MedAI Hackathon: Tackling Real Biomedical Problems With Artificial Intelligence
Today’s most pressing challenges require an integrated approach across disciplines — a concept Hariri Institute FY26 Focused Research Program (FRP) award recipients Vijaya B. Kolachalama and Jennifer Beane-Ebel embraced in their recent event, the MedAI Hackathon.
Held on April 10 and 11 at the Duan Family Center for Computing and Data Sciences, the MedAI Hackathon drew over 233 registrants and 30 multidisciplinary teams across the BU community. Using relevant biomedical data, the teams consisted of individuals from engineering, computing, data sciences, and medicine who created solutions for genuine clinical problems with the hopes of winning one of nine cash prizes.
“Our goal was to build an intensive, fun, educational, and impactful experience where multidisciplinary teams could tackle genuine clinical problems with relevant biomedical data, learn from faculty and research computing experts, and see how their ideas could contribute to publications or improved patient care,” said Kolachalama. “The overwhelming response confirmed that the BU community was hungry for this kind of convergent event.”
The inspiration for the hackathon began with Kolachalama’s and Beane-Ebel’s Focused Research Program (FRP) “AI-driven Accurate Detection Strategies for Aggressive Early-stage Lung Adenocarcinoma,” funded by Hariri Institute for Computing as part of its signature seed funding program. Calling for expertise across thoracic pathology, computer vision, and bioinformatics, the FRP drew researchers across Boston Medical Center, Chobanian and Avedisian School of Medicine, College of Engineering, and the Faculty of Computing and Data Sciences (CDS).
The FRP sought to develop a biomarker that can predict microscopic vascular invasion in both lung resection tissues and pre-surgical biopsies using digitized hematoxylin and eosin stained slides which are part of the standard pathology workflow. Yet, while the FRP engaged faculty across disciplines and institutions, Beane-Ebel noticed there was an opportunity for students to cross-collaborate more.
“At BU, we have talent across computing, data sciences, medicine, and engineering, yet connecting students with opportunities to work directly with de-identified datasets from active research labs remains difficult,” said Beane-Ebel. “We wanted to create an attractive event for trainees that would be hands-on and exciting at the intersection of cutting-edge AI methods and real-world biomedical challenges.”
The hackathon included test data and a challenge derived from Kolachalama and Beane-Ebel’s FRP, “Early-stage Lung Cancer Phenotyping.” This challenge encouraged participants to use H&E whole-slide images from lung resections and biopsies and model subtle histology patterns linked to vascular invasion and aggressive growth.
The second challenge, “Amyloid PET for Alzheimer’s Disease,” allowed participants to work with 3D amyloid PET scans and tracer labels to predict standardized Centiloid burden.
The last challenge, “Acute Tubular Injury from Proteomics,” challenged teams to predict acute tubular injury from pre-biopsy SomaScan protein profiles and clinical covariates in the Boston Kidney Biopsy Cohort.
All three challenges were drawn from ongoing research within Kolachalama’s and Beane-Ebel’s labs and at the Medical Campus. Teams could tackle as many challenges as they liked, with most teams addressing one or two of the three, using Python, PyTorch, TensorFlow, scikit-learn, JAX, and TerrierGPT on BU-supported stacks where possible.
Research Computing Services dedicated 40 GPUs and over 10,000 remote computing jobs were submitted to the BU Shared Computing Cluster on Saturday alone.
In Saturday’s awards ceremony, teams received 1st ($1000), 2nd ($500), and 3rd ($250) prizes for each challenge. Winners were selected based on quantitative performance metrics calculated on held-out test sets (using standard machine metrics relevant to each challenge).
“This event was highly impressive,” said Yannis Paschalidis, Distinguished Professor (ECE, BME, SE) at the College of Engineering; Founding Professor of Computing & Data Sciences; and Director of the Hariri Institute for Computing. “The collaboration among students from different disciplines paired with expert faculty guidance throughout the challenges cultivated an innovative and creative learning environment—preparing students for future convergent research projects and modern day problem solving.”
Holding big checks branded for the occasion, the winning teams celebrated their hard work as they posted for pictures.
“The hackathon demonstrated the power of bringing together students from nearly every corner of BU to solve real biomedical problems through AI,” said Kolachalama. “It generated real energy around convergent research and showed how student innovation can feed directly into ongoing scientific work.”
Looking ahead, Kolachalama and Beane-Ebel seek to grow this into a larger event, occurring annually or bi-annually — exploring external sponsorships, the possibility of a full hackathon week, and the inclusion of students from other local universities.
“Our goal is to scale while keeping the core focus on building real, actionable solutions so participants can develop ideas that move forward into products,” said Beane-Ebel.
This event was co-sponsored by the Hariri Institute for Computing, Boston University Chobanian & Avedisian School of Medicine, Department of Medicine and Evans Center for Interdisciplinary Biomedical Research, School of Public Health Center for Health Data Science, Center for Brain Recovery, and Information Services & Technology (IS&T). Learn more about the event and its results at the MedAI Hackathon webpage.