The Brain Imaging and Neurophysiology Database: BINDing multimodal neural data into a large-scale repository

Area of Science:

• Neuroscience
• Medical Imaging
• Data Science

Background:

• Neuroimaging research is limited by data scale and diversity.
• Existing repositories often lack multimodal integration and comprehensive clinical metadata.

Purpose of the Study:

• To introduce the Brain Imaging and Neurophysiology Database (BIND), a large-scale, multimodal clinical neuroimaging repository.
• To provide a comprehensive resource for advancing neuroimaging research, machine learning, and multimodal brain studies.

Main Methods:

• Integrated de-identified neuroimaging data (MRI, CT, PET, SPECT) from multiple institutions.
• Utilized Bio-Medical Large Language Models to extract structured clinical metadata from brain reports.
• Linked imaging data with existing EEG and polysomnography recordings.

Main Results:

• BIND comprises 1.8 million brain scans from 38,945 patients, spanning all ages and diverse neurological conditions.
• Extracted structured metadata for 84,960 brain reports, categorized into standardized pathology classifications.
• The database is BIDS-organized and available in NIfTI format.

Conclusions:

• BIND represents a significant advancement in neuroimaging research by offering unprecedented scale, diversity, and multimodal integration.
• This freely accessible resource will accelerate discoveries in clinical neuroscience by facilitating large-scale studies and machine learning applications.