MRI Image Optimization Using Voxel Clustering Centroids
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current magnetic resonance imaging (MRI) techniques face challenges in optimizing images quickly within time-bound clinical situations, particularly due to the high number of voxels required for accurate B1+ field mapping, which can lead to computational inefficiencies and reduced image quality when using methods like grid decimation, voxel aggregation, or randomization.
Innovation Solution
A computer-implemented method that maps images into identically sized three-dimensional voxels, represents them in a Euclidean n-dimensional space, clusters voxels with similar characteristics to create homogenous clusters, determines the center or centroid of each cluster, and uses these clusters as super-voxels to generate an optimized diagnostic image, thereby adapting voxel size to the magnetic field distribution and improving image homogeneity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the number of voxels is reduced for faster processing, then computational time is reduced, but image quality and mapping precision deteriorate
Solution Approach 1:
The patent segments the large set of voxels into multiple clusters based on spatial proximity and B1+ field similarity. Each cluster is represented by a centroid voxel, creating a hierarchical structure that reduces the number of voxels processed while preserving the essential spatial and field characteristics of the original data set
Solution Approach 2:
The patent creates representative copies (centroids) of voxel groups that capture the essential characteristics of each cluster. These centroid voxels serve as simplified representations that maintain the B1+ field mapping accuracy while significantly reducing computational complexity
2Loss of time
If traditional voxel aggregation or grid decimation is used to reduce voxel count, then processing time is reduced, but image homogeneity and information integrity deteriorate
Solution Approach 1:
The patent uses the B1+ field values and spatial positions of all original voxels as feedback to determine the optimal clustering configuration. This feedback mechanism ensures that the clustering process preserves the essential field characteristics and spatial relationships, preventing information loss during the reduction process
Solution Approach 2:
The patent changes the parameter representation from individual voxel values to cluster centroid values that are derived from multiple original voxels. This parameter transformation maintains the essential information while reducing the data dimensionality and processing requirements
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach allows for rapid optimization of MRI images with minimal impact on quality, reducing the number of voxels from 10,000 to 100, enhancing image homogeneity and reducing computational time, while maintaining the integrity of the original image information.
Implementation Method 1
clustering the voxel space by grouping together voxels having similar characteristics to create homogenous clusters; determining the centre or centroid of each cluster
Implementation Method 2
One method of dealing with such B1+ magnetic field inhomogeneities is to use an RF shimming approach. This is where rather than a single transmit channel Ch for all of the coil elements, at least two (NCh≥2) are provided. This allows the same waveform p(t) to be transmitted on each channel... But enables the channels to be transmitting signals having amplitudes or phases ωc
Implementation Method 3
Over a number of parallel transmit channels it may also be desirable to control the RF signal from pulse to pulse dynamically, in order to improve image quality still further. This is known as Dynamic Signal Control, or DSC.
Data Source
AI summary
A computer-implemented magnetic resonance image optimisation method is disclosed. An image of an object is mapped using a static magnetic field and divided into a plurality of voxels. Each voxel is represented in a Euclidean n-dimensional space, where n≥3, and clustered by grouping together voxels having similar characteristics to create homogenous clusters. The centre or centroid of each cluster is determined, and used, or the voxel closest to either the centre or the centroid is used, as a super-voxel in an optimisation procedure. An optimised diagnostic image of the object is then generated.