The value of hippocampal volume, shape, and texture for 11-year prediction of dementia: a population-based study

Hippocampal volume and shape are known magnetic resonance imaging biomarkers of neurodegeneration. Recently, hippocampal texture has been shown to improve prediction of dementia in patients with mild cognitive impairment, but it is unknown whether texture adds prognostic information beyond volume and shape and whether the predictive value extends to cognitively healthy individuals. Using 510 subjects from the Rotterdam Study, a prospective, population-based cohort study, we investigated if hippocampal volume, shape, texture, and their combination were predictive of dementia and determined how predictive performance varied with time to diagnosis and presence of early clinical symptoms of dementia. All features showed significant predictive performance with the area under the receiver operating characteristic curve ranging from 0.700 for texture alone to 0.788 for the combination of volume and texture. Although predictive performance extended to those without objective cognitive complaints or mild cognitive impairment, performance decreased with increasing follow-up time. We conclude that a combination of multiple hippocampal features on magnetic resonance imaging performs better in predicting dementia in the general population than any feature by itself. •Investigated magnetic resonance imaging–based dementia prediction in a population setting.•Hippocampal volume, shape, and texture are predictive for dementia in this setting.•The combination of different magnetic resonance imaging–based features improves predictive value.•All features are predictive in subjects without objective cognitive complaints.•Volume and texture combined are predictive up to 6 years before diagnosis.