Simulating Longitudinal Brain MRIs with Known Volume Changes and Realistic Variations in Image Intensity

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Biology

Background:

• Longitudinal Magnetic Resonance Imaging (MRI) is crucial for tracking brain changes.
• Simulating realistic longitudinal MRI data with known volume changes, particularly for Alzheimer's disease (AD) atrophy, is challenging.
• Existing methods lack the ability to accurately reproduce intensity variations and artifacts seen in real-world longitudinal scans.

Purpose of the Study:

• To present a novel simulator for generating large databases of visually realistic longitudinal brain MRIs.
• To enable the simulation of known brain volume changes, specifically focusing on atrophy patterns in AD.
• To provide researchers with an open-source tool for developing and validating brain morphometry methods.

Main Methods:

• The simulator utilizes a biophysical model of brain deformation due to AD-related atrophy.
• A novel approach combines deformation fields from the biophysical model and non-rigid image registration.
• This combined field generates new images with specified atrophy and realistic intensity variations, mimicking real longitudinal data.

Main Results:

• The simulator successfully generates visually realistic longitudinal MRIs with controlled volume changes.
• The method effectively reproduces realistic intensity variations, including noise independence and acquisition artifacts.
• The resulting simulated databases offer ground truth volume changes for tool development.

Conclusions:

• The developed open-source simulator facilitates the creation of tailored MRI databases for research.
• This tool is expected to advance the development of more robust brain morphometry tools.
• The platform encourages further research into advanced models of brain deformation and atrophy generation.