US-based Stanford University and Vietnamese startup VinBrain have signed a Data Use Agreement (DUA) to improve the precision of radiology interpretation using a multi-modal “RadGraph” method.

To design a state-of-the-art artificial intelligence (AI)-enabled platform for more accurate diagnosis and treatments, a tremendous amount of geographically distributed data, medical images, and patient medical records are required for system training.

The DUA grants Stanford University and VinBrain the collaborative power to improve efficiency in radiology reports and advance AI in healthcare by sharing relevant de-identified data for model training and testing, while tackling complexity, ambiguity, and limitations in medical image interpretation.

Starting with 240,000 anonymised medical images and report pairs provided by Stanford through this DUA, VinBrain will work to improve the accuracy of the RadGraph method through a research collaboration with Stanford to extract clinical entities and relation annotations from a large dataset of full-text radiology reports.

This agreement promises to comprehensively expand cooperation in the next areas of development, strengthen the relationship between VinBrain and the world's leading University, and take VinBrain further in the research of AI-developed solutions for worldwide patients.