Learning Patient-Specific Disease Dynamics with Latent Flow Matching for Longitudinal Imaging Generation

2025-12-09

AI Summary

Existing longitudinal medical imaging models suffer from latent representations lacking semantic structure, dynamic continuity, and alignment with clinical severity. To address this, we propose Δ-LFM: a framework that models individualized disease progression as a monotonic, continuous velocity field in latent space, enabling MRI sequence generation and dynamic interpretation via flow matching. We introduce the first patient-specific latent space alignment mechanism, ensuring disease trajectories extend monotonically along a single semantic axis explicitly linked to clinical scores. By integrating autoencoders with flow matching, Δ-LFM jointly optimizes generation fidelity, dynamic interpretability, and cross-patient consistency. Evaluated on three longitudinal MRI datasets, Δ-LFM significantly improves image generation fidelity and disease progression prediction accuracy. Moreover, it enables visualizable, interpretable, patient-specific trajectory inference—bridging generative modeling with clinically grounded longitudinal analysis.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation
• Computer Vision: Diffusion Models for Vision

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Understanding disease progression is a central clinical challenge with direct implications for early diagnosis and personalized treatment. While recent generative approaches have attempted to model progression, key mismatches remain: disease dynamics are inherently continuous and monotonic, yet latent representations are often scattered, lacking semantic structure, and diffusion-based models disrupt continuity with random denoising process. In this work, we propose to treat the disease dynamic as a velocity field and leverage Flow Matching (FM) to align the temporal evolution of patient data. Unlike prior methods, it captures the intrinsic dynamic of disease, making the progression more interpretable. However, a key challenge remains: in latent space, Auto-Encoders (AEs) do not guarantee alignment across patients or correlation with clinical-severity indicators (e.g., age and disease conditions). To address this, we propose to learn patient-specific latent alignment, which enforces patient trajectories to lie along a specific axis, with magnitude increasing monotonically with disease severity. This leads to a consistent and semantically meaningful latent space. Together, we present $Δ$-LFM, a framework for modeling patient-specific latent progression with flow matching. Across three longitudinal MRI benchmarks, $Δ$-LFM demonstrates strong empirical performance and, more importantly, offers a new framework for interpreting and visualizing disease dynamics.

Problem

• Model continuous monotonic disease progression from medical images
• Align patient-specific latent representations with clinical severity indicators
• Generate interpretable longitudinal imaging for personalized treatment insights

Innovation

• Flow Matching models disease as velocity field
• Patient-specific latent alignment ensures monotonic severity progression
• Latent space structured for interpretable disease dynamics visualization