Diffusion Models for Computational Neuroimaging: A Survey

TL;DR

Diffusion-based approaches offer a principled way to handle data scarcity, noise, and high dimensionality in neuroimaging by learning flexible priors that support generation, reconstruction, and cross-modal translation across MRI, fMRI, CT, DTI, EEG, and PET. The paper surveys foundational formulations (DDPM, Score-SDE, DDIM, Latent Diffusion Models) and conditioning strategies (classifier-free vs inference-time guidance), then maps these to eight neuroimaging tasks with domain-specific design choices. It provides a structured taxonomy of task-related variations (starting point, conditioning, generation target) and summarizes representative works across data generation, reconstruction, super-resolution, cross-modality translation, neural disorder diagnosis, tumor segmentation, visual decoding, and speech decoding, accompanied by a public repository. The discussion identifies open challenges and directions—representation learning, privacy-preserving federated computation, integration with foundation models, and causal inference—that will shape how diffusion models advance practical neuroimaging research and clinical applications.

Abstract

Computational neuroimaging involves analyzing brain images or signals to provide mechanistic insights and predictive tools for human cognition and behavior. While diffusion models have shown stability and high-quality generation in natural images, there is increasing interest in adapting them to analyze brain data for various neurological tasks such as data enhancement, disease diagnosis and brain decoding. This survey provides an overview of recent efforts to integrate diffusion models into computational neuroimaging. We begin by introducing the common neuroimaging data modalities, follow with the diffusion formulations and conditioning mechanisms. Then we discuss how the variations of the denoising starting point, condition input and generation target of diffusion models are developed and enhance specific neuroimaging tasks. For a comprehensive overview of the ongoing research, we provide a publicly available repository at https://github.com/JoeZhao527/dm4neuro.

Figures (1)

• Figure 1: An overview of diffusion models for neuroimaging. Based on fundamental diffusion formulations and conditioning mechanisms, variations on denoise starting point, input condition and generation target are developed for different neuroimaging data and tasks.