Helical CT Motion Estimation Using AI-Based Partial Angle Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Helical computed tomography (CT) scans face significant challenges in motion artifact reduction due to patient movement during scans, which existing techniques fail to adequately address, especially given the unique conditions of a helical CT scan where the field of view continuously changes.
Innovation Solution
A deep-learning based method for motion estimation and compensation that involves selecting center-points along the radiation source trajectory, identifying pairs of sections, reconstructing partial images, and performing image registration to estimate deformations representative of object motion during the scan, thereby generating a motion-compensated image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional motion compensation techniques are used for circular CT scans, then motion artifacts can be reduced in circular scan geometry, but these techniques fail to address the unique conditions of helical CT scans where the field of view continuously changes
Solution Approach 1:
The helical scan trajectory is segmented into multiple discrete sections along the z-axis, with motion estimation performed independently for each section. This segmentation allows the application of circular scan motion compensation techniques to each local section while accounting for the continuous change in field of view across the entire helical trajectory.
Solution Approach 2:
The system dynamically adjusts the motion compensation parameters for each section based on the local field of view conditions. By making the motion estimation and compensation adaptive to each section's specific geometry, the technique successfully applies circular scan methods to the dynamic helical scan environment.
2Measurement precision
If image registration is performed on full images, then comprehensive motion information can be obtained, but computational complexity and processing time increase significantly
Solution Approach 1:
The image registration process is applied to multiple small patches within each section rather than entire images. Each patch provides local motion information, and the combination of these local measurements achieves comprehensive motion estimation while significantly reducing computational complexity compared to full-image registration.
Solution Approach 2:
Instead of performing exhaustive full-image registration, the method uses partial image patches that provide sufficient motion information for accurate estimation. This partial action approach achieves the necessary measurement precision with much lower computational requirements.
3Measurement precision
If motion estimation is performed for each section independently, then local motion variations are captured accurately, but the overall motion compensation consistency across sections may be compromised
Solution Approach 1:
The motion estimation results from multiple independent section analyses are merged through a consistent mathematical framework. By combining the local motion measurements while enforcing global consistency constraints, the system achieves both accurate local motion capture and consistent global motion compensation across all sections.
Solution Approach 2:
The system uses feedback from the registered patch information to adjust and refine the global motion model. The local motion estimates inform and correct the overall motion compensation, ensuring consistency across sections while maintaining accuracy in capturing local motion variations.
Data Source
AI summary
An improved system and method for estimating and compensating for motion by reducing motion artifacts produced during image reconstruction from helical computed tomography (CT) scan data. In a particular embodiment, the reconstruction may be based on helical partial angle reconstruction (PAR) and the registration may be performed utilizing one or more artificial intelligence (AI) based methods.