Covariate-Conditioned One-Step Diffusion for Multi-Horizon 3D Brain MRI Prediction in Aging and Neurodegeneration
Authors: Natenaile Asmamaw Shiferaw, Ali Farki, Deepika Koundal, Jussi Tohka
Organizations: A.I. Virtanen Institute for Molecular Sciences, University of Eastern Finland, Kuopio, Finland · University of Petroleum and Energy Studies (UPES), Dehradun, India
Predicting longitudinal brain changes from magnetic resonance imaging (MRI) can help characterize structural progression associated with aging and neurodegeneration. However, diffusion-based forecasting typically requires iterative denoising, limiting inference efficiency. We propose a covariate-conditioned one-step diffusion framework for multi-horizon prediction of 3D gray-matter density (GMD) maps. Rather than directly generating future anatomy, the model predicts longitudinal GMD changes through residual prediction in a single denoising step. The predicted changes are added to the baseline GMD map to reconstruct future anatomy. The model is conditioned on baseline anatomy, age, diagnosis, APOE4 allele count, and continuous follow-up time. Training combines mean squared error with a Pearson correlation loss to improve the spatial agreement between predicted and observed GMD changes. A unified model supports prediction horizons from 2 to 10 years. On the held-out ADNI test set, our method achieves a Δ-Pearson of 0.478, mean squared error (MSE) of 5.176×10−3, and structural similarity index measure (SSIM) of 0.979, with Δ-Pearson reaching 0.582 at 10 years. Without retraining, external evaluation on AIBL yields a Δ-Pearson of 0.450, MSE of 4.016×10−3, and SSIM of 0.985. Ablation studies further demonstrate the benefits of residual prediction, covariate conditioning, and correlation-based supervision. Code: https://github.com/natenaile/OneStepBrainMRI.
Figures & tables
Figure 1: Overview of the proposed framework. (a) Training with noisy longitudinal GMD changes, anatomical conditioning, and subject-specific information. (b) One-step inference for predicting future GMD maps without iterative denoising.
Interval
Pairs
Participants
Age (years)
ADNI (Internal Dataset)
2 y
3440
1434
73.3±7.0
3 y
1942
858
73.5±6.8
4 y
1568
676
73.0±6.6
5 y
1044
448
73.0±6.4
6 y
871
402
72.9±6.5
Table 1: Longitudinal cohort characteristics of ADNI and the external AIBL test set. Participants are counted uniquely within each interval; age is mean ± SD.
Table 2: Summary of the experimental configuration.
Figure 2: Qualitative multi-horizon GMD predictions. Columns represent follow-up intervals from 2 to 9 years, while rows show ground-truth GMD maps, predicted maps, and absolute error maps in axial, sagittal, and coronal views.
Whole-Brain Change
Reconstructed GMD
Regional Δ -Pearson
Horizon
Δ -Pearson ↑
MSE ↓ ( ×10−3 )
MAE ↓ ( ×10−2 )
Pearson ↑
PSNR ↑ (dB)
SSIM ↑
Hippocampus ↑
Entorhinal Cortex ↑
ADNI (Main Test Set)
2 y
0.444±0.060
4.322±1.954
4.622±0.987
0.936
26.124
0.982
0.455
0.479
3 y
0.464±0.067
4.959±2.340
4.951±1.118
0.927
25.543
0.980
0.494
0.503
4 y
0.478±0.068
5.313±2.579
5.118±1.194
0.923
25.367
0.978
0.513
0.515
5 y
0.498±0.072
5.916±2.900
5.435±1.285
0.914
24.972
0.976
0.531
0.536
Table 3: Multi-horizon performance on the ADNI main test set and the external AIBL test set. whole-brain change metrics are reported as mean ± SD.
