Temporal Flow Matching for Learning Spatio-Temporal Trajectories in 4D Longitudinal Medical Imaging

2025-08-29

AI Summary

Medical imaging temporal modeling faces two key challenges: (1) existing methods predominantly exploit only single-timepoint context, and (2) they lack support for fine-grained 3D spatial prediction. To address these, we propose Temporal Flow Matching (TFM), the first unified generative trajectory modeling framework capable of handling irregularly sampled, multi-prior, and 4D volumetric longitudinal data. TFM explicitly learns latent temporal distributions via time-aware encoding and conditional flow matching, and naturally degenerates to recent-image prediction—ensuring both theoretical consistency and practical utility. Evaluated on three public longitudinal medical imaging datasets, TFM establishes new state-of-the-art performance in 4D medical image forecasting, significantly outperforming prior spatiotemporal models. It demonstrates strong robustness to acquisition variability and cross-disease generalization. This work provides a reliable foundation for disease progression modeling, treatment planning, and developmental trajectory analysis.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Planning, Routing, and Scheduling: Temporal Planning
• Computer Vision: Multi-modal Vision

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs

Abstract

Understanding temporal dynamics in medical imaging is crucial for applications such as disease progression modeling, treatment planning and anatomical development tracking. However, most deep learning methods either consider only single temporal contexts, or focus on tasks like classification or regression, limiting their ability for fine-grained spatial predictions. While some approaches have been explored, they are often limited to single timepoints, specific diseases or have other technical restrictions. To address this fundamental gap, we introduce Temporal Flow Matching (TFM), a unified generative trajectory method that (i) aims to learn the underlying temporal distribution, (ii) by design can fall back to a nearest image predictor, i.e. predicting the last context image (LCI), as a special case, and (iii) supports $3D$ volumes, multiple prior scans, and irregular sampling. Extensive benchmarks on three public longitudinal datasets show that TFM consistently surpasses spatio-temporal methods from natural imaging, establishing a new state-of-the-art and robust baseline for $4D$ medical image prediction.

Problem

• Modeling disease progression in 4D medical imaging
• Overcoming limitations in spatio-temporal prediction methods
• Learning temporal distributions with irregular sampling support

Innovation

• Temporal Flow Matching for generative trajectory learning
• Supports 3D volumes and irregular temporal sampling
• Unified framework with fallback to nearest image prediction