Digitized Scintillation Pulse Fitting via Parameter Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing algorithms for fitting digitized scintillation pulse signals in PET systems are inefficient, unstable, and compromise on accuracy, leading to suboptimal time and energy resolution in medical imaging.
Innovation Solution
A computer-implemented method using the Levenberg-Marquardt algorithm with a double-exponential model for fitting digitized scintillation pulse signals, utilizing a scintillation pulse database with specific voltage and time thresholds, and initial parameters to improve fitting efficiency, accuracy, and stability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing least square algorithms (Newton method, gradient method) are used for fitting digitized scintillation pulse signals, then the fitting process can be performed, but the processing time is too long and efficiency is low
Solution Approach 1:
The patent transforms the original least square fitting problem into a different parameter space by introducing a transformation function that converts the non-linear fitting problem into a linear one, enabling faster computation while maintaining accuracy
2Productivity
If two parameters in a double-exponential model are selected for fitting to improve time efficiency, then processing speed increases, but pulse integrity is compromised and accuracy deteriorates
Solution Approach 1:
The patent applies a parameter transformation that allows all parameters of the double-exponential model to be determined through linear fitting rather than non-linear iteration, achieving both full parameter accuracy and computational efficiency simultaneously
3Reliability
If Levenberg-Marquardt algorithm is applied to fit digitized scintillation pulse signals, then fitting accuracy and stability are improved, but computational complexity increases
Solution Approach 1:
The patent extracts and separates the complexity by transforming the non-linear fitting problem into a linear one, removing the need for complex iterative algorithms like Levenberg-Marquardt while achieving comparable or superior accuracy and stability
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
A method for fitting digitized scintillation pulse signals, comprising following steps: selecting a scintillation pulse prior model to serve as double-exponential model; selecting sampling points for classical pulses in a scintillation pulse database, the sampling points comprising eight voltage thresholds sequence and eight time points sequence; carrying out time translation of the curve of the double-exponential model to obtain a time array; setting four initial default parameters of a target function to be 1 and storing them as an initial parameter array; inputting the time array, a scintillation target function and the like into a Levenberg-Marquardt fitting function, and fitting to obtain fitting parameters; sequentially repeating each scintillation pulse in the scintillation pulse database to complete the fitting of all scintillation pulses; integrating the fitted target function to obtain the energy value of a k-th pulse; and solving the equation to extract the time and energy information of the pulse. The energy resolution obtained by means of the present invention is remarkably improved, so that the accuracy and stability of information extraction are ensured while the processing efficiency is increased.