Feature list
Everything RadFuzion does
Every shipped feature on one page, from DICOM routing and HL7 to reporting, structured results and fleet management. Search it, filter it or print it.
- features
- 288
- areas
- 15
- work with AI off
- 248
- new in 6.8
- 24
Current with RadFuzion 6.8.3. AI features are optional and off until you turn them on.
288 features in 15 areas
Receive & route
RadFuzion accepts studies from any modality or PACS and sends each exam where it belongs, using rules you can see, test and change.
DICOM C-STORE receiver
Accepts studies from any modality, PACS or gateway over standard DICOM C-STORE, with C-ECHO verification and a configurable AE title, port and thread count.
Extra listening ports
Listen on more ports under the same AE title for separate feeds. Port changes apply without a restart.
Automatic firewall and port-conflict alerts
On Windows, opens the firewall for the receiver ports itself. If a port is already in use, it warns you in the UI and sends an alert.
Exam completion detection
Waits until a study stops arriving, after a timeout you set, so exams go out whole and not image by image.
Hold for HL7 order
Hold incoming images until the matching HL7 order arrives, so downstream systems get images and order together.
Group exams by accession or Study UID
Tracks and routes each exam by accession number (the default) or by Study Instance UID, whichever fits your sites.
Tag-based routing rules
Route on any DICOM tag with equals, contains, starts with, greater or less than, exists and missing. One condition can list several values.
Route by Calling or Called AE title
Rules can match on the AE title that sent the exam or on the RadFuzion AE title it was sent to, so one server can handle several feeds differently.
Priorities, stop on match and multi-destination routing
Drag rules into priority order. Each can send to several destinations and stop the rules below it.
Scheduled (time-window) rules
Limits a rule to set hours and days of the week in any time zone, including overnight windows such as 22:00 to 06:00, for after-hours or weekend routing.
Drop rules
Discards matching traffic, such as a test feed or a duplicate stream, and marks it as dropped instead of leaving it looking unrouted.
Companion objects follow the exam
An exam's SR, KO, PR and secondary captures go wherever its rule sends it, even when the rule filters on modality.
DICOM tag changes
Set, fill in when blank, copy or remove tags, with references to other tags and fallbacks. The viewer, HL7 and resends all see the change.
Lookup tables
Rewrite a tag from a table pasted from a spreadsheet, such as turning a code a tech typed into the right Study Description.
Required-tag validation and Fix Tags
Holds exams missing required values such as patient name, DOB or MRN, raises a Dashboard alert, and lets staff fix the tags and release the exam.
Study UID merger and SUID Keeper
Merge a study that arrives under several Study UIDs, and keep late images under the original UID, so the PACS never splits the exam.
Generate a new Study UID or a missing accession
Give an exam a new Study UID, or a numbered accession when it arrives without one, used everywhere the exam goes.
Exam change history
Each exam's Changes tab and Transfers labels record what was changed and how: tags modified, accession generated, Study UID merged, anonymized, converted and more.
Per-rule processing switches
A rule's Processing tab turns on RadSR, DICOM to HL7, RadMatch, SR to ORU and prior fetching for exactly the exams it matches.
Per-rule notifications
A rule can send its matching exams' events to chosen Slack, Google Chat, email or webhook channels.
Create a rule from an exam
Right-click an exam in Transfers to start a rule from its header values (modality, institution, description, AE titles), with no patient details.
Create rules with AI
Describe a routing rule in plain English. The AI drafts it in a short conversation and shows its assumptions, and you review it before saving.
Rule management tools
Clone rules, import and export them as JSON, enable or disable all at once, and see a warning on any rule that depends on a setting that is off.
Test rules (routing simulator)
Shows which rules a real exam or made-up tag values would match, where it would go, and every tag change, lookup and transcode, without sending anything.
Destinations, forwarding & cloud
Send exams to any PACS, cloud store or folder, with forwarding that compresses, retries and waits out downtime so nothing is lost.
DICOM C-STORE destinations
Sends to any PACS, workstation or AI node over DICOM, with an optional Calling AE per destination for PACS that accept only certain callers.
Cloud and DICOMweb destinations
Send to Google Cloud Healthcare, Azure DICOM, any DICOMweb (STOW-RS) server, Google Cloud Storage, AWS S3 and local or network folders.
Bulk Google Cloud import
Creates Google Cloud destinations from service-account JSON files in one step, using your saved defaults for project, location and dataset.
Connection tests that say why
Test or Test All checks each destination for real, with C-STORE, C-FIND or cloud permissions, and gives the reason for any failure.
Skip prefetched priors per destination
Lets a destination take routine traffic but not the priors RadFetch pulls in, so an archive doesn't get duplicates.
Parallel, tuned forwarding
Sends several exams at once in size-capped batches, with timeouts that scale to the payload and per-modality settings for mammography and ultrasound cine.
Transcoding and compression
Re-encodes to JPEG 2000, HTJ2K, JPEG Lossless, JPEG Baseline, RLE or uncompressed on the way out, and falls back safely when a codec won't round-trip.
Durable retries with clear failure reasons
Retries survive restarts, and every failure gets a plain reason that says whether it is temporary or needs a fix.
Destination-down hold and replay
Destinations are checked around the clock. Exams for a down destination wait and go out the moment it recovers.
