Motion Compensated X-Ray Image Reconstruction Using Latent Vector Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing X-ray-based imaging methods, such as cone beam computed tomography (CBCT), face challenges in achieving high-quality image reconstruction due to patient movement, which results in artifacts like stripes or blurring.
Innovation Solution
The method involves parameterizing the patient's movement trajectory in an abstract latent space using latent vectors and employing a trained algorithm to generate physically realistic movement trajectories, thereby preventing optimization from getting stuck in local minima.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If motion compensation is performed using traditional parameterization methods (rotations and translations for each projection image), then motion compensation can be achieved, but the computational complexity becomes very high due to the large number of parameters required
Solution Approach 1:
The patent transforms the motion parameterization from traditional rigid body parameters (rotations and translations for each projection image) to a reduced set of latent vector parameters. This parameter transformation significantly reduces the dimensionality of the optimization problem while maintaining the ability to model patient motion, thereby reducing computational complexity without sacrificing motion compensation reliability
Solution Approach 2:
The patent introduces a latent vector as an intermediary representation that bridges the complex motion trajectories and the optimization process. The latent vector serves as a compressed representation of motion that can be efficiently optimized, and from which the actual motion parameters can be derived, thus mediating between the complexity of real motion and the needs of computational optimization
2Adaptability or versatility
If the cost function is non-convex to capture complex motion trajectories, then motion compensation can account for various motion patterns, but the optimization may get stuck in local minima leading to unreliable results
Solution Approach 1:
The patent changes the parameterization approach by using latent vectors that are optimized through a simplified objective function. This parameter transformation allows the optimization to navigate the complex motion space more effectively, reducing the likelihood of getting stuck in local minima while still capturing diverse motion patterns through the latent space representation
Solution Approach 2:
The patent replaces the traditional mechanical optimization approach (direct optimization of motion parameters with complex cost functions) with a latent space-based optimization approach. This substitution simplifies the optimization landscape by working in the latent vector space rather than directly in the motion parameter space, making the optimization more reliable while maintaining adaptability to various motion trajectories
3Manufacturing precision
If traditional optimization methods are used to solve the motion compensation problem, then the full motion parameters can be optimized, but the computational intensity becomes very high for a large number of projection images
Solution Approach 1:
The patent fundamentally changes the number of parameters to be optimized by transforming from per-projection-image parameters to a reduced latent vector representation. This parameter reduction dramatically decreases the computational power requirements while maintaining the precision of image reconstruction through the latent space optimization approach
Data Source
Figure 1~2
Figure 3~4
AI summary
To generate a motion-compensated image reconstruction (14) in an X-ray-based imaging procedure, an optimal object motion trajectory (10') is determined by applying a trained algorithm (12) to an optimal latent vector, and the motion-compensated image reconstruction (14) is generated depending on projection images under the assumption that the object (6) moved according to the optimal object motion trajectory (10') during the execution of the imaging procedure.The optimal latent vector is determined by minimizing a given cost function using the latent vector as an optimization parameter, wherein for each optimization step with a present latent vector, a present object motion trajectory is determined by applying the trained algorithm (12) to the present latent vector, and a present value of the cost function is calculated depending on the multitude of projection images under the assumption that the object (6) has moved according to the present object motion trajectory during the execution of the imaging procedure.