Position Spectrum Segmentation for Deformed PET Detector Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for establishing position spectra in PET imaging equipment suffer from inaccuracies due to nonlinearity, inconsistent crystal specifications, and Compton scattering, leading to blurry edges, severe deformation, and low accuracy in identifying position spectra generated by different radiation detectors.
Innovation Solution
A method involving pre-processing with Gaussian templates, Hessian matrix calculation, morphological expansion, peak search, and event clustering to generate a crystal lookup table, ensuring accurate identification of position spectra.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional scatter plot methods are used to establish position spectrum, then the process is simple and intuitive, but the accuracy is low due to nonlinearity and Compton scattering
Solution Approach 1:
The patent applies preliminary action by performing pre-processing on the position spectrum before peak detection. Specifically, it uses Gaussian filtering to smooth the spectrum and enhance the visibility of peak points, and applies morphological operations to prepare the data structure. This preliminary processing simplifies the subsequent peak detection process while improving accuracy, resolving the contradiction between measurement precision and device complexity.
Solution Approach 2:
The patent introduces an intermediary approach by using a two-stage processing pipeline: first using Gaussian filtering as an intermediate step to preprocess the position spectrum, then applying peak detection algorithms on the enhanced data. This intermediary processing layer mediates between the raw data and the final analysis, improving peak detection accuracy without directly increasing the complexity of the main algorithm.
2Measurement precision
If multiple mean value deletion methods are used during local peak search, then the accuracy improves, but false deletions occur and true peak points may be missed
Solution Approach 1:
The patent implements feedback mechanisms through iterative peak detection and validation processes. After initial peak detection, the algorithm performs validation steps that compare detected peaks against the preprocessed spectrum and use feedback from morphological operations to correct false detections. This feedback loop ensures that only valid peak points are accepted, preventing false deletions while maintaining high accuracy.
Solution Approach 2:
The patent replaces mechanical iterative deletion methods with a more elegant mathematical approach using Gaussian filtering and morphological operations. Instead of mechanically deleting mean values and risking false deletions, the system uses continuous mathematical transformations that smoothly enhance peak points and preserve true peaks while removing artifacts, thereby improving both accuracy and reliability.
3Ease of operation
If semi-automatic methods with watershed algorithm are used, then peak points can be obtained, but edges become blurred and deformation occurs
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting processing parameters based on the input data characteristics. It uses adaptive Gaussian filtering with variable sigma values and adjusts morphological operation parameters to match the specific features of each position spectrum. This adaptive parameter adjustment allows the system to maintain clear edges and reduce deformation while achieving fully automatic processing, resolving the contradiction between ease of operation and segmentation 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
Enables fully automatic, high-efficiency, and high-accuracy segmentation of position spectra with clear edges, effectively handling position spectra from various structural detectors.
Implementation Method 1
processing the initial position spectrum by means of convolution with the first Gaussian template to generate the first position spectrum
Implementation Method 2
finding Hessian matrices by a convolution of each second Gaussian template and the first position spectrum
Implementation Method 3
the second position spectra are morphologically expanded to generate expanded position spectra
Implementation Method 4
performing a peak search on the second position spectra to obtain N peak points
Implementation Method 5
clustering single events based on the peak points in the sixth position spectrum to form a final position spectrum
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
The invention is related to a method, device, and computer storage medium for identifying position spectrum. The method comprises: pre-processing an initial position spectrum to generate a first position spectrum; extracting feature from the first position spectrum to generate a plurality of second position spectra; performing a peak search on the second position spectra to obtain N peak points, which form a third position spectrum; globally numbering the peak points in the third position spectrum to form a sixth position spectrum; and clustering single events based on the peak points in the sixth position spectrum to form a final position spectrum. The device comprises a pre-processing unit, a feature extraction unit, a peak search unit, a global numbering unit and an event clustering unit. The above method may be realized when executing programs in a computer storage medium. In the invention, any position spectrum may be fully automatically processed, with the capacity of high efficient and high accurate identification of the position spectrum with fuzzy edges and severe deformation, and it is possible to realize the effective identification of the position spectrum generated by various structural detectors.