Abstract
Alzheimer's disease (AD)-related primary progressive aphasia (PPA) exhibits considerable heterogeneity in clinical presentation and neuroimaging patterns. No studies have quantitatively assessed cerebral perfusion patterns or systematically evaluated the internal heterogeneity of linguistic and neuroimaging features in this population. This study aimed to investigate cerebral hypoperfusion patterns and elucidate their correlation with diverse linguistic features in patients with AD-related PPA using a data-driven approach. Eleven patients with AD-related PPA and 34 with non-AD-related PPA were categorized based on cerebrospinal fluid biomarkers, and their single-photon emission computed tomography (SPECT) data were analyzed. Cerebral hypoperfusion was assessed across 56 regions of interest (ROIs) covering the entire cerebral hemisphere. Sparse principal component (sPC) analysis was performed on the AD-related PPA group to identify distinct patterns of cerebral perfusion reduction and correlate these components with clinical assessments of linguistic abilities. AD-derived sPCs were identified, reflecting hypoperfusion patterns in the left temporoparietal, frontal, and temporal pole regions, corresponding to regions typically associated with logopenic, nonfluent, and semantic variants. In both AD-PPA and non-AD-PPA, the sPC corresponding to the anterior temporal region was associated with semantic comprehension deficits, whereas that corresponding to the frontal region was linked to nonfluent speech and Kana writing impairment. sPC-based hierarchical clustering revealed clusters corresponding to logopenic, nonfluent, and semantic variants, with the anomic subtype distinguished from logopenic PPA. AD-positive cases were distributed across these clusters, emphasizing AD-PPA heterogeneity. These findings suggested that AD-related PPA heterogeneity is reflected in distinct cerebral perfusion patterns, which correlate with varying linguistic deficits.