Related Experiment Video
Updated: Jul 27, 2026
17:06
Co-analysis of Brain Structure and Function using fMRI and Diffusion-weighted Imaging
Published on: November 8, 2012
26.4K
ReMiND: Recovery of missing neuroimaging using diffusion models with application to Alzheimer's disease
Chenxi Yuan1,2, Jinhao Duan3, Kaidi Xu3
1Department of Biostatistics, Epidemiology & Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, United States.
Imaging Neuroscience (Cambridge, Mass.)
|August 13, 2025
Summary
This study introduces ReMiND, a novel 3D diffusion model for imputing missing structural MRI data in Alzheimer's disease research. ReMiND accurately reconstructs whole-brain images, improving biomarker development and atrophy rate estimation.
Area of Science:
- Neuroimaging
- Medical Research
- Artificial Intelligence
Background:
- Missing data, due to missed visits or technical issues, is a significant challenge in longitudinal Alzheimer's disease (AD) studies using structural Magnetic Resonance Imaging (MRI).
- This data gap hinders the development of high-quality, imaging-based biomarkers for AD.
- Accurate imputation of missing MRI data is crucial for reliable longitudinal analysis.
Purpose of the Study:
- To introduce a novel 3D diffusion model, ReMiND (Recovery of Missing Neuroimaging using Diffusion models), for imputing missing structural MRI data in AD research.
- To evaluate ReMiND's performance against established imputation methods like forward filling and variational autoencoders.
- To demonstrate ReMiND's capability in generating accurate whole-brain MRI images and estimating atrophy rates.
Main Methods:
- Developed a novel 3D diffusion model (ReMiND) to generate conditional whole-brain structural MRI images.
- The model uses past and/or future observed MRI scans to impute missing data points in longitudinal trajectories.
- Compared ReMiND's performance against forward filling and variational autoencoder-based imputation methods using experimental results.
Main Results:
- ReMiND generates 3D structural MRI with high similarity to ground-truth images.
- Imputed images from ReMiND show lower differences in volume estimation compared to forward filling and autoencoders.
- ReMiND provides more accurate estimations of atrophy rates over time in key brain regions than comparator methods.
Conclusions:
- ReMiND effectively imputes missing structural MRI data at single time points within longitudinal studies.
- The 3D diffusion model outperforms alternative methods in generating whole-brain images and accurately estimating neurodegenerative changes.
- ReMiND offers a promising solution for addressing missing data challenges in Alzheimer's disease neuroimaging research.
