Fig. 1

Methodological overview. A Recruitment from three centers in the Latam database consisted of demographically matched 18 bvFTD patients, 39 AD patients, and 57 HCs, while 100 bvFTD patients, 100 AD patients, and 100 HCs were downloaded to obtain a demographically matched HIC dataset. B Model inputs consisted of subject-specific FC in the AAL-90 atlas, group averaged atrophy patterns as 90 w-scores, and a 90 region DTI structural atlas. C Whole-brain modeling scheme fitting the \(\sigma\) feedback inhibition parameter that multiplies the atrophy vector in the feedback inhibition current \({J}_{n}\) equation, tuned by Bayesian optimization (D) Model output consisting of simulated BOLD signals transformed from the excitatory firing rates using the Balloon-Windkessel model. E Hypothesis testing including the evaluation of the \(\sigma\) parameter, associations with YWD and sex, brain region-specific FC changes with YWD, and \(\sigma\) parameter variability analysis across datasets. FC: functional connectivity. YWD: years with disease; AD: Alzheimer’s disease. bvFTD: behavioral variant frontotemporal dementia. HC: healthy controls. DTI: diffusion tensor imaging