Align-cDAE: Alzheimer's Disease Progression Modeling with Attention-Aligned Conditional Diffusion Auto-Encoder

Ayantika Das; Keerthi Ram; Mohanasankar Sivaprakasam

cs.CVMar 2, 2026

Align-cDAE: Alzheimer's Disease Progression Modeling with Attention-Aligned Conditional Diffusion Auto-Encoder

Authors: Ayantika Das, Keerthi Ram, Mohanasankar Sivaprakasam

Organizations: Department of Electrical Engineering, Indian Institute of Technology Madras, Chennai, India · Sudha Gopalakrishnan Brain Centre, Indian Institute of Technology Madras, Chennai, India

Abstract

Generative AI framework-based modeling and prediction of longitudinal human brain images offer an efficient mechanism to track neurodegenerative progression, essential for the assessment of diseases like Alzheimer's. Among the existing generative approaches, recent diffusion-based models have emerged as an effective alternative to generate disease progression images. Incorporating multi-modal and non-imaging attributes as conditional information into diffusion frameworks has been shown to improve controllability during such generations. However, existing methods do not explicitly ensure that information from non-imaging conditioning modalities is meaningfully aligned with image features to introduce desirable changes in the generated images, such as modulation of progression-specific regions. Further, more precise control over the generation process can be achieved by introducing progression-relevant structure into the internal representations of the model, lacking in the existing approaches. To address these limitations, we propose a diffusion autoencoder-based framework for disease progression modeling that explicitly enforces alignment between different modalities. The alignment is enforced by introducing an explicit objective function that enables the model to focus on the regions exhibiting progression-related changes. Further, we devise a mechanism to better structure the latent representational space of the diffusion auto-encoding framework. Specifically, we assign separate latent subspaces for integrating progression-related conditions and retaining subject-specific identity information, allowing better-controlled image generation. These results demonstrate that enforcing alignment and better structuring of the latent representational space of diffusion auto-encoding framework leads to more anatomically precise modeling of Alzheimer's disease progression.

Figures & tables

Fig. 1: Flow diagram indicating the desirable information to be modeled in the latent representational space of our approach.

Fig. 2: Left to right: The condition encoder ( $\mathcal{C}$ ) integrates progression information with the latent representation of baseline image ( $x_{b}$ ) produced by the encoding component ( $\mathcal{E}$ ), guiding the denoising decoder ( $\mathcal{D}$ ) to generate follow-up image ( $\hat{x}_{f}$ ). Cross-attention ( $A_{l}$ ) is computed between the conditioning vector and decoder layers, and the objective functions enforce alignment of $A_{l}$ with the progression-specific mask ( $M$ ).

Methods	PSNR (dB) ( $\uparrow$ )	SSIM ( $\uparrow$ )	MSE ( $\downarrow$ )
	CN/ MCI & AD	CN/ MCI & AD	CN/ MCI & AD
Naive Baseline	27.25 $\pm$ 2.12/ 26.75 $\pm$ 2.07	0.93 $\pm$ 0.021/ 0.92 $\pm$ 0.021	0.0021 $\pm$ 0.001/ 0.0024 $\pm$ 0.001
IPGAN [ 24 ]	25.86 $\pm$ 2.12/ 25.31 $\pm$ 2.13	0.92 $\pm$ 0.032/ 0.91 $\pm$ 0.034	0.0030 $\pm$ 0.001/ 0.0034 $\pm$ 0.002
BrLP [ 17 ]	26.71 $\pm$ 1.02/ 26.20 $\pm$ 1.14	0.79 $\pm$ 0.022/ 0.79 $\pm$ 0.025	0.0029 $\pm$ 0.001/ 0.0030 $\pm$ 0.001
DE-CVAE [ 4 ]	27.32 $\pm$ 2.98/ 26.99 $\pm$ 2.83	0.65 $\pm$ 0.090/ 0.63 $\pm$ 0.082	0.0023 $\pm$ 0.001/ 0.0024 $\pm$ 0.001
SITGAN [ 23 ]	28.73 $\pm$ 3.25/ 28.09 $\pm$ 3.23	0.94 $\pm$ 0.033/ 0.93 $\pm$ 0.034	0.0019 $\pm$ 0.001/ 0.0022 $\pm$ 0.001

TABLE I: Quantitative evaluations of Align-cDAE with the baseline methods in terms of image-level metrics (PSNR, SSIM, MSE). Statistical significance (

p<0.01

) is marked with ( ∗ ).

