Functional Neuroimaging of Sentence Comprehension in Frontotemporal Dementia

Abstract:
As part of an ongoing investigation of the neural basis for grammatical and short-term memory components of sentence comprehension, we studied eight patients with mild frontotemporal dementia (FTD). These patients typically have some sentence comprehension difficulty. We monitored cerebral activity with BOLD fMRI in these patients while they determined the agent of the action in grammatically simple subject-relative (SR) sentences and grammatically complex object-relative (OR) sentences that were semantically unconstrained. To manipulate the short-term memory demand of sentence comprehension, half of each type of sentence contained a short (three-word) distance between the antecedent noun phrase (NP) and the "gap" where the NP is interpreted; half of each sentence type contained a longer (seven-word) antecedent-gap distance. Patients were exposed to eight 40-second blocks of each type of sentence. The printed sentences were presented in a word-by-word fashion, at a rate of 750 msec/word (n=3) or 1000 msec/word (n=5), depending on the ability of the patient to maintain performance at the pace of the task. Scans included a baseline task involving pseudofont target detection. Patients were imaged on a 1.5T GE Echospeed scanner, and gradient echo echoplanar images were obtained to detect alterations in blood oxygenation (5mm slice thickness, effective TE 50msec, 64 x 40 matrix, voxel size of 3.75 x 3.75 x 5mm). The images were registered, aligned to Talairach pace, smoothed with 8mm and 2.8sec Gaussian kernels, and analyzed with SPM96 using appropriate protection for multiple comparisons.

FTD patients showed significant activation of left posterolateral temporal cortex for SR short sentences (OR short and OR long with a smaller extent threshold). Previous study of 7 healthy subjects demonstrated left posterior superior temporal recruitment in all four sentence types in comparison to the baseline task. Left temporal cortex thus appears to retain some ability to support sentence comprehension in FTD. During SR short, SR long and OR long sentences, FTD patients recruited right posterolateral temporal cortex. Healthy subjects also recruited the right posterior superior temporal cortex during SR long and OR long sentences, and it was postulated that this area may serve as a short-term memory buffer to retain sentence material transiently during a long antecedent-gap linkage. Healthy subjects recruited Broca's area in left inferior frontal cortex during OR long sentences. We hypothesized that Broca's area may be recruited in order to block a default canonical construction and re-construct a non-canonical sentence representation in the setting of a lengthy antecedent-gap linkage. While FTD patients recruited left anterior middle temporal cortex adjacent to Broca's area, they did not recruit Broca's area during OR-long sentences. The absence of Broca's area recruitment in FTD suggests one factor contributing to their sentence comprehension difficulty. FTD patients differed from healthy subjects since FTD patients recruited right parietal cortex during SR long, OR short, and OR long sentences, and left medial frontal cortex for SR short, OR short, and OR long sentences. The additional recruitment of left anterior temporal, right parietal, and left medial frontal regions in FTD may represent cortical reorganization in the form of collateral sprouting to functional regions adjacent to compromised brain regions. Alternately, this may represent the attempt to develop other, albeit less effective, cognitive strategies to comprehend grammatically complex sentences.