Iterative Planar Distribution Estimation for Low-Dose Spectral Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques for spectral imaging struggle to effectively reduce noise, particularly system noise and quantum noise, which can result in unclear images when the radiation dose is low, especially in the estimation of planar distributions for substances.
Innovation Solution
A planar distribution obtaining unit that uses multiple radiation images captured at different energies, employing likelihood and prior probability calculations to maximize the probability of the output image, thereby reducing noise and improving image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If simple arithmetic operations are used for estimation, then the process is simple and fast, but noise cannot be completely removed and image quality deteriorates at low radiation doses
Solution Approach 1:
The patent changes the processing parameters from simple arithmetic operations to iterative calculations using likelihood and prior probability. This transforms the estimation process into a statistical optimization problem that can effectively suppress noise while maintaining computational feasibility through iterative approximation.
Solution Approach 2:
The patent replaces the mechanical arithmetic operation system with a probabilistic calculation system. By substituting deterministic arithmetic with statistical methods (maximum likelihood estimation and Bayesian inference), the system achieves superior noise reduction capability while maintaining processing efficiency.
2Object-affected harmful factors
If radiation dose is reduced, then patient exposure is minimized, but quantum noise increases and image quality deteriorates
Solution Approach 1:
The patent converts the harmful quantum noise into a manageable statistical variable. By modeling noise as a probabilistic phenomenon and applying maximum likelihood estimation, the system transforms the adverse effect of quantum noise into a solvable optimization problem, enabling high-quality imaging at low radiation doses.
Solution Approach 2:
The patent introduces probabilistic models and iterative calculation algorithms as intermediaries between the raw noisy data and the final image. These intermediate processing steps act as mediators that filter and refine the information, separating signal from noise while preserving diagnostic quality.
3Measurement precision
If iterative calculation with likelihood and prior probability is used, then noise reduction and image quality improve, but computational complexity increases
Solution Approach 1:
The patent employs dynamic iterative calculation where the solution evolves progressively through multiple iterations. Each iteration refines the estimate by incorporating both likelihood information from the data and prior probability constraints, allowing the system to adaptively converge to the optimal solution rather than requiring a single complex calculation.
Solution Approach 2:
The patent applies preliminary action by incorporating prior probability information before the actual estimation process. This pre-processing step provides initial constraints and guidance that streamline the subsequent iterative optimization, reducing the computational burden by narrowing the search space from the outset.
Data Source
AI summary
An image processing apparatus includes a planar distribution obtaining unit configured, using a plurality of radiation images captured using different radiation energies as input images, to obtain, by a sequential approximation solution by an iterative calculation, an output image representing a planar distribution for a substance contained in the input image. Based on a likelihood and a prior probability using one of a pixel value of a pixel of interest or a pixel value of a peripheral pixel as an input, the planar distribution obtaining unit decides the output image such that a probability that an output is a value of the planar distribution when an input is the input image that is maximized.