# Tumor Board AI

Tumor Board AI is a free, public-beta oncology clinical decision-support tool that
simulates a multi-specialist tumor board using a multi-agent AI architecture. It is
built and maintained by Fabrice M'Vondo.

> IMPORTANT: Tumor Board AI is an informational tool, NOT a medical device and NOT a
> substitute for professional medical judgment. Always confirm decisions with a
> qualified oncology team. Full disclaimer: https://tnmstage.com/DISCLAIMER

## What it does
- AJCC TNM staging (8th and 9th editions) for 12 cancer types.
- Multi-agent consensus treatment recommendations (medical, surgical, and radiation
  oncology perspectives).
- Clinical trial matching against a curated evidence corpus.
- Drug-drug interaction screening.
- Chemotherapy dosing calculations.

## Cancer types supported
BLADDER, BRAIN, BREAST, CERVICAL, MELANOMA, MYELOMA, NSCLC (non-small cell lung),
OVARIAN, PROSTATE, TESTICULAR, THYROID, UTERINE.

## Evidence corpus
The system retrieves from a curated, versioned corpus of 732 sources (917 indexed
passages), comprising:
- 259 clinical trials
- 73 FDA drug labels
- 375 PubMed abstracts
- 25 NCCN guidelines

## Architecture
A coordinator agent orchestrates specialist sub-agents (medical oncologist, surgical
oncologist, radiation oncologist, pathologist, radiologist). Each specialist reviews
the case and the shared evidence corpus; the coordinator synthesizes a consensus.
Reasoning is performed by a local LLM (Microsoft Phi-4-mini-reasoning) served via
llama.cpp.

## Key endpoints (public, informational)
- Web app: https://tnmstage.com/
- Disclaimer: https://tnmstage.com/DISCLAIMER
- API health: https://tnmstage.com/api/v1/health
- OpenAPI docs: https://tnmstage.com/docs

## Founder
Fabrice M'Vondo — founder. LinkedIn: https://www.linkedin.com/in/fabrice-m-vondo-56b17522