MRI Bias Correction via Symmetric Two-Image Differential Registration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for correcting intensity bias in MRI images are inadequate, as they often remove true intensity variations along with artifacts, leading to inconsistent atrophy estimation depending on which image is bias-corrected, and fail to simultaneously correct both images during registration.
Innovation Solution
A B-spline free-form deformation based two-image differential bias correction method that simultaneously corrects both images for bias during registration, ensuring symmetry in bias correction by enforcing that the average bias correction applied to each image is equal and opposite, thereby reducing dependency on the choice of image for bias modeling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If bias correction is applied to only one image during registration, then the processing complexity is reduced, but the atrophy measurement precision becomes dependent on which image is corrected, leading to inconsistent results
Solution Approach 1:
The patent applies asymmetric bias correction where the moving image is corrected using a bias field estimated from the fixed image, creating a deliberate asymmetry in the correction process. This resolves the contradiction by establishing a consistent protocol that eliminates dependency on arbitrary image selection while maintaining computational efficiency.
Solution Approach 2:
The patent performs preliminary bias field estimation from the fixed image before registration, and then applies this pre-computed bias correction to the moving image during registration. This preliminary action separates the bias estimation step from the registration step, improving both efficiency and measurement consistency.
2Object-generated harmful factors
If differential bias correction is applied after rigid registration, then some bias removal is achieved, but true intensity variation is also removed, reducing measurement accuracy
Solution Approach 1:
The patent applies local differential bias correction by estimating the bias field locally from the fixed image and applying it specifically to the moving image in corresponding regions. This localized approach preserves true intensity variations while removing bias artifacts, as the correction is adapted to local image characteristics rather than applied globally.
Solution Approach 2:
The patent implements dynamic bias correction where the bias field estimation and application are integrated into the non-rigid registration process, allowing the correction to adapt dynamically as the images are deformed and aligned. This dynamic approach ensures that bias correction remains accurate throughout the registration process without removing true anatomical variations.
3Stability of the object's composition
If nonparametric non-uniform intensity normalization is used, then image uniformity is improved, but the method does not account for differential bias between images, reducing atrophy measurement accuracy
Solution Approach 1:
The patent segments the bias correction process into two distinct components: (1) nonparametric non-uniform intensity normalization applied to each image individually to improve uniformity, and (2) differential bias correction applied to the moving image based on the fixed image to account for between-image bias differences. This segmentation allows both image uniformity and atrophy measurement accuracy to be achieved simultaneously.
Solution Approach 2:
The patent merges the advantages of nonparametric normalization (image uniformity) with differential bias correction (accurate atrophy measurement) by applying both methods in sequence. The combination of these two correction approaches resolves the contradiction by achieving both uniformity and measurement accuracy.
Data Source
AI summary
Computer analysis of data sets representing images such as MRI images to achieve bias correction and image registration, each image including a bias in intensity within the image of unknown magnitude, is performed by: a) inputting a digital data set of a first image and a digital data set of a second image into a computer; b) calculating a deformation of said first image that transforms said first image into a transformed image that is an optimized approximation of said second image and c) simultaneously calculating and applying a bias correction which is applied to said first image and a bias correction which is applied to said transformed image such that each of the first image and the transformed image is individually corrected for bias therein. Generally, an average of the bias correction over the first image is equal and opposite to an average of the bias correction over said transformed image.