CT Device Channel Calibration Using Adversarial Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
CT devices face issues with inaccurate estimation of high-frequency information due to defective channels, leading to streak and ring artifacts that degrade clinical diagnosis quality.
Innovation Solution
A method using a neural network to calibrate defective channels by acquiring original data, identifying to-be-recovered areas, normalizing and training with an adversarial neural network to produce accurate attenuation values, thereby improving image reconstruction quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If interpolation or averaging calculation is used to estimate attenuation at defective channels, then the calibration process is simple, but high-frequency information estimation becomes inaccurate leading to streak artifacts
Solution Approach 1:
The patent introduces an adversarial neural network as an intermediary between the defective channel data and the attenuation estimation process. The network consists of a generator that creates estimated attenuation values and a discriminator that validates them against high-frequency information requirements, thereby improving estimation accuracy without complicating the overall calibration workflow
Solution Approach 2:
The patent replaces the traditional mechanical/mathematical interpolation and averaging methods with a neural network-based system. This substitution enables the system to capture complex high-frequency patterns in the data that simple mathematical operations cannot, thereby eliminating streak artifacts while maintaining ease of operation through automated learning
2Ease of operation
If traditional interpolation methods are used for defective channel calibration, then the method is easy to implement, but significant errors occur in continuous damaged detector areas producing pale ring artifacts
Solution Approach 1:
The patent applies dynamics by making the calibration process adaptive rather than static. The neural network learns from the training data and dynamically adjusts its estimation strategy based on the pattern and extent of defective channels, allowing it to handle continuous damaged areas effectively while maintaining ease of operation through automated adaptation
Solution Approach 2:
The patent implements preliminary action by training the adversarial neural network on comprehensive datasets before actual calibration. This pre-training phase enables the network to learn from various defect patterns including continuous damaged areas, so that when calibration is performed, the network is already prepared to handle these challenging cases accurately
Data Source
AI summary
A method for calibrating defective channels of a CT device involves in a step S10, acquiring original data collected by the CT device; in a step S20, capturing to-be-recovered areas from the original data, wherein the to-be-recovered areas contain the defective channels of the CT device; in a step S30, inputting data of the to-be-recovered areas to a neural network for training so as to generate training results; and in a step S40, using the training results to repair the to-be-recovered areas. The method eliminates effects of artifacts caused by defective channels on image reconstruction.