MRI Coil Normalization via Reference Image Masking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional coil normalization techniques in MRI face issues such as reduced image contrast, amplified noise in low signal regions, and artifacts in normalized images, particularly in phase sensitive inversion recovery (PSIR) sequences, due to inadequate smoothing and T1-weighting in reference images.
Innovation Solution
The method involves separating the reference image into low- and high-signal regions, applying a strong smoothing function, and iteratively combining masking and smoothing to create a normalization map without T1-contrast, thereby maintaining true infarct-to-myocardium contrast and reducing artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional coil normalization techniques are used, then coil sensitivity correction is achieved, but image contrast is reduced and noise is amplified in low signal regions
Solution Approach 1:
The patent applies different processing strategies to different regions of the image based on signal intensity. High signal regions undergo standard normalization, while low signal regions are identified and protected from excessive noise amplification through region-specific thresholding and adaptive processing. This local differentiation resolves the contradiction by maintaining correction accuracy where safe while protecting vulnerable regions.
Solution Approach 2:
The patent dynamically adjusts normalization parameters based on local signal characteristics. By changing the processing parameters (such as normalization factors and threshold values) according to regional signal intensity, the system achieves accurate coil sensitivity correction in high signal areas while preventing noise amplification in low signal areas, thus resolving the contradiction between correction precision and noise control.
2Object-affected harmful factors
If strong smoothing is applied to reference images, then noise is reduced, but T1-contrast and image detail are lost
Solution Approach 1:
The patent applies smoothing selectively rather than uniformly across the entire image. By identifying regions where smoothing is beneficial versus regions where detail preservation is critical, the system applies appropriate processing strength locally. This resolves the contradiction by reducing noise where appropriate while maintaining T1-contrast and anatomical detail where needed.
Solution Approach 2:
The patent applies partial smoothing - using smoothing strength proportional to local needs rather than uniform excessive smoothing. The degree of smoothing is modulated based on local image characteristics, applying just enough smoothing to reduce noise while stopping before T1-contrast and fine details are lost, thus resolving the contradiction between noise reduction and information preservation.
3Measurement precision
If reference images with T1-weighting are used for normalization, then normalization is achieved, but artifacts are introduced in the normalized image
Solution Approach 1:
The patent extracts and removes the T1-weighting component from the reference image before using it for normalization. By separating the coil sensitivity information from the T1-contrast information, the system uses only the sensitivity map for normalization while discarding the T1-weighting that causes artifacts. This resolves the contradiction by maintaining normalization accuracy while eliminating artifact introduction.
Solution Approach 2:
The patent introduces an intermediary processing step that converts the T1-weighted reference image into a T1-neutral sensitivity map. This intermediary transformation removes the harmful T1-weighting while preserving the useful coil sensitivity information, allowing accurate normalization without artifact introduction, thus resolving the contradiction between normalization accuracy and artifact prevention.
Data Source
AI summary
A method for correcting image inhomogeneity includes acquiring a non-normalized image and a reference image using receiver coils. A high-signal mask and a low-signal mask are created. Each pixel in the high-signal mask is set to a predetermined integer value if the reference image pixel at the same specific location has a value above a threshold value. Each pixel in the low-signal mask is set to the predetermined integer value if the reference image pixel at the same specific location has a value below or equal to the threshold value. A coil normalization map is created by smoothing the reference image with filters. Then, an iterative procedure is performed to update the coil normalization map using the high-signal mask and the low-signal mask. Following the iterative procedure, the non-normalized image is divided by the current coil normalization map to yield a normalized image.