Stuck-forward recovery
Restarts any exam stuck mid-send after a crash or restart, sends each exam in one pass at a time so nothing goes twice, and lists stuck forwards on the Dashboard.
Image cache with pinning
Keeps recently forwarded exams on disk within size, age and free-space limits, for resends, rule tests, RadRead and the viewer.
Live Transfers board
Every exam's status by destination, with live progress, transfer syntax, compression and change labels, searchable across all history.
Exam actions
Resend or force-forward an exam, cancel a send, re-run OCR or SR, pin it in the cache, delete it, open it in the viewer, or retry every exam no rule matched.
Manual DICOM upload
Drag in files or a whole folder from a workstation and send them straight to one destination, or through the routing rules if you choose.
PDF to DICOM upload
Turns a PDF, such as an outside report or requisition, into a DICOM Encapsulated PDF with the patient and exam details you enter, and sends it to a destination.
DICOM tools & Query/Retrieve
Find any study on any archive, move it where it needs to go, and fix, merge, split, convert or de-identify it on the way.
Search every archive from one screen
Search DICOM PACS, DICOMweb, Google Cloud Healthcare, Google Cloud Storage, Azure DICOM and local storage at once, or pick the sources.
Flexible patient search
Find exams by accession, MRN, patient name in any format (partial, comma or caret), date of birth, date range, modality and description.
Duplicate archive results merged
When an archive returns the same study several times, it is listed once and flagged so the archive owner can be told.
Image counts on demand
Get Image Counts looks up how many images each listed study has before you pull it.
Push studies from one system to another
Select studies and send them straight to any destination in the transfer syntax it needs. RadFuzion keeps that destination through retries and resends.
Send to RadFuzion and process with rules
Retrieve studies into RadFuzion and let your routing rules decide where they go, instead of a single target.
Live push activity
A Q/R Push Activity panel shows each study's incoming, forwarded and expected images and its status while it moves.
Rerun or reprocess a remote study
Right-click a result to rerun it through the routing rules or reprocess its OCR or SR.
Plain-language C-MOVE errors
When a retrieve fails, RadFuzion explains the likely cause (AE title, port, permissions) instead of showing a raw DICOM status code.
Per-source AE titles and throttling
Give each query source its own calling and Move Destination AE titles, and cap simultaneous C-FIND and C-MOVE threads per source.
Query/Retrieve server for the image cache
Approved workstations can find and pull cached studies from RadFuzion with C-FIND, C-MOVE or C-GET during PACS or cloud downtime. Off by default.
Download a study as a DICOMDIR package
Download any found study as a standard DICOMDIR package that any DICOM viewer or PACS can import.
Tag Editor
Fix the accession, MRN, birth date, Study UID or any tag for one image, a series or a whole study, then push it back or on to another system.
Replace a study in Google Cloud
For Google Cloud Healthcare, which won't overwrite, the Tag Editor deletes the old copy and pushes the corrected one as a tracked job.
Edit tags before a push
From Query/Retrieve, open a study in the Tag Editor and correct it before it goes to the target system.
Merge two exams
Combine two exams that arrived split into one study under the chosen master's accession, Study UID and patient details.
Split series into a new study
Move selected series out of an exam into a new study with its own accession, Study UID and description.
Anonymization profiles
Remove, replace, hash, blank or date-shift any tag with reusable profiles, assigned per routing rule. Originals are never changed.
Enhanced MR/CT to classic images
Split Enhanced multi-frame MR and CT into classic single-frame images for older PACS, CD viewers and AI, per rule and on every disc.
DSA subtraction before forwarding
Renders mask-subtracted XA/XRF runs as a new series any viewer can show. Send them alone or alongside the originals.
Global search (Ctrl+K)
One header search box finds exams, activity, log entries and pages, with recent exams shown before you type.
Pick a recent exam to test with
Every test box that takes an accession offers recent exams to pick from, with no patient names shown.
One-click copy of accession and MRN
Click any accession or MRN in Query/Retrieve, the Tag Editor, Merge/Split, the viewer and other tables to copy it.
HL7, worklist & results
RadFuzion speaks HL7 in both directions: it takes orders, serves the modality worklist, builds orders for image-first sites, and turns final reports into ORU results.
HL7 MLLP listener
Receives HL7 orders from your RIS or interface engine on a port you choose (2575 by default) and acknowledges each message automatically.
Configurable HL7 field mapping
Choose which HL7 fields hold the accession, MRN, patient, exam, schedule, station, priority, reason for exam, allergies and alerts.
Order cancels handled
A canceled order comes off the worklist and stops any prior fetch already running for that accession.
HL7 messages log
A searchable list of every order received, failed or canceled, showing each extracted field next to its HL7 path and the raw message.
HL7 listener health at a glance
An HL7 icon in the header and a warning banner show right away when the listener is stopped, in error or can't open its port.
Durable order history
Each order's original HL7 is kept by accession after it leaves the worklist, and pruned on your retention schedule.
DICOM Modality Worklist server
Modalities query RadFuzion for scheduled exams (MWL C-FIND) built from your HL7 orders, including allergies and alerts.
Worklist access control
Only modalities that call the right worklist AE title, and optionally only those on an allowed list, can query it.
Per-station worklists
Each scanner sees only its own modalities and the orders meant for it, by station AE or station name, with an optional strict mode for unmapped stations.
