Representational Similarity Analysis (RSA)
Representational similarity analysis (RSA) is used to analyze the response similarity between evoked fMRI responses in selected regions-of-interest (ROIs). For each ROI a representational distance (or dissimilarity) matrix (RDM) is computed and graphically displayed containing distance measures (usually 1-correlation) between pairs of distributed activity patterns representing different experimental conditions. The distance information stored in an RDM is often visualized using multi-dimensional scaling (MDS) plots; MDS attempts to place each condition at a location in a 2-dimensional space that maximally satisfies the pairwise distances to all other conditions. Note that the calculated similarity structure between conditions can itself be related to RDMs from other ROIs producing second-level RDMs that help to understand the specific representational principles of a brain region by revealing what distinctions between stimuli are emphasized and what distinctions are de-emphasized in a specific ROI (or at a certain level of a computational model). Since the comparison of first-level RDMs does not require voxel-level correspondence, data from other sources can be easily integrated in second-level analyses, including integration of RDMs from multiple subjects, other measurement modalities and computational models.
BrainVoyager provides two dialogs for first and second-level analyses, respectively. The Representational Similarity Analysis dialog (evoked from the Analysis menu) is used to caclualte RDMs from fMRI data that needs to be provided as a set of condition-specific volume maps (VMP file). In addition to the volume maps, a set of ROIs need to be provided (VOI file). For each provided ROI, one RDM will be calculated using the volume map data for the voxels specified in the ROI. The Second-Level RSA dialog (evoked from the Analysis menu) is used to relate RDMs from the first level with each other. Since the second-level analysis requires as input only representational distance (dissimilarity) matrices (simple text files), the data may come not only from other subjects but also from other modalities or from computational models. The comparison of brain-derived RDMs with those from computational models is a common motivation for second-level RSA analysis (e.g. Kriegeskorte et al., 2008; Nili et al., 2014). For further details on how to use the two provided RSA dialogs, use the links below:
- First-level RSA using the Representational Similarity Analysis dialog.
- Second-level RSA using the Second-Level RSA dialog.
While RSA can be applied to many experiments, it is particularly well suited for designs with many conditions. For a detailed description and discussion of RSA, see Kriegeskorte et al. (2008).
Kriegeskorte N, Mur M, Bandettini P (2008). Representational similarity analysis - connecting the branches of systems neuroscience. Frontiers in Systems Neuroscience, 2, 4-10.
Nili H, Wingfield C, Walther H, Su L, Marslen-Wilson W, Kriegeskorte N (2014). A toolbox for representational similarity analysis. PLoS Comput Biol 10(4): e1003553.
Copyright © 2017 Rainer Goebel. All rights reserved.