AbstractBackgroundSubjects with cognitive impairment and dementia often suffer from neuropsychiatric symptoms (NPS). Recent studies reported the association between brain structural damages and NPS. However, little is known about how domains of NPS map to region‐specific multimodal macro‐ and micro‐structural brain changes and whether the patterns vary across different cognitive stages. Importantly, most existing evidence is cross‐sectional. Here, we sought to investigate how baseline cortical thickness (CTh) and white matter (WM) microstructure influenced longitudinal progression of NPS in predementia and dementia.MethodLongitudinal neuropsychiatric inventory (NPI) total severity and subsyndrome scores (Hyperactivity, Psychosis, Affective, Apathy) over 5 years were reported in 334 memory‐clinic patients (187 cognitive impairment no dementia (CIND) and 147 dementia). All participants underwent T1‐weighted structural and diffusion MRI at baseline. A free‐water (FW) imaging method was applied to derive individual‐level WM‐FW and tissue‐compartment maps from diffusion‐MRI. The grey matter (GM) CTh was derived using Freesurfer. We examined the association of baseline WM‐FW, WM‐tissue or GM‐CTh with baseline/longitudinal NPI (total severity and sub‐syndrome) using general linear models.ResultIn CIND, higher total NPI severity was associated with widespread WM‐FW increases (fronto‐parietal projection, temporal and callosal fibres) and more focal tissue fractional‐anisotropy (FAt) reductions (fronto‐parietal/callosal fibres) cross‐sectionally (Fig‐1). Specifically, such WM‐measures were associated with affective and hyperactive sub‐syndromes, but not with psychosis and apathy. No NPI‐cortical thickness association was detected in CIND. By contrast, in dementia, GM‐CTh in frontal executive and default mode and central sensory networks, but not WM‐alteration, was related to higher total NPI severity (Fig‐2). Also, GM‐CTh was associated with apathy and psychosis, but not with affective and hyperactive sub‐syndromes.Furthermore, baseline WM‐measures were associated with longitudinal changes in total NPI severity (Fig‐3) and affective and hyperactive in CIND. By contrast, no association was found between GM‐CTh and longitudinal NPI (total and subsyndromes) in dementia.ConclusionThis study demonstrated that WM‐abnormalities, which were early marker of disease, were related to NPS in CIND, while GM‐CTh was associated with NPS in dementia. Early identification and interventions should be developed for timely management of affective and hyperactive disorders, followed up by close monitoring of the progression of cerebral structural disruptions.