Fig. 3: Left to right: In sub-figures (A) and (B), the columns (a)-(f) present a method-wise comparison of predicted follow-up images ( $\hat{x}_{f}$ ) along with their error maps computed with respect to the ground-truth follow-up ( $|\hat{x}_{f}-x_{f}|$ ) and ground truth baseline ( $|x_{b}-\hat{x}_{f}|$ ), respectively. Column (g) in sub-figure (A) shows the ground-truth follow-up ( ${x}_{f}$ ), while in sub-figure (B) it also shows the ground-truth follow-up and baseline difference ( $|x_{b}-x_{f}|$ ). Top to bottom: For both sub-figures, the upper and lower two rows correspond to subjects aged 83.3 and 84.3 years, respectively, from the AD disease category. In sub-figure (A), the green box highlights that Align-cDAE better predicts progression-related changes.

Methods	MAE ( $\downarrow$ )
	Hippocampus	Amygdala	Lateral Ventricles
IPGAN [ 24 ]	0.3348 $\pm$ 0.0338	0.3283 $\pm$ 0.0260	0.5445 $\pm$ 0.5240
BrLP [ 17 ]	0.1960 $\pm$ 0.0552	0.1731 $\pm$ 0.0529	0.3702 $\pm$ 0.1012
DE - CVAE [ 4 ]	0.1871 $\pm$ 0.0727	0.1183 $\pm$ 0.1117	0.1747 $\pm$ 0.1470
SITGAN [ 23 ]	0.1161 $\pm$ 0.0288	0.0211 $\pm$ 0.0201	0.1436 $\pm$ 0.1053
cDAE	0.0550 $\pm$ 0.0500	0.0200 $\pm$ 0.0400	0.1250 $\pm$ 0.1200

TABLE II: Volumetric assessment of Align-cDAE with baseline methods in terms of MAE in different anatomical regions. Statistical significance (

p<0.01

) is marked with ( ∗ ).

Fig. 4: Left to right: (a) Ground truth, (b) cDAE, and (c) Align-cDAE. Top to bottom: (i) Ground-truth and predicted follow-up images, (ii) (a) Ground truth baseline and (b)-(c) model attention maps overlaid on images, and (iii) Difference between ground-truth/ predicted follow-up and ground-truth baseline images.

Data Mixture	Accuracy (%)
$100\%$ RD $+0\%$ GD	$81.92$
$75\%$ RD $+25\%$ GD	$82.02$
$50\%$ RD $+50\%$ GD	$82.15$
$25\%$ RD $+75\%$ GD	$82.00$
$0\%$ RD $+100\%$ GD	$81.93$

TABLE III: Cognitive state classification accuracy under different mixtures of real baseline data (RD) and generated follow-up data (GD).

Methods	PSNR( $\uparrow$ ) (dB)	SSIM( $\uparrow$ )	FID( $\downarrow$ )	Accuracy( $\uparrow$ ) (%)
	CN/ MCI & AD			Classify CN vs. AD
Align-cDAE (w/o L_attn_imax)	28.58/ 28.54	0.945/ 0.938	7.56/ 7.88	80.99
Align-cDAE	29.82/ 29.52	0.950/ 0.944	6.85/ 6.93	81.93

TABLE IV: Ablation study comparing the image similarity and diversity metrics, along with downstream classification performance.

