Stacked Autoencoder for Mixed Tracer PET Image Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for reconstructing PET concentration distribution images of mixed tracers require separate injections and scans, leading to increased patient time, expense, and safety concerns, with poor image quality and inability to handle simultaneous injections due to hardware limitations.
Innovation Solution
A method using stacked autoencoders to reconstruct dynamic PET concentration distribution images by injecting two tracers simultaneously, employing coincidence counting matrices and the ML-EM algorithm, followed by training and fine-tuning neural networks to separate and reconstruct individual tracer images from mixed data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If traditional direct fitting or kinetic parameter estimation methods are used for mixed tracers PET image reconstruction, then calculation speed is improved, but image quality deteriorates
Solution Approach 1:
The patent introduces an interval matrix as an intermediary component that bridges the relationship between projection data and concentration distribution. By decomposing the system matrix into interval matrices representing different tracer components, the method enables simultaneous reconstruction of multiple tracers while maintaining both computational efficiency and image quality through the mediating mathematical structure
Solution Approach 2:
The patent segments the mixed tracer system into distinct component tracers by decomposing the system matrix into multiple interval matrices (A1, A2, ..., An), each representing a specific tracer's contribution. This segmentation allows independent reconstruction of each tracer's concentration distribution while maintaining the overall mixed tracer reconstruction efficiency
2Measurement precision
If traditional methods are used for mixed tracers PET imaging, then individual tracer imaging is achieved, but patient time and expense increase
Solution Approach 1:
The patent merges multiple tracer imaging procedures into a single simultaneous acquisition by combining the projection data from multiple tracers injected at different times. The interval matrix method mathematically separates the mixed signals while reconstructing individual tracer distributions in one unified process, eliminating the need for separate scans and significantly reducing patient time and expense
Solution Approach 2:
The patent creates a universal reconstruction framework that handles multiple tracers simultaneously through a single integrated algorithm. The method can process n different tracers with different injection times within one reconstruction process, making the system multi-functional and eliminating the need for separate imaging procedures for each tracer
3Measurement precision
If hardware-based photon separation is attempted for mixed tracers, then tracer separation is achieved, but device complexity increases due to hardware limitations
Solution Approach 1:
The patent replaces complex hardware-based photon separation mechanisms with a mathematical software solution. By using interval matrix decomposition and iterative reconstruction algorithms, the system achieves tracer separation through computational methods rather than physical hardware modifications, significantly reducing device complexity while maintaining separation accuracy
Solution Approach 2:
The patent introduces interval matrices as mathematical intermediaries that enable tracer separation without hardware modification. These interval matrices serve as mediators between the mixed projection data and the separated tracer images, providing a software-based solution that avoids the complexity of hardware-based photon energy discrimination or temporal separation systems
4Measurement precision
If interval between tracer injections is increased for traditional algorithms, then TAC curve analysis accuracy is improved, but productivity decreases
Solution Approach 1:
The patent performs preliminary decomposition of the system matrix into interval matrices before the reconstruction process. By pre-processing the data to separate tracer contributions mathematically, the method eliminates the need for long intervals between injections while maintaining TAC curve analysis accuracy, thereby improving reconstruction efficiency and productivity
Solution Approach 2:
The patent implements a dynamic reconstruction approach that adapts to different tracer injection scenarios. The iterative algorithm dynamically adjusts the reconstruction process based on the mixed projection data from multiple tracers with different injection times, maintaining accuracy without requiring fixed long intervals between injections
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 effective reconstruction of PET concentration distribution images for both tracers with simultaneous injection, improving image quality and reducing patient burden while addressing hardware limitations.
Implementation Method 1
The positrons generated during the decay react annihilate with negative electrons to generate a pair of gamma photons with energy of 511 kev in opposite directions
Implementation Method 2
The photons are recorded by the ring probe to generate projection data
Data Source
AI summary
The present invention discloses a method for reconstructing dynamic PET concentration distribution image of dual-tracer based on stacked autoencoder. The invention introduces deep learning into dynamic tracer PET concentration distribution image reconstruction, and the process is mainly divided into two stages of training and reconstruction. In the training phase, train multiple autoencoders using the concentration distribution images of mixed tracers taken as input, and the concentration distribution images of the two tracers taken as labels to build the stacked autoencoder. In the reconstruction phase, the concentration distribution images of the individual tracer can be reconstructed by inputting the concentration distribution images of the mixed traces to the well trained stacked autoencoder. The present invention realizes the reconstruction of the dynamic PET concentration distribution images of mixed tracers from the data-driven point of view, and effectively solves the problems of poor reconstruction effect and inability of simultaneous injection.