Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Semi-disentangled spatiotemporal implicit neural representations of longitudinal neuroimaging data for trajectory classification

Agampreet Aulakh, Nils D. Forkert, Matthias Wilms

TL;DR

The paper tackles irregularly sampled longitudinal MRI data by modeling subject-specific brain aging trajectories as continuous functions using Implicit Neural Representations (INRs). It introduces a semi-disentangled spatiotemporal INR with separate spatial and temporal streams and an encoder-based classifier that operates on INR parameter groups P_{space}, P_{time}, and P_{com} to classify aging trajectories. A diffusion-model–based trajectory simulator plus an age-dependent ODE deviation model creates healthy and AD-like synthetic trajectories for controlled evaluation. The approach demonstrates strong trajectory reconstruction fidelity and classification performance, notably achieving 81.3% accuracy under irregular sampling with the time stream, outperforming a baseline SFCN, and highlights the value of operating directly in INR parameter space for longitudinal neuroimaging analysis.

Abstract

The human brain undergoes dynamic, potentially pathology-driven, structural changes throughout a lifespan. Longitudinal Magnetic Resonance Imaging (MRI) and other neuroimaging data are valuable for characterizing trajectories of change associated with typical and atypical aging. However, the analysis of such data is highly challenging given their discrete nature with different spatial and temporal image sampling patterns within individuals and across populations. This leads to computational problems for most traditional deep learning methods that cannot represent the underlying continuous biological process. To address these limitations, we present a new, fully data-driven method for representing aging trajectories across the entire brain by modelling subject-specific longitudinal T1-weighted MRI data as continuous functions using Implicit Neural Representations (INRs). Therefore, we introduce a novel INR architecture capable of partially disentangling spatial and temporal trajectory parameters and design an efficient framework that directly operates on the INRs' parameter space to classify brain aging trajectories. To evaluate our method in a controlled data environment, we develop a biologically grounded trajectory simulation and generate T1-weighted 3D MRI data for 450 healthy and dementia-like subjects at regularly and irregularly sampled timepoints. In the more realistic irregular sampling experiment, our INR-based method achieves 81.3% accuracy for the brain aging trajectory classification task, outperforming a standard deep learning baseline model (73.7%).

Semi-disentangled spatiotemporal implicit neural representations of longitudinal neuroimaging data for trajectory classification

TL;DR

The paper tackles irregularly sampled longitudinal MRI data by modeling subject-specific brain aging trajectories as continuous functions using Implicit Neural Representations (INRs). It introduces a semi-disentangled spatiotemporal INR with separate spatial and temporal streams and an encoder-based classifier that operates on INR parameter groups P_{space}, P_{time}, and P_{com} to classify aging trajectories. A diffusion-model–based trajectory simulator plus an age-dependent ODE deviation model creates healthy and AD-like synthetic trajectories for controlled evaluation. The approach demonstrates strong trajectory reconstruction fidelity and classification performance, notably achieving 81.3% accuracy under irregular sampling with the time stream, outperforming a baseline SFCN, and highlights the value of operating directly in INR parameter space for longitudinal neuroimaging analysis.

Abstract

The human brain undergoes dynamic, potentially pathology-driven, structural changes throughout a lifespan. Longitudinal Magnetic Resonance Imaging (MRI) and other neuroimaging data are valuable for characterizing trajectories of change associated with typical and atypical aging. However, the analysis of such data is highly challenging given their discrete nature with different spatial and temporal image sampling patterns within individuals and across populations. This leads to computational problems for most traditional deep learning methods that cannot represent the underlying continuous biological process. To address these limitations, we present a new, fully data-driven method for representing aging trajectories across the entire brain by modelling subject-specific longitudinal T1-weighted MRI data as continuous functions using Implicit Neural Representations (INRs). Therefore, we introduce a novel INR architecture capable of partially disentangling spatial and temporal trajectory parameters and design an efficient framework that directly operates on the INRs' parameter space to classify brain aging trajectories. To evaluate our method in a controlled data environment, we develop a biologically grounded trajectory simulation and generate T1-weighted 3D MRI data for 450 healthy and dementia-like subjects at regularly and irregularly sampled timepoints. In the more realistic irregular sampling experiment, our INR-based method achieves 81.3% accuracy for the brain aging trajectory classification task, outperforming a standard deep learning baseline model (73.7%).

Paper Structure

This paper contains 10 sections, 2 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Materials and Methods
  3. Semi-Disentangled Spatiotemporal Implicit Neural Representations
  4. Brain Aging Trajectory Classification
  5. Brain Aging Trajectory Simulation
  6. Experimental Setup
  7. Results and Discussion
  8. Conclusion
  9. Acknowledgments.
  10. Disclosure of Interests.

Figures (3)

  • Figure 1: Overview of proposed methods. Left: Subject-specific trajectory modelling with a semi-disentangled spatiotemporal INR. Right: Organization of INR parameters in space-only, time-only, and combined streams, their row-wise embedding in latent space, and subsequent trajectory classification.
  • Figure 2: Left: Exemplary simulated brain age deviation mappings (dashed lines) from our nonlinear ODE model. Solid lines denote the mean mapping and shaded regions indicate standard deviation across 450 simulations. Right: Visualization of sampled subject data as both healthy and AD-like trajectories.
  • Figure 3: Examples of reconstructed scans for simulated subjects with AD-like (orange) and healthy (blue) brain aging trajectories. For both AD-like and healthy trajectories, the top row shows the ground-truth scans and the bottom row shows their corresponding reconstructions. Grey dotted outlines indicate data not used for training. Solid outlines (orange/blue) indicate interpolated reconstructions and dashed outlines (orange/blue) indicate extrapolated reconstructions.