Live presentation

Asynchronous discussion


Hierarchical decomposition of a large-scale database of cognitive neuroimaging activation maps using graph-theoretical tools.

Matías Palmucci

  • CABA,
  • Argentina
  • Matías Palmucci ¹
  • , Enzo Tagliazucchi ¹
  • 1 Computational Cognitive Neuroscience Lab (COCUCO), Departamento de Física, FCEyN, UBA
  • 2 Escuela de Ciencia y Tecnología, Universidad Nacional de San Martín (UNSAM), Buenos Aires
  • 3 Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET)

Since the seminal work by Biswal et al. (Biswal et al., 1995) – later expanded using multivariate methods by Beckmann and colleagues (Beckmann et al., 2005) – it is known that spontaneous brain activity recorded using fMRI presents a spatio-temporal organization consistent with well-defined neural systems. This correspondence was revealed for the first time by Smith and colleagues (Smith et al., 2009), who compared the independent components obtained from a database of fMRI task activation maps ( with those obtained from resting state fMRI data. The striking correspondence between both sets of components suggested that spontaneous brain activity recapitulates spatio-temporal patterns that might be required for the rapid reaction to environmental demands. A total of 3072 association test maps of activation meta-analysis were downloaded from ( of which we classified 400 maps corresponding to different terms associated with cognitive processes. Combining graph-theoretical tools with modularization optimization algorithms, we performed a hierarchical clustering of these maps and observed task-positive and negative clusters at a coarse-level, which were then subdivided into maps associated with well-defined functions. In contrast with the work by Smith et al., the correspondence between task-derived maps and resting state networks was only manifest at an intermediate resolution.