aiqbench analysis
This research introduces STAMP, a novel Siamese Masked Autoencoder framework that integrates stochastic processes to model temporal dynamics in longitudinal medical imaging. By reframing reconstruction loss as a conditional variational inference objective, the method effectively captures non-deterministic disease progression patterns where deterministic approaches fail. The approach demonstrates superior performance in predicting late-stage outcomes for conditions like Age-Related Macular Degeneration and Alzheimer's Disease using OCT and MRI data.
Stochastic Siamese MAE Pretraining for Longitudinal Medical Images
arXiv:2512.23441v2 Announce Type: replace Abstract: Temporally aware image representations are crucial for capturing disease progression in 3D volumes of longitudinal medical datasets. However, recent state-of-the-art self-supervised learning approaches like Masked Autoencoding (MAE), despite their strong representation learning capabilities, lack temporal awareness. In this paper, we propose STAMP (Stochastic Temporal Autoencoder with Masked Pretraining), a Siamese MAE framework that encodes temporal information through a stochastic process by conditioning on the time difference between the 2 input volumes. Unlike deterministic Siamese approaches, which compare scans from different time points but fail to account for the inherent uncertainty in disease evolution, STAMP learns temporal dynamics stochastically by reframing the MAE reconstruction loss as a conditional variational inference objective. We evaluated STAMP on two OCT and one MRI datasets with multiple visits per patient. STAMP pretrained ViT models outperformed both existing temporal MAE methods and foundation models on different late stage Age-Related Macular Degeneration and Alzheimer's Disease progression prediction which require models to learn the underlying non-deterministic temporal dynamics of the diseases.
Authors:Taha Emre, Arunava Chakravarty, Thomas Pinetz, Dmitrii Lachinov, Martin J. Menten, Hendrik Scholl, Sobha Sivaprasad, Daniel Rueckert, Andrew Lotery, Stefan Sacu, Ursula Schmidt-Erfurth, Hrvoje Bogunović
Abstract:Temporally aware image representations are crucial for capturing disease progression in 3D volumes of longitudinal medical datasets. However, recent state-of-the-art self-supervised learning approaches like Masked Autoencoding (MAE), despite their strong representation learning capabilities, lack temporal awareness. In this paper, we propose STAMP (Stochastic Temporal Autoencoder with Masked Pretraining), a Siamese MAE framework that encodes temporal information through a stochastic process by conditioning on the time difference between the 2 input volumes. Unlike deterministic Siamese approaches, which compare scans from different time points but fail to account for the inherent uncertainty in disease evolution, STAMP learns temporal dynamics stochastically by reframing the MAE reconstruction loss as a conditional variational inference objective. We evaluated STAMP on two OCT and one MRI datasets with multiple visits per patient. STAMP pretrained ViT models outperformed both existing temporal MAE methods and foundation models on different late stage Age-Related Macular Degeneration and Alzheimer's Disease progression prediction which require models to learn the underlying non-deterministic temporal dynamics of the diseases.
Provisional Accept at IEEE TMI. Code is available in this https URL
Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
arXiv:2512.23441 [cs.LG]
(or arXiv:2512.23441v2 [cs.LG] for this version)
Full text under the publisher's open licence · 217 words
Related coverage
MMAP: Multimodal Missing-Aware Pretraining for Longitudinal Alzheimer's Prediction
arXiv · Sep 23, 2026
The pretraining domain outweighs the training objective in setting the privacy-utility trade-off of differentially private medical image analysis
arXiv · Jul 28, 2026
Synthesize Realistic 3D Medical Images at Scale to Ship Pre‑Trained Models
NVIDIA · May 22, 2026
Evaluating the trustworthiness of the Fr\'echet Inception Distance with stochastic embedding representations
arXiv · Aug 19, 2026