Auto-clear when images arrive
An order leaves the worklist as soon as its images arrive, so modalities only see exams still to be done.
Worklist orders page
View, filter and search worklist orders, show only today's, and add, edit or delete orders by hand.
Worklist connection log
Every modality query is recorded with its AE titles, IP address, outcome and result count, so worklist problems are quick to troubleshoot.
HL7 orders to placeholder studies
Create a placeholder Secondary Capture study, with your logo, from each HL7 order and send it to the destination named in the order.
Placeholder history and regenerate
Review recent placeholder studies and regenerate any of them with one click.
DICOM to HL7 orders
For image-first sites with no order feed, build an HL7 ORM from each exam's images, per routing rule, so the worklist, RadSR and RadRead still work.
Order lands before images
Sends the generated order to MLLP, HTTP(S) or S3 destinations and holds image forwarding briefly so the receiving system has the order first.
Manual ORM send and send log
Send an ORM for any accession on demand, test each destination, and review every ORM sent.
Per-rule HL7 field changes
A routing rule can change any field of the HL7 order it sends, from text, a DICOM tag or another field, by overwriting, appending or filling blanks.
Basic Text SR to ORU
Final reports that arrive as DICOM SR become HL7 ORU results for your RIS, from a routing rule, for priors or by hand. Included with every license.
Recognizes report SRs
Tells a radiologist's report apart from measurement, dose and CAD SRs, so only actual reports become results.
Correct result status, never sent twice
Unsigned reports go out as preliminary, signed ones as final and changed ones as corrections. The same report is never sent twice.
ORU profiles and delivery
Profiles set which SRs are sent, the message layout and the targets (MLLP, Amazon S3 or Mirth REST), with a test send.
ORU field mappings with preview
Set any ORU field to text, a DICOM tag or another HL7 field, and preview the result on a real exam before saving.
ORU sent log with automatic retries
A Sent tab lists every SR handled (sent, skipped and why, or retrying) with its message and Send again. Unreachable targets are retried for up to an hour.
SR to OT for PACS without SR support
Also send report SRs as Secondary Capture pages in the same study, labeled non-diagnostic, for a PACS that cannot display SR.
Reporting (RadRead)
RadRead is a reading and dictation station for downtime and preliminary reads, with templates, voice, pick lists, report checks and delivery built in.
Downtime mode switch
One visible switch decides whether signed preliminary reports are delivered (downtime) or the module runs in a safe test mode where nothing is ever sent.
Radiologist reading worklist
Orders, Unread and Completed tabs show what needs reading, image availability, cache status, delivery status and who has each exam open.
Worklist intake filters and sandbox mode
Limit RadRead to STAT and stroke orders, leave out modalities, or keep a rolling window so a standby station never builds a backlog.
Read exams without an HL7 order
Exams that arrive with images but no HL7 order get a reading order built from their DICOM headers. A real HL7 order replaces it automatically.
Automatic image check and manual matching
RadRead finds each order's images in the image cache and connected archives, and staff can match a study by name, MRN or birth date.
Pre-cached images that open instantly
Unread exams are downloaded and prepared on the server ahead of time, and the local image cache is read first, so studies open immediately.
Viewer on a second monitor
The DICOM viewer opens in its own window, moves itself to the second screen, and follows along as the radiologist goes from exam to exam.
Report editor with templates and fill-in fields
Each exam opens on a structured template with bracketed fields to tab or speak through, plus undo and redo, formatting, cleanup and autosaved drafts.
Report template library
One template per exam description, learned from the headings of incoming final reports and editable in Settings. Works with AI off.
On-premises speech recognition (MedASR)
Medical speech recognition runs on the RadFuzion server by default, so no dictation audio leaves the site. Several radiologists can dictate at once.
Cloud speech and live dictation
Optional OpenAI transcription for low-CPU servers, with a site word list. Live dictation shows each phrase as it is spoken.
SpeechMike and PowerMic support
Philips SpeechMike, Nuance PowerMic and foot controls work over USB in Chrome or Edge, with hold-to-talk and buttons for sign and next field.
PowerScribe-style voice commands
Say next field, go to impression, select a phrase to correct it, scratch that, insert template, spoken punctuation, new paragraph or pick two.
Keyboard shortcuts and push-to-talk
Mic toggle, push-to-talk on any key, field navigation, Smart Findings, Smart Impression and sign all have shortcuts, with noise suppression and gain controls.
Trained words
Teach RadRead to fix words the recognizer keeps mishearing, by hand or, with AI on, from suggested sound-alike variants.
Pick lists
Each radiologist's most-used sentences for the current section appear under the line. Click, press Alt+1 to 6, or say pick two. Needs ORU Style Training.
Live report checks
While dictating, RadRead flags a wrong side, anatomy that does not match the patient's sex, findings-impression conflicts, units and missed follow-up.
AI review before signing
At Complete, AI rereads the report for a wrong side or sex, a mismatch with the reason for exam and misheard words. It can be turned off.
Unviewed series warning
Signing warns when some series were never scrolled through in the viewer, with Complete Anyway available.
Smart Findings
Dictate findings freely in any order into one box, and AI places each one in the right section of the report, with the changes highlighted.
Smart Impression in the radiologist's own style
AI drafts the impression from the findings in that radiologist's own style, learned from their past final reports through ORU Style Training.
