B0 Inhomogeneity Mapping Using Deblurred MRI Difference Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Neural networks used for determining B0 inhomogeneity maps in magnetic resonance imaging are susceptible to out-of-distribution errors and can produce erroneous data, leading to misleading medical images.
Innovation Solution
A method using a convolutional neural network to deblur magnetic resonance images with varying demodulating frequencies, followed by fitting a smooth manifold to difference images to determine the B0 inhomogeneity map, reducing errors and inaccuracies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a neural network is used to determine B0 inhomogeneity maps, then the automation and speed of the process is improved, but the reliability and accuracy deteriorate due to out-of-distribution errors and neural hallucinations
Solution Approach 1:
The patent introduces difference images as an intermediary between the neural network output and the final B0 map. The neural network generates preliminary B0 estimates, which are then processed through difference image calculation and smooth manifold fitting to eliminate errors and hallucinations, producing a reliable final B0 inhomogeneity map
Solution Approach 2:
The patent implements a feedback mechanism where the neural network's preliminary B0 estimates are evaluated through difference image analysis. The smooth manifold fitting process uses this feedback to correct deviations and hallucinations, iteratively refining the B0 map until convergence to a reliable solution
2Manufacturing precision
If multiple demodulating frequencies are applied to deblur magnetic resonance images, then the manufacturing precision of the deblurred images is improved, but the device complexity increases
Solution Approach 1:
The patent segments the complex deblurring problem into multiple independent processing streams, each handling a specific demodulating frequency. By processing different frequency components separately and then combining results through difference image analysis, the system achieves high precision while managing complexity through modular organization
3Reliability
If difference images are calculated and smooth manifold fitting is applied, then the reliability of B0 inhomogeneity mapping is improved, but the processing time and computational complexity increase
Solution Approach 1:
The patent applies partial action by using smooth manifold fitting only on the difference images rather than processing all image data. This selective application of computational techniques achieves high reliability by focusing computational resources on the critical error-correcting steps while minimizing overall processing time
Data Source
AI summary
Disclosed herein is a medical system (100, 300) comprising a memory (110) storing machine executable instructions (120) and a convolutional neural network (122) configured for outputting a predetermined number of deblurred magnetic resonance images (126) that are slices of a deblurred magnetic resonance imaging data set in response to receiving a set of partially deblurred magnetic resonance images for each of the slices. The execution of the machine executable instructions causes a computational system (104) to: receive (200) the set of partially deblurred magnetic resonance images; receive (202) the predetermined number of deblurred magnetic resonance images in response to inputting the set of partially deblurred magnetic resonance images for each of the slices into the convolutional neural network; calculate (204) a set of difference images (128) for each of the slices by calculating a difference between the deblurred magnetic resonance image and each of the set of partially deblurred magnetic resonance images; and calculate (206) a determined B0 inhomogeneity map (130) for each of the slices by fitting a smooth manifold to B0 values determined from the set of difference images, the documentation frequency map, and the assigned demodulating frequnecy for each of the set of difference images.