CT Image Reconstruction Noise Intensity Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current CT devices face challenges in accurately detecting noise intensity when applying the projection data division method to parallel beam reconstruction due to correlation between even-numbered and odd-numbered view images after fan-parallel conversion, leading to difficulties in balancing noise and signal intensity and reducing radiation exposure.
Innovation Solution
The method involves performing fan-parallel conversion on projection data, dividing it into sets, generating difference images, calculating pixel value variation indices, and correcting these indices with pre-calculated correction values to accurately determine noise intensity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If fan-parallel conversion is performed to enable parallel beam reconstruction, then spatial uniformity of resolution is improved, but correlation occurs between even-numbered and odd-numbered view images causing noise intensity detection accuracy to deteriorate
Solution Approach 1:
The patent applies preliminary action by pre-calculating correction values using simulation data before actual noise intensity measurement. The system stores correction values corresponding to different imaging conditions (tube voltage, tube current, rotation speed, focal spot size, detector element width, scan object diameter) and selects the appropriate correction value based on current imaging parameters. This preliminary preparation eliminates the need for complex real-time calculations and enables accurate noise intensity measurement despite the correlation introduced by fan-parallel conversion.
2Measurement precision
If projection data is divided into even-numbered and odd-numbered sets for noise intensity calculation, then noise intensity can be estimated, but correlation between the sets after fan-parallel conversion causes covariance to become non-zero reducing measurement accuracy
Solution Approach 1:
The patent uses copying by creating simulation data that replicates the fan-parallel conversion process and its associated correlation effects. Through Monte Carlo simulation, the system generates synthetic projection data and reconstructs images to calculate correction values that capture the statistical properties of the correlated difference images. These correction values are then applied to actual measurements, effectively copying the behavior of the simulation to real-world data.
3Object-affected harmful factors
If parameter values in filtering techniques are adjusted to decrease noise intensity, then noise is reduced, but signal intensity also decreases making it difficult to maintain both noise and signal quality
Solution Approach 1:
The patent implements feedback by using the accurately measured noise intensity (corrected using simulation-based correction values) to dynamically adjust filtering parameters. The system measures actual noise intensity in the reconstructed image, compares it with desired noise levels, and adjusts the strength of noise reduction filters accordingly. This feedback loop enables precise control of noise reduction while preserving signal intensity, avoiding the trial-and-error approach that leads to signal loss.
Data Source
AI summary
In order to determine noise intensity more accurately by a projection data division method using data after fan-parallel conversion, a plurality of projection data obtained by irradiating a scan object with radiations are received and subjected to prescribed data conversion; data after the conversion is divided into two or more sets; a reconstructed image is generated for each set of data; and the generated reconstructed image for each of the sets is subtracted to generate a difference image. An index indicating pixel value variation for at least one prescribed region on the difference image is calculated, the value of the index is corrected by a previously calculated correction value, and the corrected index value is regarded as noise intensity of the region.