ORU Style Training
A separate HL7 port collects each radiologist's signed final reports from the RIS by NPI, to drive style-matched impressions, pick lists and templates.
Auto-complete from final reports
When the RIS sends back a final report for an exam still unread in RadRead, it is marked completed automatically so nobody reads it twice.
CXR Draft
AI drafts chest X-ray findings from the views on screen, rewritten in the radiologist's voice, into Smart Findings for review and never directly into the report.
Downtime CXR pre-reads
During downtime, every chest X-ray on the Unread list is drafted in the background before anyone opens it, with a daily limit on AI spend.
Ultrasound measurement extract
Draw a box around an ultrasound measurement panel in the viewer, and AI reads the values into the report with full anatomical labels.
Sign and next, skip exam
Sign and jump straight to the next unread exam with the viewer following, or skip an exam back to the queue, from the screen, keyboard or mic.
Exam locking for multiple readers
An exam open by one radiologist shows as in use to others, stale locks clear themselves, and admins can force-unlock.
Addendum and correction
Completed reports are read-only. Radiologists can add an addendum and admins can make a correction, both sent with the right result status.
HL7 ORU delivery of reports
Signed reports go to the RIS or EHR as HL7 ORU results over MLLP, signed with the radiologist's name and NPI, with a test-send button.
Fax delivery by facility
Signed reports are faxed as PDFs through RingCentral or any SIP fax provider to each facility's own fax number, with a test fax and a delivery log.
Grace period, recall and retries
Delivery waits 30 seconds after signing so a report can be recalled or edited, then retries failures, with Retry now and Resend on the worklist.
Branded PDF reports
Each preliminary report renders as a PDF with your facility's header, a preliminary-report notice and the signer, downloadable from the worklist.
Radiologist roster and role
Keep radiologists with credentials and NPI, import them from CSV, and give readers a Radiologist role with no access to routing or settings.
Report retention and purge
Completed preliminary reports are purged automatically after a set number of days, with a storage check and a manual purge.
Structured reports (RadSR)
RadSR reads ultrasound, DEXA, echo, OB and dose measurements from SRs and worksheets and delivers them as results, with AI or with learned profiles that need no AI at all.
Measurements read from DICOM SR
Reads ultrasound, DEXA, echo, OB and CT dose SR measurements, with laterality and the chosen sample, and recovers values from non-conformant SRs.
OCR for image-only worksheets
Reads measurements off screenshot-only ultrasound and DEXA worksheets, keeping table rows and left/right columns together, then AI structures them.
Per-rule SR/OCR control
Routing rules decide which exams RadSR reads, OCR only chosen series or only when there is no SR, and pin a specific workflow per facility.
Native RadSR workflows
A built-in measurements-to-results pipeline, no n8n needed: match exams by description, modality and source, exclude others, and run in test mode first.
AI maps measurements to your forms
AI reads SR and OCR measurements and fills your form fields in each field's unit, ready for HL7, REST or Fuzion Reporting.
Field maps from anywhere
Define form fields by CSV upload, the in-app editor, a Google Sheet, any URL, a central RadFuzion or the customer portal, refreshed only when they change.
Smart form selection
Picks the right form for each exam, keeping exams to forms of their modality. AI picks are remembered, and you can pin a form per description.
Plain-language adjustments
Tune extraction in plain English, such as "treat mass and nodule as one field", and compare results with and without it side by side.
Describe a workflow
Describe a workflow in a sentence and AI drafts it in test mode for review. It never fills in hosts, ports or credentials.
DICOM header workflows (no AI)
Send any DICOM header value, or mammography views (RCC, LMLO and so on), straight into the HL7 message. These are deterministic and never counted toward RadSR usage.
Fill in a missing study description
When an exam's DICOM has no study description, AI works one out from the OCR or SR text so the exam still finds its form.
Automatic mode: learns, then runs without AI
RadSR learns a mapping profile per scanner and form from AI results. After 5 matching exams it maps them with no AI call, spot-checking about 1 in 50.
Profiles only (fully AI-free)
Map exams only with saved profiles and never call AI. Exams from unknown scanners are skipped, or optionally sent to AI.
What Automatic would do
Workflows still on AI only keep learning in the background and show 30-day numbers for how many exams profiles would have handled.
Visual mapper
Drag SR measurements onto form fields, take suggestions by name, or learn from past AI results. Set the unit, decimals and default per field.
Smart tools and derived values
No-code unit conversion, totals, averages, ranges, if/then and formulas, with presets like T-score class, calcium score, CAC-DRS and LVEF.
AI-written smart tools, verified
AI writes a tool for a field only it was filling, such as a calcium score category, and the tool is kept only if it matches every past AI value.
AI on top of a profile
An optional small AI step adds risk categories or a short impression to empty fields, following your instructions. It never changes measured values.
Share mapping profiles between sites
Share profiles through a versioned RadHub library, or export and import them as files. Sites can also pull form fields from one central hub.
HL7 results output
Adds results as OBX/OBR to the exam's HL7 order and sends it by MLLP, Mirth REST, Amazon S3 or the worklist, with background retries.
REST API output with vendor formats and FHIR R4
Post each exam's values as JSON to any web address, as RadSR JSON, FHIR R4 Observations or a custom template, with a live preview.
