AI
auto_awesomeAI-poweredAI woven through the workflow — not bolted on.
RadFuzion puts AI where it saves you time: write routing rules in plain English, ask your dashboard and logs questions in chat, search studies by describing them, and let RadSR draft impressions and findings. It’s optional, it’s configurable, and it never sees a patient’s identity.
No PHI is sent to AI. Period.
Before any prompt leaves your environment, RadFuzion replaces the accession number — the only potentially identifying value in the request — with a random token (like ACC-TKN-K7Q2M9XW4RTB). No patient names, MRNs, or dates of birth are ever included. The token-to-accession mapping stays inside your database, and AI responses are mapped back locally.
- check_circleAccession numbers tokenized before every request
- check_circleNo names, MRNs, or DOBs in any prompt
- check_circleToken↔accession mapping never leaves your database
- check_circleBring your own provider — or run a fully local model so nothing leaves your network at all
Work in plain language
Talk to your router
The tedious parts — building rules, hunting through logs, querying studies — become a sentence.
Routing rules in plain English
Describe a routing rule the way you’d say it out loud — “send all outside CTs to the cloud archive and the reading PACS” — and RadFuzion drafts the rule for you to review and save.
Dashboard assistant
Ask your dashboard questions in chat — “how many exams completed in the last hour?”, “what’s the failure rate today?” — and get a straight answer instead of digging through tables.
Troubleshooting assistant
Ask about your logs — “why did this accession fail to forward?” — and get a direct explanation with the relevant events, so you fix issues in seconds, not minutes.
Natural-language study search
Type what you’re looking for instead of filling out query fields. RadFuzion turns it into a proper DICOM query across your sources.
AI reporting
From images and dictation to a draft report
RadSR and RadRead use AI to get radiologists to a signable draft faster — in their own voice.
Smart Impression
Generate an impression from the findings — in the individual radiologist’s own style, learned from their prior reports.
Smart Findings
Turn free dictation into clean, structured findings sections, ready to review and sign.
Report templates
Build a modality- and exam-appropriate preliminary report template on the fly, so every read starts from the right scaffold.
OCR → structured reports (RadSR)
RadSR uses AI to turn OCR’d report text into a structured report (SR). From there, structured reports — dictated or built from OCR — can auto-fill form fields in your PACS or dictation platform.
Read the RadSR field guidearrow_forwardSmarter automation
Decisions made for you — when you want them
Routing and retrieval decisions that stay under your control and your confidence thresholds.
Intelligent routing
After your manual rules run, AI can suggest a destination for anything left unmatched — gated by a confidence threshold you set, so it never overrides your rules.
AI prior-study selection
When a new order arrives, RadFetch uses AI to pick the relevant priors worth prefetching — the right comparisons, without dragging everything.
Read the RadFetch field guidearrow_forwardYour model, your rules
Bring your own AI provider
Point RadFuzion at the provider you trust — OpenAI, Anthropic, or Azure OpenAI — or run a fully local model with LM Studio or Ollama so no data leaves your network at all. AI features are optional and off until you enable them.
Talk through your optionsarrow_forwardSee the AI on your own workflow
We’ll show you the natural-language rules, the assistants, and the reporting AI running against a setup like yours — with the PHI safeguards in place.
mailBook a walkthrough