Standardizing CT data with BIDS: Applications in Lung and Brain Imaging

Area of Science:

• Medical Imaging Informatics
• Radiology
• Artificial Intelligence in Medicine

Background:

• Computed Tomography (CT) imaging is increasingly used in fields like neuroradiology and thoracic imaging.
• Standardization of CT data is crucial for developing and validating Artificial Intelligence (AI) tools in medicine.
• Existing data structures like the Brain Imaging Data Structure (BIDS) are vital for neuroimaging but need extension for CT.

Purpose of the Study:

• To demonstrate the extension of the Brain Imaging Data Structure (BIDS) to accommodate Computed Tomography (CT) data.
• To facilitate the development and validation of AI tools for medical applications using standardized CT datasets.
• To promote Open Science principles by making CT data interoperable, accessible, and reusable.

Main Methods:

• Conversion of existing CT datasets (OASIS-3 and National Lung Screening Trial - NLST) into the BIDS format.
• Development of a BIDS App for lung cancer risk prediction utilizing the Sybil AI tool.
• Implementation of a framework for data standardization and AI model validation.

Main Results:

• Successful conversion of OASIS-3 and NLST datasets to BIDS format.
• Demonstration of a functional BIDS App for AI-driven lung cancer risk prediction.
• Establishment of a standardized framework promoting data sharing and AI model validation.

Conclusions:

• The extension of BIDS to CT data provides a standardized format for medical imaging data.
• This framework supports the development, validation, and deployment of AI tools in clinical settings.
• Enhanced data accessibility and reusability accelerate AI research, improve diagnostic accuracy, and aid clinical decision-making.