Diffusion Autoencoder

See all 11 papers

Nov 8, 2025

AD-DAE: Alzheimer's Disease Progression Modeling with Unpaired Longitudinal MRI using Diffusion Auto-Encoders

Generative modeling frameworks have emerged as an effective approach to capture high-dimensional image distributions from large datasets without requiring domain-specific knowledge, a capability essential for disease progression modeling. Recent generative approaches have attempted to capture progression by mapping images to a latent space and guiding representations to generate follow-up images from previous time points. However, these methods impose constraints on distribution learning, resulting in latent spaces with limited controllability for generating follow-up images without paired subject-specific longitudinal guidance. In order to enable controlled movements in the latent representational space and generate progression images from a previous time-point image without subject-specific guidance, we introduce a conditionable Diffusion Auto-encoder framework that forms a compact latent space capturing high-level semantics and providing means to control generation. Our approach leverages this latent space to condition and apply controlled shifts to the representations of previous time-point images by isolating progression and subject identity information for generating follow-up images. The shifts are implicitly guided by correlating with progression attributes and constraining to Alzheimer's disease specific regions, without paired longitudinal guidance. We validate the generations through image quality metrics, volumetric progression analysis, and downstream tasks in Alzheimer's disease datasets from different sources. This demonstrates the effectiveness of our approach for Alzheimer's progression modeling and longitudinal image generation.

Ayantika Das, Arunima Sarkar, Keerthi Ram +1

Department of Electrical Engineering, Indian Institute of Technology Madras (IITM), Chennai 600036, India · Sudha Gopalakrishnan Brain Centre (SGBC), IITM, Chennai 600036, India

Medical Image Synthesis Alzheimer's Disease Progression Prediction Diffusion Autoencoder Conditional Generative Modeling Diffusion Models

Ask PDF

Jan 21, 2026

Anatomically Guided Latent Diffusion for Brain MRI Progression Modeling

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

Latent Diffusion Models Longitudinal Medical Image Analysis Medical Image Synthesis Alzheimer's Disease Progression Prediction Conditional Diffusion Models

Ask PDF

Apr 17, 2026

CLIMB: Controllable Longitudinal Brain Image Generation using Mamba-based Latent Diffusion Model and Gaussian-aligned Autoencoder

Latent diffusion models have emerged as powerful generative models in medical imaging, enabling the synthesis of high quality brain magnetic resonance imaging scans. In particular, predicting the evolution of a patients brain can aid in early intervention, prognosis, and treatment planning. In this study, we introduce CLIMB, Controllable Longitudinal brain Image generation via state space based latent diffusion model, an advanced framework for modeling temporal changes in brain structure. CLIMB is designed to model the structural evolution of the brain structure over time, utilizing a baseline MRI scan and its acquisition age as foundational inputs. Additionally, multiple conditional variables, including projected age, gender, disease status, genetic information, and brain structure volumes, are incorporated to enhance the temporal modeling of anatomical changes. Unlike existing LDM methods that rely on self attention modules, which effectively capture contextual information from input images but are computationally expensive, our approach leverages state space, a state space model architecture that substantially reduces computational overhead while preserving high-quality image synthesis. Furthermore, we introduce a Gaussian-aligned autoencoder that extracts latent representations conforming to prior distributions without the sampling noise inherent in conventional variational autoencoders. We train and evaluate our proposed model on the Alzheimers Disease Neuroimaging Initiative dataset, consisting of 6,306 MRI scans from 1,390 participants. By comparing generated images with real MRI scans, CLIMB achieves a structural similarity index of 0.9433, demonstrating notable improvements over existing methods.

Duy-Phuong Dao, Muhammad Taqiyuddin, Jahae Kim +4

Department of Artificial Intelligence Convergence, Chonnam National University, South Korea · Department of Nuclear Medicine, Chonnam National University, South Korea · University of South Australia, Australia +2

Latent Diffusion Models Longitudinal Medical Image Analysis Medical Image Synthesis Neuroimaging Brain MRI Diffusion Models

Ask PDF

Abstract

Diffusion Autoencoder

Explore similar work

AD-DAE: Alzheimer's Disease Progression Modeling with Unpaired Longitudinal MRI using Diffusion Auto-Encoders

Anatomically Guided Latent Diffusion for Brain MRI Progression Modeling

CLIMB: Controllable Longitudinal Brain Image Generation using Mamba-based Latent Diffusion Model and Gaussian-aligned Autoencoder