DCNN Attenuation Map Generation for PET/CT Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current PET imaging techniques face accuracy limitations in attenuation estimation due to varying effective X-ray energy and inadequate material classification, leading to beam-hardening artifacts and metal artifacts, which affect the accuracy of CT reconstruction and PET image quality.
Innovation Solution
A deep convolutional neural network (DCNN) is trained to directly translate CT images into attenuation maps for 511 keV gamma rays, capturing signatures of different material types and artifacts, and can be used with spectral CT data or conventional CT data to generate accurate attenuation maps for PET reconstruction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional method using CT numbers and material classification is used, then the process is simple and fast, but the accuracy of attenuation estimation is limited due to beam-hardening artifacts and metal artifacts
Solution Approach 1:
The patent replaces the traditional mechanical/mathematical system of CT number conversion and material classification with a deep learning-based neural network system. The neural network directly translates CT images into attenuation maps for 511 keV gamma rays, capturing material signatures and artifacts automatically, thereby achieving higher accuracy without manual intervention in the complex conversion process.
Solution Approach 2:
The patent changes the energy parameter from the variable effective X-ray energy in CT scans to a fixed 511 keV gamma ray energy parameter. By training the neural network to output attenuation coefficients specifically for 511 keV photons, the system resolves the beam-hardening artifact problem caused by energy variation while maintaining computational efficiency.
2Measurement precision
If spectral CT with material decomposition is used, then accurate attenuation maps can be generated, but the scanning time and hardware complexity increase
Solution Approach 1:
The patent performs preliminary action by training the neural network on a comprehensive dataset of CT images with known ground truth attenuation maps during the development phase. This pre-training allows the network to encode complex material decomposition logic and artifact correction algorithms, enabling it to accurately translate conventional single-energy CT scans into PET attenuation maps without requiring time-consuming spectral scanning during actual patient imaging.
3Adaptability or versatility
If conventional CT data is used, then the workflow is simple and widely applicable, but beam-hardening and metal artifacts significantly reduce PET image quality
Solution Approach 1:
The patent converts the harmful beam-hardening and metal artifacts present in conventional CT scans into beneficial training signals for the neural network. By including these artifacts in the training data alongside ground truth attenuation maps, the network learns to recognize and correct these artifacts during inference, effectively transforming the previously harmful factors into training resources that improve generalization across different scan protocols and patient populations.
Data Source
AI summary
A method is provided for generating an attenuation map for PET image reconstruction. The method includes training a deep convolutional neural network (DCNN) model by minimizing a loss function between initial input image data and the attenuation map generated by a spectral CT scan as supervised data. Further, the method includes obtaining PET data from a scan of a subject and reconstructing a PET image from the PET data and an attenuation map output from the DCNN. The initial input image data can be from a conventional CT scan with or without beam-hardening correction or the input image data can be from a phantom, a simulation, or a SPECT image.