AIDA — AI & Data Analytics group in Radiation Oncology, Mayo Clinic, Rochester, MN.
Launched in January 2025, our multi-disciplinary team pioneers the rapid translation of AI innovations into real-world patient care, delivering high-impact solutions that reshape radiation oncology.
Projects
The PSA Control Tower creates a unified dashboard that uses AI to predict prostate cancer recurrence sooner. It is one of the first Mayo Clinic projects to use Mayo Clinic Platform (MCP) data. Presidential Strategic Impact Fund award winner and VIBE AI award finalist.
The Daily Dose is an LLM-generated email sent to all physicians at RadOnc, enterprise-wide, providing context and history summaries for all appointments in the day. VIBE AI award finalist.
Key collaborator: Jason Holmes, Ph.D.
The OncoTimeline is an interactive visualization platform that unifies the longitudinal cancer care journey within a single dashboard. It integrates structured treatment records, free-text clinical notes, and dynamic biomarker trends to support rapid review of complex oncology histories. The interface includes an embedded AI chat assistant that enables natural language queries.
Contouring CTVs for spine SBRT is a time-consuming process. Existing AI tools segment bones but do not generate customizable, guideline-based CTVs. This project creates the first spine CTV auto-segmentation model.
Key collaborators: Dalton Griner, Ph.D., Anne Rajkumar, M.D., John Lucido, Ph.D.
This project uses LLMs to generate automatic treatment summaries for patients. Each summary is sent into Epic, is editable, and must be signed off by a clinician before being filed.
This project develops clinically actionable AI predictions to support head & neck cancer treatment decisions. These predictions will improve accuracy in staging, guiding treatment intensity, and predicting patient-specific toxicity.
Patients receiving treatment at Mayo can have prior radiation therapy at outside institutions. Not flagging this can lead to severe issues. This project uses large language models to automatically read large amounts of outside notes and flag prior radiation therapy.
A key concern when deploying models such as the Mayo auto-segmentation model is performance drift due to changes in imaging standards or changes in how users interact with the model. This project creates an automated monitoring system that computes metrics such as Dice score and reports them in a dashboard.
An enormous amount of data exists in free-text notes and cannot be efficiently queried or searched. This project uses LLMs to extract this data and make it accessible, allowing large-scale statistical studies without manual chart reviews.
This project develops a neural network for predicting optimal dose in proton patients at both the plan and field level.
Radiotherapy treatment planning often requires multiple rounds of back-and-forth between physicians and dosimetrists regarding trade-offs. This project develops a novel AI method that generates an entire range of plans so that clinicians can explore the predicted trade-offs in real time.
The Mayo Clinic GenAI program has created STRAND, a large-scale genomics foundation model. This project leverages STRAND to predict clinically significant prostate cancer from germline genomics data.
Key collaborator: Shant Ayanian, M.D.
Team
Publications
Federico Mastroleo; Mariana Borras-Osorio; Shiv P. Patel; Sarah Peterson; Renthony Wilson; Mohammad Javad Namazi; Mi Zhou; Satomi Shiraishi; Andrew Y.K. Foong; David M. Routman; Mark R. Waddle · 2026
Federico Mastroleo; Mariana Borras-Osorio; Shiv P. Patel; Sarah Peterson; Renthony Wilson; Mi Zhou; Satomi Shiraishi; Andrew Y.K. Foong; David M. Routman; Mark R. Waddle · 2026
In the Spotlight
