Back to all papers

LS-MAT: Lifespan structural magnetic resonance imaging Synthesis for Microstructural covariance profile Analysis Toolbox.

March 2, 2026pubmed logopapers

Authors

Kim H,Kim J,Kim M,Park H,Park BY

Affiliations (5)

  • Department of Brain and Cognitive Engineering, Korea University, Seoul, Republic of Korea; BK21 Four Institute of Precision Public Health, Seoul, Republic of Korea.
  • Department of Electrical and Computer Engineering, Sungkyunkwan University, Suwon, Republic of Korea.
  • Department of Artificial Intelligence Convergence, Gwangju Institute of Science and Technology, Gwangju, Republic of Korea.
  • Department of Electrical and Computer Engineering, Sungkyunkwan University, Suwon, Republic of Korea; Center for Neuroscience Imaging Research, Institute for Basic Science, Suwon, Republic of Korea.
  • Department of Brain and Cognitive Engineering, Korea University, Seoul, Republic of Korea; Center for Neuroscience Imaging Research, Institute for Basic Science, Suwon, Republic of Korea. Electronic address: [email protected].

Abstract

Human brain structure changes dynamically across the lifespan, closely linked to cognitive development and decline. T1-weighted and T2-weighted magnetic resonance imaging (MRI) are widely used to assess anatomical and microstructural properties. However, acquiring longitudinal multimodal MRI is resource-intensive, limiting characterization of age-related trajectories. Here, we propose LS-MAT, a generative framework for synthesizing personalized, multimodal, and age-conditioned structural MRIs to support lifespan neuroimaging research. The framework integrates: (i) a variational autoencoder with a generative adversarial network for efficient latent encoding, (ii) a latent diffusion model for high-resolution conditional synthesis, and (iii) a ControlNet for modality-guided structural consistency. We trained and evaluated the model on large-scale, publicly available datasets spanning ages 5-100 years. LS-MAT achieved strong performance in modality conversion, measured by peak signal-to-noise ratio, structural similarity index measure, and mean squared error. The generated images captured established developmental and aging trends, including ventricular enlargement, cortical thinning, and age-related trajectories of T1/T2-weighted ratio-based microstructural profiles. Compared with existing methods, our model outperformed previous approaches in both modality conversion and age-conditioned synthesis tasks. These findings highlight the potential of generative modeling to overcome data scarcity in lifespan neuroimaging and provide a powerful tool for studying structural brain changes. The pipeline not only supports longitudinal analyses but also enables the derivation of microstructural profile features and is openly available at https://github.com/hobacteria/LS-MAT.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.