A0978
Title: Incorporating seasonality in fMRI time series to address the learning effect
Authors: Israel Almodovar Rivera - University of Puerto Rico (United States) [presenting]
Abstract: Functional magnetic resonance imaging (fMRI) is a noninvasive tool for studying regions related to some particular tasks. These activated regions are identified by assigning a map. Reliability across subjects using the activation maps is essential to this analysis. These maps are typically constructed by assigning a general linear model with an autoregressive structure, usually an AR(1) model. However, fMRI experiments naturally have a seasonal structure since the stimulus is applied in each period. A seasonal time series for an fMRI experiment is considered, where the time application of the stimulus is applied. Then, activation maps are obtained from these time series models using classical and adaptive smoothing and thresholding approaches. Reliability was assessed using the Jaccard similarity coefficient as a modified percent activation overlap. The methodology is illustrated in several real datasets from studies involving multiple subjects. It was found that activation maps obtained using seasonal models tend to have more similarities than maps where seasonality was not considered.
