Progression as Latent Drift: Generative Forecasting of Slow-Evolving Pathologies

2026-07-09

AI Summary

This work addresses the challenge of reliably predicting subtle anatomical changes in slowly progressive neurodegenerative diseases from longitudinal MRI, where existing generative models often fail due to weak pathological signals. The authors propose Latent Drift, a framework that models disease-relevant dynamics within a compressed semantic latent space, circumventing direct synthesis of high-dimensional images. By shifting the prediction target from pixel-level identity to representations of latent change and incorporating finite scalar quantization to suppress high-frequency noise, the method effectively mitigates identity collapse and pitfalls of continuous interpolation. Evaluated on longitudinal 3D brain MRI data, Latent Drift outperforms diffusion models and autoregressive Transformer baselines in both generation fidelity and clinically relevant metrics, substantially enhancing patient-specific prediction of neurodegenerative progression.

Abstract

Forecasting the future anatomy of slow-evolving neurodegenerative diseases could enable earlier, more targeted intervention and improve clinical trial design, but it remains challenging because true progression signals are subtle in longitudinal MRI. In this low-signal regime, transferring modern generative sequence models directly is unreliable: training is dominated by stable baseline anatomy and confounded by dense, sample-specific nuisance variation. We first provide a theoretical analysis that explains these failures through two modes. Identity collapse occurs when optimization is driven toward reproducing the current anatomy, which prevents the model from learning faint temporal change. The continuous interpolation trap arises when standard smooth networks cannot separate localized biological drift from pervasive noise, which leads to spurious changes that diffuse across the volume. To address both issues, we propose Latent Drift, a progressive generative framework that learns change in a compressed semantic representation rather than synthesizing full-resolution anatomy. This design removes pixel-level identity from the prediction target and concentrates model capacity on progression-relevant dynamics. We further apply Finite Scalar Quantization to the learned change representation, which suppresses small, high-frequency nuisance fluctuations while preserving consistent structural drift. Experiments on longitudinal 3D brain MRI show that Latent Drift improves patient-specific neuro-forecasting over diffusion and autoregressive transformer baselines across generative fidelity and clinically relevant evaluation metrics. Project page: \href{https://cutepkq.github.io/latent-drift}{https://cutepkq.github.io/latent-drift}.

Problem

• neurodegenerative disease
• longitudinal MRI
• disease progression forecasting
• generative modeling
• anatomical change prediction

Innovation

• Latent Drift
• generative forecasting
• neurodegenerative progression
• finite scalar quantization
• longitudinal MRI