Related Experiment Video
Updated: Jun 5, 2026
08:51
Magnetic Resonance Imaging of Multiple Sclerosis at 7.0 Tesla
Published on: February 19, 2021
Evaluating intensity normalization on MRIs of human brain with multiple sclerosis.
Mohak Shah1, Yiming Xiao, Nagesh Subbanna
1Centre for Intelligent Machines, McGill University, Montreal, Canada. mohak@cim.mcgill.ca
Medical Image Analysis
|January 15, 2011
Summary
This study validates Nyul
Area of Science:
- Medical Imaging
- Neuroimaging
- Computer Vision
Background:
- Intensity normalization is crucial for Magnetic Resonance Image (MRI) analysis.
- Supervised segmentation and classification methods rely on standardized intensity ranges.
- Real-world clinical data presents challenges like multi-site variations and disease progression.
Purpose of the Study:
- To extensively validate Nyul's intensity normalization approach in a clinical setting.
- To assess normalization effectiveness in homogenizing intensities and improving tissue separation.
- To compare Nyul's method against linear normalization for Multiple Sclerosis (MS) lesion segmentation.
Main Methods:
- Evaluation using distributional divergence criteria.
- Comparison of decile-based piecewise linear normalization with linear normalization.
- Testing across various image segmentation algorithms (Bayesian, outlier detection, MRF-based).
Main Results:
- Nyul's method demonstrated superior intensity homogenization and tissue separation.
- The decile-based approach significantly improved MS lesion segmentation accuracy.
- Normalization effectiveness was consistent across segmentation algorithms of varying complexity.
Conclusions:
- Nyul's intensity normalization is effective for heterogeneous clinical MRI data.
- This method enhances segmentation performance, particularly for MS lesions.
- The approach is robust and independent of segmentation algorithm complexity.