Fuzion Reporting integration
Uses Fuzion Reporting Fillsets live as forms and pre-fills them by accession, so the values are in the report when the radiologist opens it.
Webhook delivery
Sends structured JSON results to your integration engine, with per-exam status, payload inspection and one-click resend.
Test delivery and test mode
Preview results without sending anything, and test each output and HL7 source against a recent exam before going live.
View SR and View Result
See each exam's SR as a readable report, text or XML, and its result grouped into sections with headline values and where it went.
How each result was made
Every result says whether a profile or the AI filled it, what was left empty, and how it compared with the AI on spot checks.
Never silently lose an exam
Interrupted exams resume after a restart, failures land in a retryable dead-letter list on the Dashboard, and any exam can be reprocessed.
Honest RadSR report counts
Usage stats count only reports with at least two real result values. Cover pages, empty SRs and header-only workflows are never counted.
Priors (RadFetch)
RadFetch pulls each patient's relevant prior exams from your archives before the radiologist opens the case, triggered by HL7 orders or by the images themselves.
HL7 order-driven prefetch
When an order arrives, RadFetch finds and pulls the patient's relevant priors from your PACS, so comparisons are waiting at read time.
Use DICOM Data (no HL7 needed)
Starts a prefetch from an exam's own DICOM tags through a routing rule, with tag profiles and a try-it-on-an-exam test.
Prefetch rules
Rules choose priors by description keywords, patient details, look-back window, how many per exam type, and newest or oldest. Ready-made defaults ship with it.
Facility-specific rules
Facility profiles give each sending site its own prefetch rules, with optional fallback to the global rules.
AI prior selection
AI reviews the patient's history and the reason for exam and adds the most relevant priors, with identifiers removed and a shadow mode to compare first.
AI-only prior rules with guidance
A RadFetch rule can leave prior choice to AI alone, with your own written guidance and a maximum number of priors.
Multiple sources and migrations
Pull from up to six archives: spread the load, fall back in order, or combine all and pull each study from the first that has it.
Gentle on your PACS
STAT orders go first, per-source limits cap load, duplicate orders are ignored, and priors already at the destination are skipped.
Complete-retrieval checks and auto-retry
Confirms every image arrived, waits for late images, retries incomplete jobs automatically, and recovers jobs whose images showed up later.
Stalled-retrieval alerts and source health
Alerts when an archive accepts a retrieval but never sends, and monitors each source's reachability with polite back-off.
Job tracking and retry anytime
Every job shows what was found, moved and received per study. Retry or cancel any job, or use Retry All Failed and Retry Validating.
Manual fetch
Search a source by MRN, name, birth date, description or date range and pull chosen studies on demand.
Prefetch insights
Success rate, average time to fetch, top failure reasons and per-rule results over 7 to 90 days.
Fetched-study webhook
Calls your web address for each prior fetched, with exam details filled into the URL or header, retries and manual resend.
Prior reports to the reading system
Reports that come back with retrieved priors as SR are sent on as HL7 ORU, so radiologists see the prior's report too.
Study normalization (RadMatch)
RadMatch turns every facility's free-text study descriptions into your standard exam codes, locally and in milliseconds, and writes them into DICOM and HL7.
Study description to exam code matching
Maps free-text study descriptions from any facility to your exam codes locally in milliseconds, with an adjustable minimum score.
Code sets with RadLex terms
Import code sets from CSV, switch the active set, and accept the RadLex license terms when a set includes RadLex content.
Mappings and term overrides
Edit learned mappings and add term overrides that persist across retrains, and test any description in Match Test.
AI fallback with review queue
When the local match is unsure, AI suggests a code. Low-confidence picks wait in a review queue. Works with OpenAI, LM Studio or Ollama.
AI-generated training data
Auto Bootstrap has AI write example descriptions for each code when a code set is uploaded, so matching works well from day one.
Apply to DICOM and HL7
Writes the matched description into forwarded exams (all, or per routing rule) and into incoming HL7 orders.
RadMatch HL7 Out
Forwards orders with the matched description and exam code written in, to MLLP, Amazon S3 or Mirth REST, with test messages.
Matching dashboard and history
See match totals per hour and day, the active code set, and a searchable history of every match.
Viewer, patient CDs & sharing
A browser viewer with MPR and 3D for preliminary reads and image review, patient discs with a no-install viewer, and secure exam sharing between sites. Non-diagnostic.
Built-in browser DICOM viewer
A Cornerstone3D viewer opens exams from Transfers, the image cache or any Query/Retrieve source, with nothing to install. Non-diagnostic.
Reader and Advanced modes
A lean Reader mode for fast reading, or Advanced for the full toolset, with keyboard shortcuts and a toolbar that adapts to the screen width.
Measurement and annotation tools
Length, angle, Cobb angle, bidirectional, circle, ellipse, rectangle, freehand, spline and livewire ROIs, probe/HU, arrows and labels.
Reading layouts and linked scrolling
1 to 10 viewport layouts, scroll-all and linked scrolling, localizer lines, scale overlay, magnify, invert, rotate and flip.
Modality-aware window presets
One-click window/level presets for CT, MR, PET, ultrasound and more, matched to the series on screen.
Cine and tomosynthesis
Cine up to 60 fps with loop or back-and-forth playback. Breast tomosynthesis streams frame by frame for smooth scrolling.
