Related Experiment Videos
Are fMRI event-related response constant in time? A model selection answer.
Sophie Donnet1, Marc Lavielle, Jean-Baptiste Poline
1Laboratoire de mathématiques, Université de Paris Sud, 91405 Orsay, France.
Neuroimage
|May 2, 2006
Summary
This study introduces a flexible model for estimating the hemodynamic response function (HRF) in functional magnetic resonance imaging (fMRI), allowing response magnitude to vary over time. The findings suggest this dynamic model offers improved accuracy for neuroimaging analyses compared to traditional fixed-magnitude approaches.
Area of Science:
- Neuroimaging
- Cognitive Neuroscience
- Biophysics
Background:
- Accurate hemodynamic response function (HRF) estimation is vital for interpreting functional magnetic resonance imaging (fMRI) data.
- Existing models often assume a constant HRF magnitude, which may not reflect biological reality where attention and activity influence responses.
Purpose of the Study:
- To develop and test a more flexible statistical model for HRF estimation in fMRI that accounts for time-varying response magnitudes.
- To compare the performance of this flexible model against a conventional fixed-magnitude model using information theory.
Main Methods:
- A maximum likelihood framework was employed to model the HRF with potentially varying magnitudes.
- An expectation-maximization (EM) algorithm was developed to estimate event magnitudes and the HRF.
- The models were evaluated on fMRI data from 32 regions of interest across eight subjects.
Main Results:
- The flexible model, allowing for time-varying HRF magnitudes, demonstrated superior performance compared to the fixed-magnitude model in the majority of tested regions.
- Information theory metrics indicated a better fit for the dynamic model.
Conclusions:
- The findings support the use of more flexible, time-varying HRF models in event-related fMRI studies.
- This approach offers enhanced precision for spatial and temporal estimation of neuronal processes, improving the analysis of neuroimaging data.