About the hackathon
The MedAI Hackathon was Boston University's inaugural clinical-AI event at the Duan Family Center for Computing and Data Sciences, themed Transform Precision Medicine with AI. Teams got real datasets across three biomedical challenges, plus GPU access and mentorship over a two-day build.
The track we competed in
We took the early-stage lung cancer phenotyping challenge, which handed teams H&E whole-slide images from lung resections and biopsies and asked for a model that predicts microscopic vascular invasion, a signal that guides staging and aggressive-growth risk.
What we built
VI-LUAD predicts vascular invasion in lung adenocarcinoma directly from whole-slide images. It tiles each slide into patches and runs a 25-model ensemble to flag invasion, trading a single fragile model for a vote that holds up better on unseen slides.
What we won
We placed 1st in the early-stage lung cancer phenotyping track.