Multimodal normative modeling in Alzheimer’s Disease with introspective variational autoencoders

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 17, 2024

Abstract Normative models in neuroimaging learn patterns of healthy brain distributions to identify deviations disease subjects, such as those with Alzheimer’s Disease (AD). This study addresses two key limitations variational autoencoder (VAE)-based normative models: (1) VAEs often struggle accurately model control distributions, resulting high reconstruction errors and false positives, (2) traditional multimodal aggregation methods, like Product-of-Experts (PoE) Mixture-of-Experts (MoE), can produce uninformative latent representations. To overcome these challenges, we developed a introspective VAE that enhances modeling by achieving more precise representations anatomy both the space reconstructions. Additionally, implemented Mixture-of-Product-of-Experts (MOPOE) approach, leveraging strengths PoE MoE efficiently aggregate information improve abnormality detection space. Using biomarkers from Neuroimaging Initiative (ADNI) dataset, our proposed demonstrated superior controls outperformed baseline methods detecting outliers. Deviations calculated aggregated effectively integrated complementary multiple modalities, leading higher likelihood ratios. The exhibited strong performance Out-of-Distribution (OOD) detection, clear separation between cohorts. Z-score specific dimensions were mapped feature-space abnormalities, enabling interpretable identification regions associated AD pathology.

Language: Английский

A Multiform Heterogeneity Framework for Alzheimer’s Disease Based on Multimodal Neuroimaging

Biological Psychiatry, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 1, 2024

Language: Английский