MPR with MIP, MinIP and thick slab
Axial, coronal and sagittal reconstructions for CT and MR, with crosshairs, a 3D cursor, slab thickness and MIP, MinIP or average blending.
3D volume rendering
3D views with tissue presets, cinematic lighting, a crop box, opacity curve, fly-through, spin, 360-degree video and PNG snapshots.
Spine labeling and key images
Label vertebrae starting from the region in the series description and jump every viewport to a level, and mark key images for a key-images-only view.
Readable structured reports with ORU send
SR reports show as readable text in the sidebar, and Preview ORU and Send ORU turn a final report into HL7.
Presentation states and DSA
Applies saved presentation states (GSPS) from the modality and shows angiography subtraction (DSA) images correctly.
Scanned documents as PDF
Download a scanned-document or secondary-capture series as a PDF, or every OT and eDoc in the study as a ZIP.
Tags panel and in-viewer tag editing
See every DICOM tag and the compression in use, and edit a tag for one image, the series or the whole study.
Compressed image support
Displays JPEG 2000, HTJ2K and color images directly in the browser.
Burn exams to CD, DVD or Blu-ray
Burn one or many selected exams to a single disc, with the size and recommended media shown first. Uses the RadFuzion Burn companion app.
Portable viewer on every patient disc
Discs include a no-install, non-diagnostic viewer for Windows and Mac. Images load into memory only, leaving no patient data behind.
Share an exam to another site
Send an accession and its images to a partner RadFuzion over HTTPS, with no DICOM ports opened between sites.
Receive, route and resend shared exams
Received exams can run through your routing rules and stay in a short-term repository for resending or pushing to any destination.
Trusted connections and share history
Trust partner sites by system certificates or a pinned certificate, and track every inbound and outbound share.
Fleet & remote (RadHub)
RadHub watches and runs every RadFuzion site from one place, over RadHub Connect when there is no VPN, without storing patient data on the hub.
Fleet overview
See every site on one page as cards or a table: exams failing, in flight or complete, plus CPU, memory, disk, send queue, version and a Needs attention list.
Fleet-wide exam list
One list of every exam that arrived or changed at any site in the last 24 hours, with filters and a search across every site's full history.
Fix exams at any site from the hub
Open an exam from any site to see where it went and which rules matched, then Resend, Stop, Reprocess, Pin or Delete it from the hub.
Fleet statistics
Volume, busy hours, failures and reasons, forwarding time, modalities and top senders, for all sites or one, over 24 hours to 30 days.
Fleet-wide RadSR results
Every site's RadSR results in one searchable list, with whether each result went out or failed and the full result.
Cross-site activity trail
An Activity tab shows the activity logs of every site together, so you don't have to sign in to each server.
No patient data stored on the hub
The hub asks each site for fresh exam lists and stats every time and keeps none of it. Fleet searches are recorded in the Access Log.
Remote control of any site
Pick a site in the header and use its RadFuzion screens from the hub. The header turns blue while you work remotely, and each site can opt out. The viewer is left out to save bandwidth.
Manage each site's rules, destinations and settings
Edit one site's routing rules, destinations and settings from the hub, test its destinations, and compare its configuration with the hub's.
Site actions from the hub
Open a site's UI, view or refresh its config, restart its service, purge its database, or check for updates, from a menu on each site.
Update the whole fleet
See each server's version and pending update, and update one server or all of them, including sites on RadHub Connect.
Updates for sites without internet
RadHub keeps signed updates and sends them to sites on a VPN or closed network. Each site checks the signature before it installs.
Usage reporting for offline sites
RadHub carries the signed usage reports of sites that can't reach radfuzion.com and brings back a signed receipt, so nobody uploads usage files by hand.
RadEngine7 in RadHub
See the health, channels and alerts of each site's paired RadEngine7, and restart channels, turn them on or off, or retry failed messages from the hub.
AI through the hub
Sites whose network blocks AI providers can send their AI requests through the hub, which forwards them without ever seeing or supplying an API key.
Full-window and wall display mode
Open all of RadHub in its own window for a second monitor or a wall screen. The selected tab stays in the address, so you can bookmark it.
Connect sites without VPN or open ports
Hub, sites and RadEngine7 reach each other through RadFuzion's relay using outbound HTTPS only, so there's no VPN, port forwarding or fixed IP to set up.
End-to-end encrypted relay
Traffic is encrypted end to end between your own servers, and the relay can't read it. The relay only links servers on the same customer license.
One-time connect codes
Join a site to the hub by pasting a code that works once and expires after 7 days. Removing a site cuts it off at once.
Keeps working if the relay is down
If the relay can't be reached, each server keeps working locally and reconnects by itself. Direct addresses keep working alongside the relay.
Operations & reliability
RadFuzion shows what needs attention, tells you why something is set up but won't run, alerts the right channel, and gives you the tools to fix it.
Operations dashboard
Live exam flow by stage, throughput, success rate, processing time, receiver status, destination health, failure hotspots and disk space on one screen.
Needs-attention cards with one-click fixes
Failed or stuck forwards and failed OCR/SR jobs are listed on the Dashboard, where you can retry, cancel or force-complete them.
"Set up but won't run" checks
When a rule or setting depends on something that is off, RadFuzion says where and why, and links to the fix. The Dashboard lists them all.
