Poststroke aphasia recovery assessed with functional magnetic resonance imaging and a picture identification task

Stroke patients often display deficits in language function, such as correctly naming objects. Our aim was to evaluate the reliability and the patterns of poststroke language recovery using a picture identification task during functional magnetic resonance imaging (fMRI) at 4 T. Four healthy subjects and 4 subjects with left middle cerebral artery stroke with chronic (>1 year) aphasia were enrolled in the study. In each subject, 10 fMRI scans were performed over a 10-week period using a picture-identification task. The active condition involved presenting subject with a panel of 4 figures (eg, drawings of 4 animals) every 6 seconds and asking the subject to indicate which figure matched the written name in the center. The control condition was a same/different judgment task with pairs of geometric figures (squares, octagons, or combination) presented every 6 seconds. Thirty-second active/control blocks were repeated 5 times each, and responses were recorded. The stoke subjects and controls had similar demographic characteristics, including age (46 vs 53 years), personal handedness (Edinburg Handedness Inventory, 89 vs 95), familial handedness (93 vs 95), and years of education (14.3 vs 14.8). For the active condition, the controls performed better than the stroke subjects (97.7% vs 89.1%; P < .001); the 2 groups performed similarly for the control condition (99.5% vs 98.8%; P = .23). On fMRI, the controls exhibited bilateral, L > R positive blood oxygenation level-dependent (BOLD) activations in frontal and temporal language areas and symmetric retrosplenial and posterior cingulate areas and symmetric negative BOLD activations in bilateral frontotemporal language networks. In contrast, the stroke subjects exhibited positive BOLD activations predominantly in peristroke areas and negative BOLD activations in the unaffected (right) hemisphere. Both groups displayed high activation reliability (as measured by the intraclass correlation coefficient [ICC]) in the left frontal and temporal language areas, although in the stroke subjects the ICC in the frontal regions was spread over a much larger peristroke area. This study documents the utility of the picture-identification task for poststroke language recovery evaluation. Our data suggest that adult stroke patients use functional peristroke areas to perform language functions.