Convolutional Neural Network for Multi-Energy CT Material Decomposition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional material decomposition techniques from multi-energy computed tomography (MECT) data often amplify noise, decrease spatial resolution, and generate image artifacts, while being limited to differentiating only 2-3 base materials, which is insufficient for the complexity of the human body.
Innovation Solution
A system utilizing a trained convolutional neural network (CNN) to transform MECT data into virtual non-contrast enhanced, virtual non-calcium, and virtual monoenergetic images by predicting material mass densities, capable of differentiating multiple materials, including iodine, calcium, blood, adipose tissue, and hydroxyapatite, using a fidelity term and image-gradient-correlation regularization term in the loss function.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional material decomposition techniques are used, then material differentiation is achieved, but noise is amplified and spatial resolution decreases
Solution Approach 1:
The patent replaces conventional iterative material decomposition algorithms with a deep learning-based convolutional neural network (CNN) system. The CNN is trained on multi-energy CT data to directly predict material decomposition results, substituting the traditional mechanical/algorithmic decomposition process with an intelligent system that learns optimal decomposition strategies from training data, thereby avoiding noise amplification and resolution loss.
Solution Approach 2:
The patent implements preliminary training of the CNN model using multi-energy CT data from phantoms with known material compositions. This pre-training phase allows the system to learn accurate material decomposition mappings before actual clinical use, enabling the network to perform robust decomposition on new data without requiring iterative refinement that amplifies noise.
2Adaptability or versatility
If conventional material decomposition techniques are used, then 2-3 base materials can be differentiated, but this is insufficient for human body complexity
Solution Approach 1:
The patent designs a universal CNN-based material decomposition system that can simultaneously differentiate multiple material types (including iodine, calcium, uric acid, and soft tissues) within a single unified framework. The network is configured with multiple output channels, each corresponding to a different material class, enabling it to perform multiple decomposition functions concurrently rather than requiring separate specialized algorithms for each material type.
3Measurement precision
If conventional material decomposition techniques are used, then material quantification is achieved, but computational resources are excessive and processing is slow
Solution Approach 1:
The patent replaces computationally intensive iterative decomposition algorithms with a pre-trained CNN model that performs material quantification through direct forward propagation. Once trained, the network executes decomposition in a single pass through the input data, eliminating the need for repeated iterative calculations while maintaining accurate material quantification through learned feature representations.
Data Source
AI summary
In accordance with some embodiments, systems, methods, and media for material decomposition and virtual monoenergetic imaging from multi-energy computed tomography data are provided. In some embodiments, a system comprises: at least one hardware processor configured to: receive a multi-energy computed tomography (MECT) image of a subject; provide the MECT data to a trained convolutional neural network (CNN); receive output from the trained CNN indicative of predicted material mass density for each of a plurality of materials at each pixel location of the MECT data, wherein the plurality of materials includes at least four materials; and generate a transformed version of the MECT data using the output.