Continuous PET Reconstruction with Coordinate Descent for Convergence Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing PET reconstruction methods, such as ML-EM and OSEM, lack control over convergence speed and are computationally complex due to large matrix dimensions, making regularization strategies inconvenient and difficult to implement.
Innovation Solution
A continuous-to-continuous data model with iterative coordinate descent strategy is employed, using 2D FFT and IFFT algorithms to accelerate calculations and allow for 'early stopping' regularization, reducing computational complexity to 8I2 log2 I per iteration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ML-EM or OSEM algorithms are used for PET image reconstruction, then statistical accuracy is improved, but convergence speed control is lost and computational complexity increases
Solution Approach 1:
The patent segments the large system matrix into smaller blocks corresponding to different detector rings and angular views. By processing the reconstruction in segmented blocks rather than treating the entire large matrix at once, the computational complexity is reduced while maintaining statistical accuracy through iterative refinement of each segment.
Solution Approach 2:
The patent performs preliminary calculations of system matrix elements and their derivatives before the main iterative reconstruction process. This pre-computation of geometric factors and sensitivity weights allows the iterative algorithm to converge faster with controlled speed, reducing the overall computational burden while preserving statistical accuracy.
2Measurement precision
If large dimensionality matrices are used in discrete-to-discrete data model, then measurement accuracy is improved, but computational complexity and implementation difficulty increase
Solution Approach 1:
The patent changes the mathematical parameters by introducing continuous coordinate representations and analytical expressions for the system matrix elements. Instead of using discrete pixel-based matrices, the invention employs continuous spatial coordinates and analytical derivatives, which simplifies the implementation of regularization strategies while maintaining measurement accuracy.
Solution Approach 2:
The patent replaces the mechanical discrete matrix operations with analytical continuous mathematical operations. By substituting the discrete-to-discrete data model with a continuous-to-continuous approach using analytical expressions and their derivatives, the implementation becomes more flexible and easier to operate, particularly for regularization applications.
3Reliability
If iterative reconstruction algorithms are used, then statistical modeling accuracy is improved, but convergence control and regularization implementation become difficult
Solution Approach 1:
The patent introduces feedback mechanisms through the analytical derivatives of the system matrix elements with respect to image parameters. These derivatives provide feedback information about the sensitivity of measurements to image changes, enabling effective regularization control and convergence speed adjustment during the iterative reconstruction process while maintaining statistical modeling accuracy.
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 significantly reduces image artifacts and distortion, improves image resolution, and decreases radiotracer dosage while maintaining image quality, offering a feasible and high-quality reconstruction method.
Implementation Method 1
a patient is initially injected with a radiotracer, which contains bio-chemical molecules. These molecules are tagged with a positron emitting radioisotope... After the decay of these radioisotope molecules, positrons are emitted from the various tissues of the body... As a consequence of the annihilation of the positrons, pairs of gamma photons are produced
Implementation Method 2
In PET scanners, these pairs of photons are registered by detectors and counted
Data Source
AI summary
An iterative statistical algorithm based on a continuous-to-continuous data model with iterative coordinate descent optimization for image reconstruction from radiation measurements obtained in emission tomography, i.e. in a Positron Emission Tomography scanner, is described in this invention. The method presented here improves the resolution of the reconstructed images and/or decreases the tracer dosage absorbed by a patient during examination. At the same time, it maintains the quality of the functional images obtained. Additionally, this method allows for the presence of the regularization as an additive term. Furthermore, this method makes it possible to control the convergence of the algorithm using a specific parameter. These improvements are due to the signals obtained regarding the given statistics of this imaging technique.