Large-scale real-world validation of an AI-driven ultrasound breast cancer screening system in rural settings

This is a demo dataset sampled from the data used in the paper "Large-scale real-world validation of an AI-driven ultrasound breast cancer screening system in rural settings". It contains 871 breast ultrasound video clips randomly sampled from the original multicentre screening study, classified into three BI-RADS-based risk levels with approximately balanced class distribution. This dataset is intended for research use in developing and benchmarking AI models for ultrasound-based breast cancer risk stratification.