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

2026-04-21 · Source: cs.AI updates on arXiv.org · Field: Health & Wellbeing — Health & Medical Research, Medical Devices & Health Technology · Depth: Expert, quick

Summary

CLIMB (Controllable Longitudinal brain Image generation via state space based latent diffusion model) is a new framework designed for modeling temporal changes in brain structure, specifically for generating high-quality longitudinal brain MRI scans. It utilizes a baseline MRI scan and its acquisition age as primary inputs, further incorporating conditional variables such as projected age, gender, disease status, genetic information, and brain structure volumes to enhance anatomical change modeling. Unlike traditional latent diffusion models that use computationally expensive self-attention modules, CLIMB integrates a Mamba-based state space model architecture to significantly reduce computational overhead while maintaining high-quality image synthesis. The framework also features a Gaussian-aligned autoencoder that extracts latent representations conforming to prior distributions without the sampling noise found in conventional variational autoencoders. Evaluated on the Alzheimer's Disease Neuroimaging Initiative dataset, comprising 6,306 MRI scans from 1,390 participants, CLIMB achieved a structural similarity index of 0.9433, outperforming existing methods.

Key takeaway

For Computer Vision Engineers developing medical imaging solutions, CLIMB offers a more computationally efficient approach to generating longitudinal brain MRI scans. You should consider integrating Mamba-based state space models and Gaussian-aligned autoencoders into your generative frameworks to improve performance and reduce resource demands, especially when modeling temporal anatomical changes for prognosis and treatment planning.

Key insights

CLIMB uses Mamba-based latent diffusion and a Gaussian-aligned autoencoder for efficient, high-quality longitudinal brain MRI generation.

Principles

• State space models reduce computational cost.
• Gaussian-aligned autoencoders minimize sampling noise.

Method

CLIMB models brain evolution using a baseline MRI and age, conditioned by projected age, gender, disease, genetics, and brain volumes, leveraging a Mamba-based latent diffusion model and a Gaussian-aligned autoencoder.

In practice

• Generate synthetic longitudinal brain MRI scans.
• Aid early intervention for neurological diseases.

Topics

• CLIMB Framework
• Longitudinal Brain Imaging
• Latent Diffusion Models
• Mamba Architecture
• Gaussian-aligned Autoencoder

Best for: Computer Vision Engineer, AI Scientist, Research Scientist, Machine Learning Engineer

Related on AIssential

• CLIMB: Controllable Longitudinal Brain Image Generation using Mamba-based Latent Diffusion Model and Gaussian-aligned Autoencoder — CLIMB uses Mamba-based latent diffusion and a Gaussian-aligned autoencoder for efficient, controllable longitudinal brain MRI generation.
• MambaBack: Bridging Local Features and Global Contexts in Whole Slide Image Analysis — MambaBack combines Mamba and Gated CNNs with spatial sampling and chunking for efficient WSI analysis.
• U$^2$Mamba: A Two-level Nested U-structure Mamba for Salient Object Detection — U$^2$Mamba employs a two-level nested U-structure with multiscale Mamba U-blocks for enhanced salient object detection.
• PIMSM: Physics-Informed Multi-Scale Mamba for Stable Neural Representations under Distribution Shift — Anchoring model memory to physical timescales via spectrum-derived parameters enhances robustness to distribution shifts.
• Coarse-to-fine Hierarchical Architecture with Sequential Mamba for Brain Reconstruction — CHASMBrain's dual-stream Mamba and coarse-to-fine fMRI encoding architecture achieves high predictive accuracy and reveals visual cortex functional specialization.
• IEEE Transactions on Neural Networks and Learning Systems, Volume 37, Issue 2, February 2026 — This issue advances neural network theory and application across diverse domains, from medical imaging to robotics and trustworthy AI.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.