RadVoice
Voice dictation and reporting for radiology departments.
A radiologist presses the microphone and speaks. The words appear in a live editor as they talk, and they can stop, click anywhere, fix a word, and carry on dictating from that point. When the dictation is finished, RadVoice can lay the report out under standard headings — reorganising what was said, never rewriting it — and the radiologist reads it, edits it, and signs it.
RadVoice is a departmental system rather than a single-user dictation tool. A technologist enters the study, a supervisor assigns it, and the assigned radiologist reports it. Every study has an owner, a status, and a timestamped history from entry to signature.
What it does
- Live dictation into a real editor. Text arrives as it is spoken. The editor stays fully usable throughout — type, delete, reposition, resume.
- Verbatim and structured reporting. Keep the dictation exactly as spoken, or have it arranged under a chosen heading set. Clinical content is never reworded, added to, or removed.
- Worklist and assignment. A study inbox for supervisors, a filterable worklist for radiologists, and a single-owner rule so two people can never open the same study.
- Full report lifecycle. Draft, preliminary, and final states, with send-back for correction. A signed report is immutable; changes afterwards are attributed addenda.
- Patient view. Every study and report for a patient in one timeline, so priors are one click away during dictation.
- Templates and vocabulary. Reusable heading sets, plus departmental and personal word lists that teach the recogniser local names, drugs, and anatomy.
- Search. Full-text search across report bodies with filters on modality, site, radiologist, date, and status.
- Complete audit trail. Every create, edit, generation, signature, and addendum is written to an immutable log, and the dictation audio is archived alongside it.
- Connection awareness. The network, the server, and the speech service are monitored separately, so a dropout during dictation is reported clearly instead of failing silently.
- Light and dark themes, a dense worklist built for long shifts, and a layout that scales correctly across screen sizes and display scaling settings.
Principles it is built on
The radiologist is the author.
The system never finalises a report on its own, never invents or rephrases clinical content, and never measures or infers anything from images — every number in a report comes from the radiologist's own dictation.
Deployment
RadVoice runs on a server inside the hospital network or on a private cloud instance, with its own database and scheduled backups. Speech and structuring can be handled by approved external services under reviewed data-processing terms, or kept entirely on local hardware — the choice is a configuration setting, not a different product.
Get in touch