Frequently asked questions.
Ten questions evaluators ask, answered in plain language and grounded in what ships today.
How is Radial different from in-house scripts?
A script answers one question, once, for the person who wrote it. Radial maintains the layer every script depends on (e.g., standardized names, governed cohorts, and metrics that stay current as cases accrue), so any question is a query, not a project, and the answer carries provenance.
How is Radial different from the tools built into my TPS?
TPS tools see the plan in front of them. Radial works across your whole archive, whatever mix of vendors produced it: every case curated into one database, comparable across systems, sites, and years.
Does my data leave the building?
No. Radial deploys as an on-premise appliance with a browser UI and no cloud dependency. PHI stays inside your institution.
Which vendors' exports are supported?
Radial imports DICOM-RT from any system that exports it — CT, MR, and PET imaging, with RTSTRUCT, RTPLAN, and RTDOSE. It also supports FHIR R4 connectivity in both directions.
How accurate is automatic structure mapping?
Mapping is evidence-based and deliberately conservative. High-confidence matches confirm automatically and are recorded like hand confirmations; everything below confidence is flagged for audit. Nothing files itself silently.
Does the dictionary train across sites?
No. It learns locally, per deployment. Your confirmations improve your institution's dictionary and stay inside it.
What does ambient monitoring actually do?
It runs continuously as new cases accrue, adjusts for case mix, and tracks your metrics on control charts. When a metric drifts, the chart raises an alert for review — you decide what it means.
Can Radial check plans against protocol limits?
Yes. Limit sheets encode a protocol's dose constraints, and cohort cases are checked against them as they accrue, with the numbers behind every call.
How is the DVH engine validated?
The engine descends from dicompyler-core, the public gold standard, and it refuses to guess: when data is incomplete or ambiguous, Radial says so rather than computing past it. Its numbers are validated against known-answer reference cases.
How is Radial licensed?
Two paths: vendors embed, build on the API, or white-label under an OEM license, and academic groups run the full platform under a research license. Both start with a conversation.