Cascaded Neural Network Attenuation Correction for PET Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current PET image reconstruction methods are time-consuming and lack real-time capabilities, hindering efficient PET scans and resource utilization due to the need for extensive data processing and lack of real-time information about the scanner's operating conditions.
Innovation Solution
A system utilizing a trained cascaded neural network for attenuation correction, which processes PET and CT images to generate attenuation-corrected PET images rapidly, enabling real-time browsing and improving scan efficiency by preprocessing, registering, and normalizing images, and using convolutional or generative adversarial neural network models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional data processing operations are performed on PET data during reconstruction, then attenuation correction accuracy is improved, but image reconstruction time increases significantly
Solution Approach 1:
The system performs preliminary actions by pre-processing PET and CT images (normalization, registration) before the main reconstruction process, and uses pre-trained neural network models to accelerate the attenuation correction process during actual scanning, thereby reducing real-time computation burden while maintaining accuracy
Solution Approach 2:
The patent replaces traditional mathematical modeling and iterative calculation methods with deep learning-based neural network models. The neural networks are trained offline using large datasets, and during clinical scanning, they rapidly infer attenuation correction results without requiring extensive real-time mathematical computations, thus substituting mechanical calculation with intelligent inference
2Manufacturing precision
If extensive data processing operations are performed during PET scan reconstruction, then image quality is improved, but scan efficiency decreases
Solution Approach 1:
The system performs necessary data processing operations in advance by pre-processing images (normalization, registration, concatenation) before scanning, and uses pre-trained neural network models that have already learned optimal processing patterns, enabling rapid inference during scanning without compromising image quality
Solution Approach 2:
Traditional iterative reconstruction algorithms and mathematical modeling operations are replaced with deep learning-based neural networks that can process images much faster while maintaining or improving quality through learned features from training data
3Ease of operation
If real-time browsing of PET images is implemented, then user control over scan process is improved, but computational resources increase
Solution Approach 1:
The system performs computationally intensive operations in advance by pre-processing images and pre-training neural network models offline, so that during scanning and real-time browsing, only lightweight inference operations are required, reducing real-time computational resource consumption while enabling user interaction
Data Source
AI summary
System for image correction in PET is provided. The system may acquire a PET image and a CT image of a subject. The system may generate, based on the PET image and the CT image, an attenuation-corrected PET image of the subject by application of an attenuation correction model. The attenuation correction model may be a trained cascaded neural network including a trained first model and at least one trained second model downstream to the trained first model. During the application of the attenuation correction model, an input of each of the at least one trained second model may include the PET image, the CT image, and an output image of a previous trained model that is upstream and connected to the trained second model.