Jove
Visualize
Contact Us

Related Experiment Video

Updated: Jun 16, 2026

Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping
10:25

Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping

Published on: September 25, 2019

Intensity Standardization Simplifies Brain MR Image Segmentation.

Ying Zhuge1, Jayaram K Udupa

  • 1Medical Image Processing Group, Department of Radiology, University of Pennsylvania, Philadelphia, PA 19104.

Computer Vision and Image Understanding : CVIU
|February 18, 2010
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Free-breathing Dynamic MRI Analysis of Three-dimensional Diaphragm Curvatures in Pediatric Patients with Thoracic Insufficiency Syndrome.

Radiology. Cardiothoracic imaging·2026
Same author

Open area segmentation in CT images based on pixel displacement and multi-view with application in the axillary and lower cervical regions.

Medical physics·2025
Same author

Auto-segmentation of thoraco-abdominal organs in pediatric dynamic MRI.

Medical physics·2025
Same author

Predicting the Effort Required to Manually Mend Auto-Segmentations.

IEEE journal of biomedical and health informatics·2025
Same author

Virtual cutaneous area severity index (vCASI): A comprehensive methodology for quantitative skin disease assessment on positron emission tomography images.

European journal of nuclear medicine and molecular imaging·2025
Same author

Do Rib-Based Anchors Impair Chest Wall Motion in Early Onset Scoliosis? - A Novel Investigation via Dynamic MRI.

Journal of pediatric orthopedics·2025
Same journal

Region-based image registration for remote sensing imagery.

Computer vision and image understanding : CVIU·2023
Same journal

Modeling eye movement patterns to characterize perceptual skill in image-based diagnostic reasoning processes.

Computer vision and image understanding : CVIU·2022
Same journal

Entropy-based Correspondence Improvement of Interpolated Skeletal Models.

Computer vision and image understanding : CVIU·2020
Same journal

Shape Matching and Registration by Data-driven EM.

Computer vision and image understanding : CVIU·2017
Same journal

Simultaneous Tumor Segmentation, Image Restoration, and Blur Kernel Estimation in PET Using Multiple Regularizations.

Computer vision and image understanding : CVIU·2017
Same journal

Contour based object detection using part bundles.

Computer vision and image understanding : CVIU·2017
See all related articles
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

This paper introduces a streamlined, non-iterative approach for segmenting brain tissues from magnetic resonance images. By standardizing image intensities before processing, the authors enable a simpler classification strategy that outperforms traditional, more complex methods in both accuracy and computational efficiency.

Area of Science:

  • Medical imaging informatics and intensity standardization within neuroimaging
  • Computational neuroscience and diagnostic image analysis

Background:

Variability in magnetic resonance imaging signals across different hardware platforms often hinders consistent automated analysis. This inconsistency frequently forces researchers to utilize complex, iterative algorithms to achieve reliable tissue classification. Such demanding procedures often require substantial processing power and remain difficult to integrate into standard clinical workflows. That uncertainty drove the development of more efficient preprocessing strategies to normalize signal values. Prior research has shown that raw data from distinct scanners rarely share a common numeric scale. This gap motivated the exploration of techniques that assign tissue-specific meanings to voxel intensities. No prior work had resolved the trade-off between high computational overhead and segmentation precision until now. The current study addresses these limitations by proposing a straightforward, non-iterative framework for brain tissue identification.

Purpose Of The Study:

Keywords:
neuroimaging preprocessingfuzzy connectednesstissue classificationmagnetic resonance imaging

Frequently Asked Questions

The researchers propose a non-iterative strategy utilizing vectorial scale-based fuzzy connectedness and maximum likelihood criteria. This approach classifies voxels by estimating membership values through either multivariate Gaussian models or direct histogram distributions, which contrasts with the computationally intensive expectation-maximization methods used in traditional finite mixture models.

The authors utilize intensity inhomogeneity correction and intensity standardization as preprocessing steps. These techniques ensure that voxel values possess tissue-specific numeric meanings, whereas traditional methods often struggle with signal variations across different scanners without these specific normalization procedures.

Spatial constraints are incorporated into the maximum likelihood criterion to improve classification. This technical necessity ensures that voxel grouping remains consistent with anatomical expectations, unlike simpler classifiers that ignore the spatial relationship between neighboring brain tissues.

