CRONOS: Continuous Time Reconstruction for 4D Medical Longitudinal Series

2025-12-18

AI Summary

This work addresses the problem of voxel-level continuous-time prediction for irregularly sampled 3D medical image sequences. We propose the first continuous-time sequence-to-image forecasting framework designed for multi-temporal inputs and capable of reconstructing images at arbitrary real-valued time points. Methodologically, we introduce a physics-inspired, differentiable spatiotemporal velocity field to jointly model discrete and continuous timestamps; integrate multi-context features with continuous-time embeddings; and directly learn and propagate deformations within the 3D voxel space for end-to-end dynamic modeling. Our approach achieves state-of-the-art performance on three public benchmarks—Cine-MRI, perfusion CT, and longitudinal MRI—demonstrating both superior accuracy and computational efficiency. To foster reproducibility and community advancement, we will release our code and a standardized evaluation protocol, establishing a foundational benchmark for multi-center, multi-modal continuous-time medical imaging modeling.

Technology Category

• Computer Vision: Medical and Biological Imaging
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Machine Learning: Time-Series/Data Streams

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Forecasting how 3D medical scans evolve over time is important for disease progression, treatment planning, and developmental assessment. Yet existing models either rely on a single prior scan, fixed grid times, or target global labels, which limits voxel-level forecasting under irregular sampling. We present CRONOS, a unified framework for many-to-one prediction from multiple past scans that supports both discrete (grid-based) and continuous (real-valued) timestamps in one model, to the best of our knowledge the first to achieve continuous sequence-to-image forecasting for 3D medical data. CRONOS learns a spatio-temporal velocity field that transports context volumes toward a target volume at an arbitrary time, while operating directly in 3D voxel space. Across three public datasets spanning Cine-MRI, perfusion CT, and longitudinal MRI, CRONOS outperforms other baselines, while remaining computationally competitive. We will release code and evaluation protocols to enable reproducible, multi-dataset benchmarking of multi-context, continuous-time forecasting.

Problem

• Forecasts 3D medical scan evolution over time for disease progression.
• Supports continuous time prediction from multiple past irregular scans.
• Learns spatio-temporal velocity fields for voxel-level forecasting.

Innovation

• Continuous time reconstruction for 4D medical longitudinal series
• Learns spatio-temporal velocity field for voxel-level forecasting
• Supports discrete and continuous timestamps in unified model