System warning banners
Banners warn about low or full disk, port conflicts, an HL7 listener that is down, held exams with missing tags, and a license in its grace period.
Service and worker control
Restart the whole service from the browser, or see and restart each background worker (receiver, forwarder, OCR/SR, RadFetch and others) one at a time.
Settings that save themselves
Switches save as soon as you flip them and ask first when they change how exams are handled. Other edits wait in a save bar.
Full backup and restore
One click downloads a ZIP with the settings, encryption key and database. Restore puts it back safely at the next restart.
Alerts to Slack, Google Chat, email and webhooks
Send failed forwards, system errors, held exams, stalled prior fetches and more to your channels, each with its own filters and a test button.
Notification rules
Get alerted when an exam matches conditions you choose, such as DICOM tag values, destination or called AE title, on the channels you pick.
Backlog alerts
Get alerted when the processing queue stays too deep or the AI error rate stays too high, with thresholds you set.
Recent alert history
Warnings and errors are listed at the top of the Notifications page, with repeats grouped into one counted row.
Unified exam log
Type an accession to see its whole story in order: receipt, rule matching and every send to every destination.
Activity logs, routing decisions and raw logs
Filter the activity log by date, event or accession, see why each rule matched or was skipped, read raw service logs, and export audit logs to CSV.
Processing status
Live OCR/SR and AI processing health: queue depth, active workers, throughput, timing for each stage, and dead-lettered jobs you can retry.
Compression test
Test whether an exam's images convert to a given transfer syntax before you send them to a destination that needs it.
RadFuzion Sidebar for Chrome
A free Chrome side panel next to your PACS or RIS with the dashboard, live transfers, exam search with Resend, RadSR results and failure alerts.
Security & compliance
Role-based access, a HIPAA access trail, encrypted secrets and built-in HTTPS keep patient data in the right hands.
Users and role-based access
Built-in Admin, Tech, Radiologist and RadRead Tech roles plus custom roles. A user reaches only what their role grants.
Bulk user and radiologist import
Import users and radiologists from CSV, and link user accounts to radiologists by NPI.
Secure sign-in
Passwords are stored hashed with PBKDF2-SHA256, default accounts must change their password before use, and repeated failed sign-ins lock out the source address.
HIPAA access log
A user access trail of sign-ins, patient data views, searches, exports and admin changes, kept for its own retention period and never purged with routine data.
Configuration change audit
Every settings change is recorded in the audit log.
Encrypted credentials
API keys, passwords, destination credentials and notification secrets are encrypted at rest with a key unique to each server, and they never reach the browser.
Built-in HTTPS for the web UI
An optional HTTPS listener for the web UI and API, with an auto-generated certificate for your hostnames and an option to turn off plain HTTP completely.
Custom login branding
Upload your organization's logo for the RadFuzion sign-in page.
AI (optional)
Every AI feature is optional and off until you turn it on, works with the provider you choose, and strips patient identifiers before anything leaves the server.
AI is optional, with one switch
With AI off, routing, RadSR profiles, report checks and everything else keep running on fixed logic, and AI buttons stay hidden until AI is set up.
Choice of AI provider, cloud or local
Use OpenAI, Anthropic (Claude) or Google Gemini, or run on-premises with LM Studio or Ollama. A few features need specific providers.
Per-task model choice and testing
Pick a model for each job (study descriptions, OCR, SR, rules, reporting, vision), reset to recommended models, and test every model in one click.
One-file AI setup
Import a signed provider file from RadFuzion that sets the provider, key and recommended models without anyone seeing the key.
Editable AI prompts with a test bench
Reword the instructions behind report templates, Smart Impression, Smart Findings and study-description inference, then test them on sample data.
Reliability and spend controls
Timeouts, retries with backoff on rate limits, parallel-call limits, input-size caps and daily limits keep AI dependable and costs predictable.
Remembered AI answers
Repeat questions, such as which form or exam code a study description gets, reuse earlier AI answers instead of paying again.
Accession numbers never sent to AI
Accession numbers are swapped for random tokens before a prompt leaves the server and restored in the answer.
Patient identifiers removed from AI text
Names, MRNs and birth dates are scrubbed from free text, dictated reports and OCR text before AI sees it.
No AI on patient search terms
Query/Retrieve searches are never sent to an AI provider, since search terms can include patient details.
AI Activity log
Every AI call is listed with time, feature, exam, model, status, latency and token use, including cached tokens, for cost and compliance review.
Troubleshooting AI assistant
Ask in plain words why exams failed or stalled. The assistant reads recent logs, queues and echo history, or one accession's trail, to answer.
Ask the Dashboard
Ask questions like "How many CTs completed today?" or "What's our failure rate?" and get direct answers from live metrics.
AI alert explanations and wording
An on-demand AI Report explains an alert and its related logs, and channels can have AI word alerts plainly. Patient details are never sent.
Deployment, updates & licensing
RadFuzion runs on Windows or Linux, updates itself safely with signed releases, scales to PostgreSQL, and is licensed by module with no hard cutoffs.
One-click in-app updates
The sidebar shows when a new version is out. Click it and RadFuzion downloads, installs and restarts in 1 to 2 minutes, with no installer or admin prompt.
Automatic updates in a maintenance window
Choose the days and hours, and RadFuzion installs new versions by itself in that window. It never retries a version that failed.
