MR-RATE-Atlas: 10K-100K Spatially Normalized Brain MRI Volumes

MR-RATE-atlas provides between 10,000 and 100,000 atlas-registered brain MRI volumes where imaging sequences are spatially normalized to a standard atlas-space. Developed by Forithmus as part of the MR-RATE vision-language foundation model project, it supports multimodal medical AI research. The data includes 3D medical imaging paired with language-based tasks such as visual question answering and radiology report generation.

Use Cases

• Training vision-language models to generate text descriptions from 3D brain MRI volumes
• Performing zero-shot classification on spatially normalized imaging sequences
• Developing visual question answering systems for radiology using atlas-registered coordinates

Strengths

• Standardized spatial normalization across all imaging sequences
• Large scale for medical imaging with 10,000 to 100,000 records
• Designed for multimodal vision-language tasks in radiology

Limitations

• Geographic bias likely restricted to US-based data as per region tags
• Atlas registration may remove individual anatomical nuances present in native-space imaging
• Non-commercial license restricts industrial application

Provenance

Source

Forithmus, MR-RATE project

Collection Method

Atlas-registration and spatial normalization of existing MRI sequences

Freshness

Last updated March 2026.

Geography

United States