Introduction Frontotemporal lobar degeneration (FTLD) includes heterogenous diseases: behavioural variant frontotemporal dementia (bvFTD), primary progressive aphasias (PPA), progressive supranuclear palsy (PSP) and corticobasal syndrome (CBS). We applied neuroanatomical normative modelling to quantify individual atrophy patterns and heterogeneity within and between FTLD forms. Methods We included 160 participants across FTLDNI and 4RTNI studies: controls (n = 15), bvFTD (n = 22), nfvPPA (n = 14), svPPA (n = 21), CBS (n = 43) and PSP (n = 45). Using cortical thickness and subcortical volumes from 3T MRIs, we applied normative modelling with a large healthy reference dataset (n = 58,836), further accounting for age, sex, and scanner. Outlier regions (z < – 1.96) were used to compute total outlier counts (tOC) and Hamming distances, capturing individual atrophy patterns and inter-subject dissimilarity. Results bvFTD, svPPA, CBS and PSP showed significantly higher cortical tOC than controls, with all groups showing higher subcortical tOC than controls, especially svPPA and PSP. bvFTD, svPPA, CBS and PSP had significantly higher cortical Hamming distance scores than controls, with higher scores in bvFTD and svPPA than nfvPPA and PSP. svPPA and PSP had significantly higher subcortical scores than controls and CBS. Greater disease severity (measured using the Clinical Dementia Rating—CDR for PSP and CBS, and the CDR® plus NACC-FTLD global scores for FTD variants) was associated with increased tOC and dissimilarity, highlighting the link between clinical progression and neuroanatomical heterogeneity. Conclusions The pronounced heterogeneity within and between FTLD subtypes (particularly in bvFTD) increases with disease progression and may reflect distinct underlying pathologies. This supports the development of subtype-specific biomarkers and emphasize the need for personalized diagnostic and therapeutic strategies.

UCL is consistently ranked as one of the top ten universities in the world (QS World University Rankings 2010-2022) and is No.2 in the UK for research power (Research Excellence Framework 2021).

Illustration of the quality control (QC) process and investigator-guided decision making for data quality. Green checkmarks indicate passed QC, while red cross marks indicate failed QC.