PET Image Reconstruction via FBP-Net Denoising
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current PET image reconstruction methods, such as filtered back-projection and iterative algorithms, suffer from noise and low resolution, while deep learning approaches are difficult to interpret and require large amounts of high-quality data, making it challenging to achieve high-quality PET images, especially at low counting rates.
Innovation Solution
A PET image reconstruction method combining a filtered back-projection algorithm with an improved denoising convolutional neural network, where the reconstruction problem is split into sub-problems of image reconstruction and denoising, using a learnable frequency-domain filter and residual learning to produce clear and artifact-free images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If filtered back-projection algorithm is used for PET image reconstruction, then reconstruction speed is improved, but image quality deteriorates due to noise and wheel-like artifacts
Solution Approach 1:
The patent segments the reconstruction process into two distinct stages: first applying filtered back-projection for rapid initial reconstruction, then applying a neural network specifically for denoising and artifact removal. This segmentation allows each component to optimize for its specific function, maintaining speed while improving quality.
Solution Approach 2:
The neural network acts as an intermediary component between the filtered back-projection algorithm and the final image output. It processes the preliminary reconstructed image to remove noise and artifacts, serving as a mediator that enhances image quality without requiring a complete redesign of the reconstruction pipeline.
2Manufacturing precision
If iterative reconstruction algorithms are used to suppress noise, then image quality is improved, but computation time increases and speed decreases
Solution Approach 1:
The patent performs preliminary reconstruction using the fast filtered back-projection algorithm to obtain an initial image estimate before applying the neural network for denoising. This preliminary action avoids the need for slow iterative algorithms while still achieving high image quality through the subsequent neural network processing.
Solution Approach 2:
The patent replaces the mechanical iterative computation process with a neural network-based denoising system. Instead of repeatedly applying iterative algorithms to suppress noise, the system uses a trained neural network that can rapidly remove noise and artifacts in a single pass, substituting computational mechanics with learned patterns.
3Manufacturing precision
If deep neural networks are used for PET image reconstruction, then image quality is improved, but interpretability deteriorates and data requirements increase
Solution Approach 1:
The neural network is positioned as an intermediary denoising component rather than a complete reconstruction system, making its function more transparent and interpretable. It specifically targets noise and artifact removal rather than performing the entire reconstruction process, which enhances interpretability while maintaining image quality improvement.
Data Source
AI summary
A PET image reconstruction method, including: 1) injecting a PET radioactive tracer into a biological tissue, scanning by a PET device, and detecting and counting coincidence photons to obtain an original protection data matrix; 2) establishing a measurement equation model; 3) splitting the reconstruction problem into a first sub-problem and a second sub-problem; 4) solving the first sub-problem by a filtered back-projection layer, solving the second sub-problem by an improved denoising convolutional neural network, where the filtered back-projection layer and the improved denoising convolutional neural network are connected in series to form a filtered back-projection network (FBP-Net); 5) inputting original projection data into the FBP-Net, and using an image as a tag to adjust parameters of the FBP-Net to reduce an error between an output of the FBP-Net and the tag; and 6) inputting projection data to be reconstructed into the trained FBP-Net to obtain a desired reconstructed image.