Related Experiment Video
Updated: Mar 1, 2026
09:30
Quantitative Magnetic Resonance Imaging of Skeletal Muscle Disease
Published on: December 18, 2016
20.2K
A non-iterative multi-scale approach for intensity inhomogeneity correction in MRI
Maryjo M George1, S Kalaivani1, M S Sudhakar1
1School of Electronics Engineering, VIT University,Vellore Institute of Technology, Vellore 632014, Tamil Nadu, India.
Magnetic Resonance Imaging
|May 28, 2017
Summary
This study introduces a novel, non-iterative method for correcting magnetic resonance (MR) image intensity inhomogeneity. The new approach effectively removes bias fields without prior segmentation or scanner knowledge, improving image quality for various applications.
Area of Science:
- Medical Imaging
- Image Processing
- Computational Neuroscience
Background:
- Intensity inhomogeneity is a significant challenge in Magnetic Resonance (MR) image analysis, impacting automated tasks like segmentation and registration.
- This artifact is influenced by scanner hardware and patient anatomy, necessitating robust correction methods.
- Existing methods often require segmentation or prior knowledge, limiting their general applicability.
Purpose of the Study:
- To develop a novel, non-iterative algorithm for retrospective correction of MR image intensity inhomogeneity.
- To address the limitations of current methods by eliminating the need for segmentation or prior information about the scanner or subject.
- To enhance the quality of MR images for improved downstream processing and analysis.
Main Methods:
- A multi-scale approach utilizing a Log-Gabor filter bank to extract bias fields at different scales.
- Smoothing operations followed by combining extracted fields to fit a third-degree polynomial for bias field estimation.
- Pixel-wise division of the original image by the estimated bias field to obtain the corrected image.
Main Results:
- The proposed method demonstrated superior performance compared to the state-of-the-art N4 algorithm on simulated (BrainWeb) and real (HCP) datasets.
- High correlation coefficients were observed between the extracted and ground truth bias fields across various MR modalities (T1w, T2w, PD).
- Significant reductions in coefficient of variation and coefficient of joint variation ratios were achieved, indicating improved inter-class separation and reduced intra-class variations in white and grey matter.
Conclusions:
- The developed non-iterative, multi-scale method effectively corrects intensity inhomogeneity in MR images.
- The algorithm offers a robust and generalizable solution, outperforming existing methods like N4.
- The improvements in image quality facilitate better tissue segmentation and analysis in neuroimaging studies.
