Datensatz

Zusammenfassung

How are objects represented in the human brain? This question has previously been addressed using encoding and decoding approaches to neural activity as indexed by the fMRI BOLD response. Specifically, hierarchical clustering based on similarity matrices of brain activation patterns demonstrated that object representations within the inferior temporal gyrus clustered into animate and inanimate categories in line with previous neuropsychological double dissociations. Here, we used Formal Concept Analysis (FCA) (Ganter and Wille 1999) to characterize how the relationship of BOLD activation patterns maps onto the relationship between object stimuli. FCA displays and interprets the relationship of neural object representations via concept lattices (a type of semantic graph). Each concept is defined by a set of formal objects as extent and a set of formal attributes as intent. In our application, the object stimuli were the formal objects and the binarized activations in single voxels the formal attributes. A single subject was scanned while viewing 72 grayscale pictures of animate and inanimate objects in a target detection task (Siemens Trio 3T scanner, GE-EPI, TE=40ms, 38 axial slices, TR=3.08s, 48 sessions, amounting to a total of 10,176 volume images). We modeled the BOLD responses to the stimulus presentations in an event-related fashion with a general linear model, using a separate regressor for each of the 72 stimuli. From the parameter estimate image for each stimulus, the 300 voxels that were most active for all stimuli > fixation were selected within the category-sensitive system as defined by a prior study (including inferior temporal, supramarginal, inferior frontal gyrus). Formal concept analysis was applied to patterns of thresholded and hence binarized voxel activations. In line with previous reports, formal concept analysis revealed a dissociation between animate and inanimate objects with the inanimate category also including plants and vegetables. This study demonstrates the potential strength of FCA for decoding structured relationships in fMRI data.