Hybrid Image Reconstruction Using Compressed Sensing for Stability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep learning-based tomographic image reconstruction methods suffer from instability, such as strong output artifacts and failure to detect small features, especially in sparse-view scenarios, necessitating a stabilization approach.
Innovation Solution
A hybrid image reconstruction system combining deep learning and compressed sensing stages, utilizing an analytic compressive iterative deep framework (ACID) or dual-domain residual-based optimization network (DRONE), which includes iterative refinement stages to stabilize and enhance image reconstruction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning-based reconstruction is used, then image quality is improved, but stability deteriorates due to strong output artifacts from tiny perturbations
Solution Approach 1:
The patent combines deep learning-based reconstruction with sparsity-regularized reconstruction methods into a hybrid framework. The deep learning component captures complex image priors and patterns for high-quality reconstruction, while the sparsity-regularized component provides mathematical constraints and stability through kernel awareness. This merging allows the system to achieve both high image quality and reconstruction stability by leveraging the complementary strengths of both approaches.
Solution Approach 2:
The reconstruction system uses a composite approach by integrating two different reconstruction methodologies (deep learning and sparsity-regularized methods) into a unified hybrid framework. This composite structure allows the system to benefit from the pattern recognition capabilities of deep learning while simultaneously incorporating the mathematical robustness and stability of sparsity-regularized methods, creating a reconstruction system that is both high-quality and stable.
2Measurement precision
If deep learning reconstruction is used, then image quality is improved, but small features become undetectable due to instability
Solution Approach 1:
The hybrid reconstruction framework merges deep learning with sparsity-regularized methods to preserve small features. The sparsity-regularized component's kernel awareness provides mathematical constraints that prevent the suppression of small features, while the deep learning component enhances overall image quality. This combination ensures that small features remain detectable while maintaining high image quality.
Solution Approach 2:
The system implements feedback mechanisms where the reconstruction output is iteratively refined. The sparsity-regularized component provides feedback constraints that guide the deep learning reconstruction to preserve small features, while the deep learning component provides quality enhancement feedback. This iterative feedback loop ensures that small features are maintained throughout the reconstruction process.
3Measurement precision
If deep learning reconstruction is used, then image quality is improved, but performance degrades with increased input data due to instability
Solution Approach 1:
The patent merges deep learning reconstruction with sparsity-regularized reconstruction to handle increased input data reliably. The sparsity-regularized component provides mathematical stability and kernel awareness that prevent performance degradation, while the deep learning component maintains high image quality. This hybrid approach ensures consistent performance even as input data volume increases.
Solution Approach 2:
The system dynamically adjusts reconstruction parameters based on input data characteristics. The sparsity-regularized component modifies regularization parameters to maintain stability with increased data, while the deep learning component adapts its processing to preserve image quality. This parameter adaptation allows the system to maintain both quality and reliability across varying input data conditions.
Data Source
AI summary
Generally, there is provided a hybrid image reconstruction system. The hybrid image reconstruction system includes a deep learning stage and a compressed sensing stage. The deep learning stage is configured to receive an input data set that includes measured tomographic data and to produce a deep learning stage output. The deep learning stage includes a mapping circuitry, and at least one artificial neural network. The mapping circuitry is configured to map image domain data to a tomographic data domain. The compressed sensing stage is configured to receive the deep learning stage output and to provide refined image data as output.