More Related Videos

Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
06:48

Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images

Published on: January 7, 2019

Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
14:08

Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images

Published on: April 13, 2013

Related Experiment Videos

Last Updated: Jun 16, 2026

Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping
10:25

Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping

Published on: September 25, 2019

Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
06:48

Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images

Published on: January 7, 2019

Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
14:08

Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images

Published on: April 13, 2013

The aim of this study is to present a simple, non-iterative method for segmenting brain tissues from magnetic resonance images. Researchers sought to address the significant intensity variations that typically occur across different patients and scanners. Such variations often lead to poor classification results when using standard training sets. The team hypothesized that preprocessing techniques could assign tissue-specific numeric meanings to image intensities. This motivation drove the development of a strategy that avoids the high computational costs associated with traditional iterative methods. The authors intended to demonstrate that intensity standardization plays a vital role in simplifying the segmentation process. They also aimed to compare their proposed membership estimation techniques against existing k-nearest neighbors and expectation-maximization models. This work specifically targets the challenge of maintaining segmentation accuracy while improving overall processing efficiency in clinical settings.

Main Methods:

The review approach involves evaluating two distinct membership estimation techniques against established benchmarks. Researchers utilized vectorial scale-based fuzzy connectedness alongside morphological operations to define the intracranial mask. They then estimated fuzzy membership values for each voxel using either a multivariate Gaussian model or direct histogram distributions. The study design incorporated a maximum likelihood criterion that accounts for spatial constraints during the final voxel classification phase. Investigators employed a training data set consisting of inhomogeneity-corrected and standardized images to fix model parameters. Performance metrics were derived from a cohort of 10 healthy subjects and 10 individuals with multiple sclerosis. This comparative analysis focused on assessing both the accuracy and the computational efficiency of the proposed algorithms. The team contrasted these findings with results from k-nearest neighbors and finite mixture model-based expectation-maximization techniques.

Main Results:

Key findings from the literature indicate that the proposed methods achieve greater overall accuracy than the k-nearest neighbors approach. The study demonstrates that these techniques provide significantly better efficiency than finite mixture model-based expectation-maximization methods. For healthy subject data sets, the accuracy of the proposed strategies is similar to that of the expectation-maximization model. However, the new methods show much better performance when applied to data from patients with multiple sclerosis. The gaussian model and histogram-based estimation techniques successfully assign tissue-specific numeric meanings to image intensities. These results confirm that non-iterative processing effectively simplifies the segmentation pipeline. The quantitative comparison highlights a clear advantage in balancing diagnostic precision with reduced computational costs. This evidence supports the utility of intensity normalization as a primary step for robust brain tissue classification.

Conclusions:

The proposed non-iterative framework demonstrates superior performance compared to traditional k-nearest neighbors approaches. These techniques achieve higher accuracy levels while maintaining significant computational efficiency advantages over expectation-maximization models. The authors suggest that standardizing signal values allows for a simplified classification strategy across diverse patient populations. Findings indicate that the gaussian model and histogram-based methods perform comparably to complex expectation-maximization models for healthy subjects. Notably, the new strategies provide improved results for patients diagnosed with multiple sclerosis. This synthesis implies that preprocessing steps are vital for ensuring consistent tissue-specific numeric representations in clinical data. The study confirms that removing intensity variations simplifies the subsequent segmentation process without sacrificing diagnostic quality. These results provide a robust alternative for researchers seeking to balance processing speed with segmentation accuracy in neuroimaging.

The training data set consists of inhomogeneity-corrected and intensity-standardized images. This data type allows the model to fix mean intensity vectors and covariance matrices, which is essential for the Gaussian-based membership estimation compared to raw, uncorrected image inputs.

The researchers measured performance using 10 normal subject scans and 10 multiple sclerosis patient scans. This measurement demonstrates that the proposed methods achieve higher accuracy than k-nearest neighbors and superior efficiency compared to finite mixture models, particularly in pathological cases.

The authors claim that their approach provides a more efficient alternative to expectation-maximization models. They propose that this simplicity makes the method easier to implement for clinical applications, especially when dealing with the signal variations inherent in multiple sclerosis patient data.