Pixon Map Construction for Low-Count Nuclear Image Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image reconstruction methods in nuclear imaging face challenges in accurately processing low-count data with noise, particularly in medical imaging techniques like PET and SPECT, where the quality of reconstructed images is compromised due to limited radiation detection.
Innovation Solution
The implementation of a pixon method using a Mighell-like statistical weight for evaluating contributions of pixon kernel functions to create a pixon map, which involves smoothing objects with selected kernel functions and assigning values based on statistical objects, enabling high-quality reconstruction of 3D images from low-count data sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional image reconstruction methods are used on low-count data, then the reconstruction process is simple and fast, but the image quality and accuracy are compromised due to noise
Solution Approach 1:
The patent applies local quality by making the smoothing strength spatially adaptive through the pixon map. Each voxel in the image space is assigned a specific pixon kernel function size based on local statistical properties (standard deviation and mean count values). This allows regions with high noise (low counts) to receive stronger smoothing while preserving edges and features, whereas regions with sufficient counts maintain higher resolution. The pixon map construction unit calculates local statistical weights and assigns kernel functions accordingly, resolving the contradiction between noise reduction and image fidelity.
Solution Approach 2:
The patent implements dynamics by making the reconstruction process adaptive rather than static. The pixon map is constructed dynamically based on the actual data statistics, and the pixon smoothing operation adjusts the degree of smoothing for each voxel based on locally determined parameters. This dynamic adaptation allows the reconstruction method to optimize image quality for low-count data while maintaining reasonable computational complexity through efficient algorithms.
2Object-affected harmful factors
If stronger smoothing is applied to reduce noise in low-count data, then noise is reduced, but image resolution and detail are lost
Solution Approach 1:
The patent resolves this contradiction by applying different smoothing strengths to different regions of the image. The pixon map assigns specific kernel function sizes to each voxel based on local data characteristics. In regions with very low counts where noise is problematic, larger kernels provide stronger smoothing. In regions with sufficient counts or near edges, smaller kernels preserve resolution. This local adaptation eliminates the need for uniform strong smoothing across the entire image.
Solution Approach 2:
The patent changes the parameter of kernel function size dynamically based on local statistical properties. The pixon map stores and applies varying kernel sizes (parameter changes) rather than using a fixed kernel size for the entire image. This parameter adaptation allows the smoothing operation to reduce noise where necessary while preserving resolution where possible, directly resolving the contradiction between noise reduction and detail preservation.
3Measurement precision
If adaptive pixon smoothing is applied to improve image quality, then noise is reduced and accuracy improves, but computational complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the image reconstruction process into distinct stages: (1) initial reconstruction, (2) pixon map construction based on statistical analysis, and (3) pixon smoothing operation. The pixon map itself segments the image space into regions requiring different smoothing treatments. This segmentation allows the complex adaptive operation to be broken down into manageable steps, improving computational efficiency while maintaining accuracy.
Solution Approach 2:
The patent implements preliminary action by pre-calculating the pixon map before applying the smoothing operation. The pixon map construction unit computes the statistical properties and assigns kernel functions to all voxels in advance. This preliminary preparation stores the adaptive parameters in a lookup table (the pixon map), which then guides the smoothing operation without requiring complex real-time calculations during reconstruction. This approach significantly reduces the computational burden of the adaptive smoothing process.
Data Source
AI summary
Determining a pixon map for pixon smoothing of an object based on a data set includes receiving the data set and an input object associated to the data set. Determining a pixon map further includes determining, in a series of steps, statistical objects for a set of pixon kernel functions, wherein each step includes selecting a pixon kernel function from the set of pixon kernel functions, smoothing the input object on the basis of the selected pixon kernel function, thereby creating a smoothed object, and determining the statistical object for the selected pixon kernel function on the basis of the smoothed object, the data set, and a Mighell-like statistical weight. Determining a pixon map further includes determining contributions of the pixon kernel functions to the pixon map based on the statistical objects and assigning values to the pixon map corresponding to the contributions of the pixon kernel functions.