Signed releases and automatic rollback
Only releases signed by RadFuzion are installed, and only newer versions of the same edition. If the new version doesn't stay running, the old one is put back.
Offline update upload
Servers without internet can upload the installer and its signature file in Settings and install it the same way.
Release notes inside the app
Settings › Updates and Help › Release notes show what changed in each public release, and the sidebar version links straight to them.
PostgreSQL database option
Stay on the built-in zero-setup database, or switch to PostgreSQL (16.x recommended) for high volume, with a connection test and one-click data migration.
Built for high volume
Parallel OCR/SR workers, parallel AI calls with automatic retry, crash recovery and a dead-letter queue keep processing steady as volume grows.
Runs on Windows or Linux
RadFuzion runs on Windows or 64-bit Linux, the same product either way.
Offline license activation
A signed license key works without internet. Online servers pick up renewals and module changes about every 12 hours.
Module licensing and server binding
The license turns on only the modules you bought, counts seats, and can be locked to specific machines. RadHub, RadShare and RadEngine7 come free with every license.
Grace period, never a hard cutoff
If a license lapses, RadFuzion keeps running through a grace period (14 days by default) and shows a banner, so clinical work isn't cut off suddenly.
Automatic usage reporting
Usage reports with daily counts and no patient data are signed and sent by themselves. Air-gapped sites can download the report and upload it instead.
Guided walkthroughs
New installs show short guided tips on each main page and a tour of every Settings section, which can be replayed or turned off.
Built-in help center
Searchable help inside the app for every module, with setup steps, recommended settings and security guidance.
HL7 interface engine (RadEngine7)
RadEngine7 is a Mirth-style HL7 interface engine that comes with every RadFuzion license, built so channels can be made and changed without scripting.
Full HL7 v2 engine
Channels with an inbound, a filter, a transformer and any number of outbounds, for HL7, JSON and plain-text messages.
Inbound connectors
Receive over MLLP, HTTP, folder pickup, S3, SFTP or FTP/FTPS, or from another channel, with ACK modes and duplicate-message dropping.
Outbound connectors and reliable delivery
Send over MLLP, HTTP/Mirth, folder, S3, SFTP/FTP or as a PDF report. Each destination has its own in-order queue, retries and throttling.
Drag-and-drop channel editor with live testing
Build channels on a flow chart that runs your sample message as you edit, names every HL7 v2.5.1 field and shows exactly what each system will get.
Visual field mapper
Map inbound fields onto the layout a receiving system expects, with suggested mappings and an option to send only the fields you choose.
No-code transformation steps
Set, copy and clear fields, translate codes with lookup tables, convert dates, renumber OBX, call web services and branch with if/then/else.
Order memory
Remembers values or segments from an order by accession and adds them to its later result, with a Memory page to view and clear them.
Sandboxed script step
JavaScript for the rare job the built-in steps can't do, sandboxed with no file, network or system access and strict time and memory limits.
Channel templates and saved messages
Save a vendor interface as a reusable template with site details stripped out, and start new channels from it, with no AI needed.
Message log with replay and resend
Search by accession, MRN or any text. See what came in, what each destination got and its reply, then replay, resend, cancel or download messages.
High-throughput, no-loss message storage
Every message is stored before it is acknowledged, so a power cut loses nothing. In our tests it took over 400 messages a second.
Dashboard, alerts and monitoring
A per-channel dashboard plus alerts by email, Teams, Slack, Google Chat or webhook for down listeners, failing sends, backlogs and disk.
Channel history and undo
Keeps the last 10 saves of every channel, with one-click undo, restore and recovery of recently deleted channels.
AI channel assistant with PHI masking
Ask in plain English for a channel change and get tested builder steps. Sample messages are masked first, and nothing goes live until a person saves.
Explain this channel
AI writes a plain-English outline of what a channel does and flags things worth a second look, like steps that never run.
Regex from a description
Describe a pattern in words ("everything after the second comma") and AI writes the regular expression, tested on your example.
Import from Mirth
Rebuild a Mirth channel or group export as script-free builder steps and lookup tables, checked field by field against your sample messages.
Vendor specs read by AI
Keep vendor spec sheets with each channel and ask them questions, such as "what goes in OBR-4?", with cited answers.
Safe AI access (MCP and API keys)
Outside AI tools connect over MCP or an OpenAPI-described API with masked or unmasked keys. They can only read, test and propose, and a person approves every change.
Interface Kit for your partners
A standalone offline page your interface partners use to tag sample messages and fill in a spec sheet, which you then import straight into a channel.
SQLite or PostgreSQL storage
Pick the data folder and run on SQLite or PostgreSQL. Moves and database switches happen safely at the next restart.
Users, roles, audit log and encrypted secrets
Viewer, editor and admin roles, a six-year audit log, shared credentials kept once, and every password and key encrypted at rest.
Automatic backups
Daily backup zips (14 kept) with Back up now and restore. When paired, RadFuzion also keeps the last 30.
Pairs with RadFuzion
One-click sign-in from RadFuzion, engine health on its dashboard and in RadHub, shared license and AI settings, and port-conflict warnings.
See it on your own exams
Click through the real interface in the demo, or tell us about your sites and we will show you the parts that matter to you.
RadFuzion feature list, current with release 6.8.3. 288 features. radfuzion.com/features