Real-Time Prediction of Brain States Using fMRI

LaConte, S.; Peltier, S.; Hu, X.

Abstract:

We demonstrate real-time predictive modeling of FMRI using support vector machine (SVM) classifiers. Distinguishing characteristics of this work are that i) we are using multi-slice data rather than regions of interest, ii) feature selection relies only on threshold-based brain masking, iii) classifier output is related to predicted brain state rather than detected activation and is obtained at each individual image time point, and, iv) the computational burden of training and testing SVM is not prohibitive for multi-slice studies. In a single subject, we illustrate prediction performance with and without feedback.

Reference:

abstract S. LaConte, S. Peltier, X. Hu. Real-Time Prediction of Brain States Using fMRI. In Proceedings 12th Scientific Meeting, International Society for Magnetic Resonance in Medicine, Kyoto, page 2551, 2004. [bibtex]

Bibtex Entry:

@inproceedings{Kyoto2551,
   Author = {LaConte, S. and Peltier, S. and Hu, X.},
   Title ={Real-Time Prediction of Brain States Using {f}{M}{R}{I}},
   BookTitle = {Proceedings 12th Scientific Meeting, International Society for Magnetic Resonance in Medicine, Kyoto},
   Pages = {2551},
   Abstract = {We demonstrate real-time predictive modeling of FMRI using support vector machine (SVM) classifiers. Distinguishing characteristics of this work are that i) we are using multi-slice data rather than regions of interest, ii) feature selection relies only on threshold-based brain masking, iii) classifier output is related to predicted brain state rather than detected activation and is obtained at each individual image time point, and, iv) the computational burden of training and testing SVM is not prohibitive for multi-slice studies. In a single subject, we illustrate prediction performance with and without feedback.},
 Keywords = {Kyoto2551},
   Year = {2004} }