Vadim Zotev1, Raquel Phillips1,
Ruben Alvarez1, W. Kyle Simmons1, Pat Bellgowan1,
1Laureate Institute for
The support vector machines (SVM) approach to decode patterns of whole-brain activity can be utilized for real-time fMRI neurofeedback. We propose a combined SVM and region-of-interest (ROI) neurofeedback approach. A custom rtfMRI system was used to compute an SVM classifier and measure fMRI activation in the left amygdala ROI, and to provide neurofeedback as a variable-height bar. Three healthy male subjects were asked to feel happy so as to raise the bar. We observed that ROI-based neurofeedback improves SVM performance, while SVM-based neurofeedback increases ROI activation. Combination of these two approaches benefits both and enhances rtfMRI neurofeedback training.