The Digital Mammography DREAM Challenge: Spurring the Sharing of Large, Sensitive Data in a Data Challenge

Abstract

The Digital Mammography (DM) DREAM Challenge was an ambitious data science challenge to improve the detection of cancer using DM screening data. A critical dependency for launching and hosting this challenge was the identification of data partners who had access to large sets of clinically annotated, mammography images, who would also allow this data to be shared in a challenge. We describe a creative solution to the problem of sharing sensitive and large-scale data in a public challenge by utilizing cloud-based compute and virtualization technology.