Atrophy-centered subtyping of mild cognitive impairment

Abstract

Mild cognitive impairment (MCI) is considered as the transitional phase between normal cognitive aging and Alzheimer's disease (AD). Nevertheless, trajectories of cognitive decline vary considerably among individuals with MCI. To address this heterogeneity, subtyping approaches have been developed, with the objective of identifying more homogenous subgroups and ultimately improving prognostic outcomes. To date, subtyping of MCI has been based primarily on cognitive performance measures, often resulting in indistinct boundaries between the proposed subgroups and limited validity. The degree to which markers of neurodegeneration such as brain atrophy can be used to subtype MCI into biologically and clinically meaningful subgroups remains unclear. Here we introduce and validate a data-driven subtyping method for MCI based solely upon measures of atrophy derived from structural magnetic resonance imaging (MRI). We trained a dense convolutional neural network to differentiate between patients with AD and age-matched cognitively normal (CN) subjects based on whole brain MRI features. We then deployed the trained model to classify individuals with MCI, as MCI-CN or MCI-AD, based on the degree to which their whole brain gray matter volume resembles CN-like or AD-like patterns. We subsequently validated the model-based subgroups using cognitive, clinical, fluid biomarker, and molecular neuroimaging data. Namely, we observed marked differences between the MCI-CN and MCI-AD groups in baseline and longitudinal cognitive and clinical rating scales, disease-free survival, cerebrospinal fluid (CSF) levels of amyloid beta and tau, fluorodeoxyglucose (FDG) and amyloid PET. Overall, the results suggest that patterns of atrophy in MCI are sufficiently distinct and heterogeneous, and can thus be used to subtype individuals into biologically and clinically meaningful subgroups.

Machine learning on longitudinal multi-modal data enabling the understanding and prognosis of Alzheimer’s disease progression

“…This fundamental property of our model is different from the existing learning models, which assumed patients with AD either follow a common progression pattern or were categorized as several disease subtypes and then separately modeled the progression of each disease subtype 8, 13,15 . Contrary to the finding of Yang et al 10,11 , which considered individuals expressed similar disease manifestations in the initial phrase of AD, we found that participants were initially in either stage A or B, and then progressed to stage C, suggesting that despite the advanced AD is clinically similar, heterogeneity exists in the early course of the disease. Our findings in the ADNI and AIBL cohorts coincide with the study results of another independent clinical cohort, the French MEMENTO cohort, which investigated the brain atrophy subtypes in participants of subjective cognitive complaints or MCI.…”

“…Generally, the progression of AD is over decades, in a non-linear manner, and with heterogenous clinical representations and varied deterioration rates between patients due to complicated genetic and environmental factors interactions 5,6,7 . There is an increasing interest in disease progression models of AD because of their potential application in understanding disease development mechanism 8,9 , guiding patient management, providing disease prognosis and designing treatment strategie 10,11 .…”