smooth function in the axial direction, collapsing the parameterization for estimating the transverse singles efficiencies into a single, conceptual ring. By solving the product, the mean random values are used to smooth the randoms in image reconstruction with less noise and artifacts.