BIOMARKER VALIDATED PERFUSION IMAGING PATTERN RANKING FOR PREDICTION OF ALZHEIMER'S DISEASE

Abstract

Perfusion brain imaging through single photon emission computed tomography (SPECT) is a less favoured imaging modality for dementia diagnosis, however it is widely available and cheaper than its alternatives. Clinical reporting of perfusion SPECT images is difficult and there are few validated ranking scales to aid image interpretation for dementia diagnosis. In addition, statistical parametric mapping (SPM) comparisons of the individual against a control database is seldom used clinically, despite giving improved detection of perfusion abnormality over visual assessment of raw images alone. This study aimed to identify patterns of perfusion abnormality in an unselected clinical sample using individual SPM comparisons that best indicate underlying Alzheimer’s type dementia, and validate them using cerebrospinal fluid (CSF) tau and amyloid protein biomarkers. A series of 81 unselected samples from a dementia diagnostic centre where perfusion SPECT brain imaging and Alzheimer's disease (AD) CSF biomarkers were previously collected in diagnostically ambiguous cases were analysed. Single-subject voxel based hypoperfusion maps were created for each individual when compared to a control group using statistical parametric mapping (SPM8). The voxel based maps were visually grouped based on regional hypoperfusion without prior knowledge of CSF biomarker result. The CSF biomarker status for each individual was consequently obtained, and group observations made. Individuals fell into 8 distinct groups: angular gyrus only, angular gyrus projecting frontally, angular gyrus projecting posteriorly, angular gyrus projecting temporally, frontal only, temporal only, other and no hypoperfusion. Regions in the inferior parietal lobe around the angular gyrus were the most salient for AD biomarkers, with perfusion abnormality in this region having a high likelihood of positive AD biomarkers both when alone and in combination with other regions (30/36 with abnormal amyloid, 21/36 with both abnormal amyloid and tau CSF). Interestingly, CSF biomarker status did not predict the extent of regional abnormality on perfusion imaging. Patients with frontal, temporal, other and no perfusion abnormality were rarely associated with AD type CSF biomarkers (33/45 with no abnormal CSF biomarkers). These results validate regional SPM-based perfusion SPECT patterns and may help increase confidence of diagnostic reporting in Alzheimer’s disease and other dementias.