Spatiotemporal forecasting of 3D anatomical changes is a critical generative vision challenge for modeling Alzheimer's Disease (AD) progression. While brain MRI forecasting from single-modality data has advanced, other key complementary multimodal biomarkers remain underutilized. In this work, we propose the Multimodal-Conditioned Latent Diffusion Model (MC-LDM), a novel framework for patient-specific brain MRI synthesis. MC-LDM leverages multimodal biomarkers---MRI, tau-PET, Clinical Dementia Rating (CDR), and APOE4---during training, but relies only on baseline MRI, APOE4, and patient metadata at inference, reflecting realistic clinical availability. Our contributions are threefold: (1) a Latent Diffusion Model (LDM) fine-tuned via Low-Rank Adaptation (LoRA) for data-efficient learning; (2) a multimodal conditioning that integrates molecular and clinical signals to model heterogeneous disease trajectories; and (3) auxiliary predictors estimating unobserved multimodal biomarkers from MRI features and patient covariates to condition the diffusion model, removing the need for these assessments at inference. Trained on 1,355 visits from 852 subjects in the Alzheimer's Disease Neuroimaging Initiative (ADNI), MC-LDM reduces average regional volumetric MAE by 18% over the second-best baseline on an independent test set of 664 visits. To our knowledge, MC-LDM represents the first diffusion-based framework to leverage multimodal priors for MRI forecasting. This conceptual advancement enables patient-specific trajectory generation that captures aspects of AD heterogeneity not observable from structural anatomy alone, bridging the gap between rich multimodal supervision and the strict constraints of routine clinical deployment. Code is available at: https://github.com/antonioscardace/MC-LDM.
Antonio Scardace, Silvia Ingala, Sebastiano Battiato +1
University of Catania · Copenhagen University Hospital – Rigshospitalet · Copenhagen University Hospital – Herlev and Gentofte +2
Predicting future structural MRI of a brain is challenging because longitudinal changes are often subtle and confined to specific anatomical regions, while most subject-specific brain structure remains stable over time. An effective model should therefore preserve global brain structural consistency while remaining sensitive to fine-grained disease progression. Existing latent-space-based methods improve computational efficiency, but suffer from information loss during their compression-reconstruction procedure. In contrast, direct voxel-space methods avoid latent reconstruction but commonly use a unified prediction pathway to model brain structure and progression-related changes. Subtle local changes may therefore be overshadowed by the dominant stable brain structure. To address these challenges, we propose ProgFormer, a hierarchical voxel-space Diffusion Transformer for longitudinal brain MRI prediction. ProgFormer uses a coarse pathway to perform the primary volumetric prediction from 3D patch tokens. This pathway models overall brain structure and longitudinal context. The fine pathway then uses the coarse representations as spatio-temporal grounding for voxel-level refinement within individual patches. The two pathways jointly estimate a velocity field directly in voxel space through conditional flow matching, enabling end-to-end prediction without a separately learned image autoencoder. The predicted future scan is then generated from Gaussian noise by integrating the estimated velocity field over a sequence of Euler steps. Extensive experimental results on three widely used benchmarks, ADNI, AIBL, and OASIS, under both pairwise and trajectory settings demonstrate favourable performance compared against several state-of-the-art methods.
Dexuan Ding, Yuankai Qi, Luping Zhou +3
Macquarie University · University of Sydney · University of California Merced
Accurately modeling longitudinal brain MRI progression is crucial for understanding neurodegenerative diseases and predicting individualized structural changes. Existing state-of-the-art approaches, such as Brain Latent Progression (BrLP), often use multi-stage training pipelines with auxiliary conditioning modules but suffer from architectural complexity, suboptimal use of conditional clinical covariates, and limited guarantees of anatomical consistency. We propose Anatomically Guided Latent Diffusion Model (AG-LDM), a segmentation-guided framework that enforces anatomically consistent progression while substantially simplifying the training pipeline. AG-LDM conditions latent diffusion by directly fusing baseline anatomy, noisy follow-up states, and clinical covariates at the input level, a strategy that avoids auxiliary control networks by learning a unified, end-to-end model that represents both anatomy and progression. A lightweight 3D tissue segmentation model (WarpSeg) provides explicit anatomical supervision during both autoencoder fine-tuning and diffusion model training, ensuring consistent brain tissue boundaries and morphometric fidelity. Experiments on 31,713 ADNI longitudinal pairs and zero-shot evaluation on OASIS-3 demonstrate that AG-LDM matches or surpasses more complex diffusion models, achieving highly competitive image quality and 15-20% reduction in volumetric errors in generated images. AG-LDM also exhibits markedly stronger utilization of temporal and clinical covariates (3.5-31.5x higher covariate sensitivity than BrLP) and generates biologically plausible counterfactual trajectories, accurately capturing hallmarks of Alzheimer's progression such as limbic atrophy and ventricular expansion. These results highlight AG-LDM as an efficient, anatomically grounded framework for reliable brain MRI progression modeling.
Cheng Wan, Bahram Jafrasteh, Ehsan Adeli +2
Cornell University · Weill Cornell Medicine · Stanford University +1