Brain state classification of rapid event-related fMRI using mixed models

LaConte, S.; Kadah, Y.; Hu, X.

Abstract:

We demonstrate a mixed model approach to account for two sources of variation in ER-data: between HRF variation from a voxel’s relative sensitivity to different stimulus types and within HRF variation to explain the heterogeneity of a voxel’s response to several repetitions of the same stimulus. Beyond the general utility of this approach for obtaining better HRF estimates than those obtained by least-squares, we demonstrate that it improves the epoch-by-epoch representations for vector-based classification algorithms. We compare time-locked and mixed model representations using simulations, and demonstrate that mixed models improve classification accuracy as measured by support vector classification (SVC).

Reference:

abstract S. LaConte, Y. Kadah, X. Hu. Brain state classification of rapid event-related fMRI using mixed models. In Proceedings 15th Scientific Meeting, International Society for Magnetic Resonance in Medicine, Berlin, page 3466, 2007. [bibtex]

Bibtex Entry:

@inproceedings{Berlin3466,
   Author = {LaConte, S. and Kadah, Y. and Hu, X.},
   Title ={Brain state classification of rapid event-related f{M}{R}{I} using mixed models},
   BookTitle = {Proceedings 15th Scientific Meeting, International Society for Magnetic Resonance in Medicine, Berlin},
   Pages = {3466},
   Abstract = {We demonstrate a mixed model approach to account for two sources of variation in ER-data: between HRF variation from a voxel’s relative sensitivity to different stimulus types and within HRF variation to explain the heterogeneity of a voxel’s response to several repetitions of the same stimulus. Beyond the general utility of this approach for obtaining better HRF estimates than those obtained by least-squares, we demonstrate that it improves the epoch-by-epoch representations for vector-based classification algorithms. We compare time-locked and mixed model representations using simulations, and demonstrate that mixed models improve classification accuracy as measured by support vector classification (SVC).},
 Keywords = {Berlin3466},
   Year = {2007} }