Search papers, labs, and topics across Lattice.
The paper introduces CASL-VAE, a deep contrastive latent variable model designed to learn structured latent generative factors from unpaired data, addressing the challenge of quantifying variability between target and reference populations without paired samples. By factorizing variation into continuous common latent factors and hierarchical salient latent factors, CASL-VAE effectively captures both discrete subtypes and continuous variations within those subtypes. Validation on semi-synthetic neuroimaging data shows that CASL-VAE outperforms existing clustering and generative models in subtype recovery and paired-sample generation, particularly in revealing biologically relevant heterogeneity in Alzheimer's disease.
CASL-VAE uncovers hidden structures in unpaired data, revealing critical insights into disease heterogeneity that traditional methods miss.
Quantifying variability in a target population relative to a reference population is central to many scientific and clinical problems (e.g., diseased vs. healthy). Yet, without paired data and in the presence of heterogeneous target variation, existing methods struggle to separate multiple modes of target-specific variation. We propose \textit{CASL-VAE}, a deep contrastive latent variable model that learns structured latent generative factors from unpaired data. CASL-VAE factorizes variation into continuous common latent factors shared across populations and hierarchical salient latent factors that model target-specific heterogeneity as discrete subtypes and continuous within-subtype variation. Using variational inference, we show how approximate joint likelihood optimization over reference and target domains can be performed using unpaired data, providing a principled basis for paired-sample generation and cross-domain analysis. We validate CASL-VAE on semi-synthetic neuroimaging data, demonstrating improved subtype recovery and paired-sample generation compared to baseline clustering and generative models. We also validate its ability to reveal biologically plausible heterogeneity in Alzheimer's disease.
