Deconvolution of fMRI BOLD signal in time-domain using an exponential operator and Lasso optimization
Résumé
Many techniques have been explored so far in the study of neural activations using the blood oxygenated level dependent (BOLD) signal. Among them, deconvolution methods have been developed in order to explore spontaneous brain activity when the brain is in resting-state. These techniques are powerful since they do not require a priori knowledge about timing and duration of activations [2]. In this work, we propose a regularized deconvolution technique which uses an exponential operator, whose shape and performance can be adjusted by tuning a parameter α, and the Least-Angle Regression (LARS) algorithm, by using the least absolute shrinkage and selection operator (LASSO) model.