A typical tumor board packet covers imaging from the most recent modality, prior pathology reports, current labs, active medications, treatment history, and relevant NCCN or ESMO guideline sections. At most institutions, a nurse coordinator assembles this by hand — opening four to seven different systems, downloading files, copying values. That process takes three to five hours per patient.

What we built

Fanoni Lab’s Tumor Board Pre-Read service uses seventeen specialized agents running in parallel. The imaging agent fetches DICOM studies via DICOMweb. The pathology agent parses structured data from the LIS and unstructured text from the report. The guideline agent queries our NCCN and ESMO index against the confirmed diagnosis and stage. Every claim links back to its source — the imaging series ID, the pathology accession number, the lab result timestamp.

The whole pipeline runs in under twenty minutes for a complex case. The output is a structured FHIR DocumentReference that writes back into the EHR with every source linked.

Results from the first pilot

At our first pilot site, the coordinator’s prep time dropped from 4.1 hours per patient to 22 minutes. Oncologists reported reading more of the packet before the meeting — because it was shorter, more structured, and the citations let them trust the summary rather than re-pulling source documents. The board starts on time now. At twelve oncologists blocking two hours every week, that is not a small thing.