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

fMRI pattern classification using neuroanatomically constrained boosting.

Manel Martínez-Ramón1, Vladimir Koltchinskii, Gregory L Heileman

  • 1Department of Electrical and Computer Engineering, University of New Mexico, NM 87131, USA. manel@ieee.org

Neuroimage
|March 15, 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

Probabilistic day-ahead forecasting of system-level renewable energy and electricity demand.

Nature communications·2026
Same author

Corrigendum: Real-time fMRI using multi-band echo-volumar imaging with millimeter spatial resolution and sub-second temporal resolution at 3 tesla.

Frontiers in neuroscience·2025
Same author

Real-time fMRI using multi-band echo-volumar imaging with millimeter spatial resolution and sub-second temporal resolution at 3 tesla.

Frontiers in neuroscience·2025
Same author

Sliding window functional connectivity inference with nonstationary autocorrelations and cross-correlations.

bioRxiv : the preprint server for biology·2024
Same author

Diffusion-weighted MR spectroscopy: Consensus, recommendations, and resources from acquisition to modeling.

Magnetic resonance in medicine·2023
Same author

Adaptive Sparse Gaussian Process.

IEEE transactions on neural networks and learning systems·2023

This study introduces a new Adaboost method for functional MRI (fMRI) pattern classification across diverse subjects and acquisition methods. The enhanced technique significantly improves classification accuracy and robustness compared to traditional approaches.

Area of Science:

  • Neuroimaging
  • Machine Learning
  • Cognitive Neuroscience

Background:

  • Functional MRI (fMRI) pattern classification aims to identify neural substrates of cognitive tasks.
  • Challenges include high dimensionality, limited datasets, inter-individual variability, and acquisition method dependence.
  • Previous methods were often limited to individual subjects.

Purpose of the Study:

  • To develop a novel approach for improved multiclass classification across subjects, field strengths, and fMRI methods.
  • To enhance the generalization ability and robustness of fMRI classification.
  • To assess the utility of the developed method for real-time fMRI applications.

Main Methods:

  • Spatially normalized fMRI activation maps were segmented using a neuroanatomical atlas.

Related Experiment Videos

  • Local classifiers were applied to each functional area.
  • A modified Adaboost technique aggregated outputs from region-specific classifiers for weighted multiclass classification.
  • Classification accuracy was assessed in a group study with diverse tasks and acquisition parameters.

    • Main Results:

    • The boosted classifier achieved misclassification rates between 3.5% and 10%.
    • Single classifiers showed misclassification rates between 15% and 23%.
    • The boosted classifier demonstrated superior generalization and robustness.

    Conclusions:

    • The novel Adaboost approach significantly improves multiclass fMRI classification accuracy and robustness across varied conditions.
    • The method's computational speed makes it suitable for real-time fMRI analysis.
    • This technique offers a more reliable way to interpret dynamic brain activation patterns.