Distinctive Progression Patterns of Brain Structural Damage Aid Classification of Frontotemporal Dementia Variants

Background: Frontotemporal dementia (FTD) encompasses diverse clinical phenotypes, primarily characterized by behavioral and/or language dysfunction. A newly characterized variant, semantic behavioral variant FTD (sbvFTD), exhibits predominant right temporal atrophy with features bridging behavioral variant FTD (bvFTD) and semantic variant primary progressive aphasia (svPPA). This study investigates the longitudinal structural MRI correlates of these FTD variants, focusing on cortical and subcortical structural damage to aid differential diagnosis and prognosis. Methods: Seventy-one FTD patients (bvFTD = 45, sbvFTD = 11, svPPA = 15) and 37 healthy controls participated in a prospective study involving up to 24 months of serial neurological, neuropsychological, and 3 T MRI assessments. Cortical thickness and subcortical/cerebellar volumes were analyzed with linear mixed-effect models. Support vector machine (SVM) models were used to classify subjects using baseline and longitudinal patterns of structural damage. Results: At baseline, sbvFTD showed right-predominant temporal pole involvement associated with significant right frontal atrophy. Longitudinally, bvFTD showed widespread bilateral cortical and basal ganglia damage, svPPA demonstrated steady temporal lobe progression, and sbvFTD progressed primarily in left temporal and frontal regions with limited right hemisphere involvement. Baseline cortical thickness of frontal regions predicted subsequent functional decline in bvFTD and sbvFTD. A multiclass SVM model provided a good diagnostic classification accuracy, with similar results when using baseline data only (82%) and adding longitudinal data (83%). Conclusions: This study delineates the unique structural MRI features and progression of FTD variants, highlighting sbvFTD as a distinct entity with early extra-temporal involvement. These findings support the development of diagnostic and prognostic tools leveraging neuroimaging biomarkers.