Jove
Visualize
Contact Us
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

Related Experiment Videos

Adaptive Markov modeling for mutual-information-based, unsupervised MRI brain-tissue classification.

Suyash P Awate1, Tolga Tasdizen, Norman Foster

  • 1School of Computing, Scientific Computing and Imaging Institute, University of Utah, 50 South Central Campus Drive, Salt Lake City, UT 84112, USA. suyash@cs.utah.edu

Medical Image Analysis
|August 22, 2006
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

Self-supervised contrastive learning enables robust electrocardiogram-based cardiac classification.

Heart rhythm O2·2026
Same author

Differential profiles of motor dysfunction in amnestic versus non-amnestic mild cognitive impairment - The Vietnam Era Twin Study of Aging.

Neuroscience·2026
Same author

Predicting Ventricular Arrhythmia in Myocardial Ischemia Using Machine Learning.

Computing in cardiology·2026
Same author

Changes in the Neighborhood Built Environment and Chronic Health Conditions in Washington, DC, in 2014-2019: Longitudinal Analysis.

JMIR formative research·2025
Same author

A Survey of Augmentation Techniques for Enhancing ECG Representation Through Self-Supervised Contrastive Learning.

Computing in cardiology·2025
Same author

Machine Learning Prediction of Blood Potassium at Different Time Cutoffs.

Computing in cardiology·2025

This study introduces an adaptive statistical model for brain tissue classification in magnetic resonance (MR) images. The novel method offers automatic, accurate tissue segmentation by analyzing image neighborhoods.

Area of Science:

  • Medical Imaging
  • Computational Neuroscience
  • Biostatistics

Background:

  • Accurate brain tissue classification is crucial for neurological disorder diagnosis and treatment planning.
  • Existing magnetic resonance (MR) image analysis methods often struggle with intensity nonuniformity and require manual parameter tuning.
  • Developing robust and automated methods for MR image segmentation remains a significant challenge in medical image analysis.

Purpose of the Study:

  • To present a novel, adaptive statistical method for automated brain tissue classification in MR images.
  • To address limitations of current methods, including intensity nonuniformity and the need for manual parameter selection.
  • To demonstrate the method's adaptability to diverse MR image types and acquisition artifacts.

Main Methods:

Related Experiment Videos

  • Utilizes a general, adaptive statistical model based on nonparametric Markov random fields to represent MR image neighborhoods.
  • Employs a data-driven strategy for nonparametric modeling of Markov statistics.
  • Minimizes an information-theoretic metric on probability density functions for optimal classification, with automatic parameter tuning.
  • Incorporates atlas-based initialization for complete automation.

Main Results:

  • The proposed method effectively classifies brain tissues in MR images, adapting to various image types and artifacts.
  • It implicitly corrects for intensity nonuniformity, performing well on T1-weighted data without explicit correction.
  • Experiments on real, simulated, and multimodal data show superior performance compared to state-of-the-art techniques.
  • The method demonstrates robustness and accuracy in automated brain tissue segmentation.

Conclusions:

  • The adaptive, nonparametric Markov modeling approach provides a powerful and flexible framework for brain tissue classification.
  • This automated method offers significant advantages over existing techniques, improving accuracy and reducing manual intervention.
  • The findings have implications for advancing quantitative analysis in neuroimaging research and clinical applications.