Temporal Conditioning for Longitudinal Brain MRI Registration and Aging Analysis.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Longitudinal brain analysis is crucial for understanding aging and diseases.
• Current brain MRI registration methods struggle with sparse data, accuracy-time trade-offs, and future state prediction.
• Existing techniques often require densely sampled time series and struggle with temporal coherence.

Purpose of the Study:

• To introduce TimeFlow, a learning-based framework for accurate and temporally coherent longitudinal brain MRI registration.
• To enable prediction of future brain states from limited longitudinal data.
• To develop an annotation-free method for analyzing neurodegenerative trajectories versus normal aging.

Main Methods:

• TimeFlow utilizes a U-Net backbone with temporal conditioning to model neuroanatomy over age.
• It employs inter-/extrapolation consistency constraints for deformation fields and images.
• The framework models neuroanatomy as a continuous function of age, enabling extrapolation.

Main Results:

• TimeFlow accurately forecasts future brain states and improves registration accuracy compared to state-of-the-art methods.
• The framework demonstrates temporal consistency and continuity without explicit smoothness regularizers.
• It successfully differentiates neurodegenerative trajectories from normal aging without segmentation.

Conclusions:

• TimeFlow provides an accurate, data-efficient, and annotation-free framework for longitudinal brain MRI analysis.
• The method enables forecasting of brain changes beyond the observed study period.
• It facilitates novel analyses of brain aging and chronic diseases by eliminating segmentation requirements.