Exploring semantic memory areas by means of parametric models of f-MRI data

In 10 right-handed volunteers, BOLD f-MRI activity was acquired during a protocol of mental speech production tasks. A task lasted 30 sec and consisted in recalling as many as possible animals (AN) or tools (TO). The protocol include 3 AN tasks and 3 TO tasks, each preceded by 30 sec rest (RE). By means of EPI the cerebral volume was acquired with a repetition time of 3 sec. After adjustment in stereotactic coordinates, spatial and temporal filtering, the time series of average activity in spherical regions with a 5 min radius were extracted relevant to Broca's (BR) area, the left mesial (MFG) and lateral (LFG) fusiform gyrus. An interaction model, estimating a modulation index of the connection between either MGF or LFG with BR was able to detect an increased MFG/BR connectivity during TO and increased LFG/BR connectivity during AN; however, this model failed in a subject by subject analysis. On the contrary a simpler correlation model, which separately analyzed the two tasks, was able to detect consistent changes in 7/10 MSG areas and in 9/10 LFG areas. We conclude that with an appropriate data analysis model the application of functional imaging can be widened from pure physiological studies to the clinical analysis in lesions or pathologies involving semantic memory and speech areas.