Thera Soft, software development
FREN
EventProduct launchOctober 2026 · 4 min read

Voxia, the TsXcare voice assistant, makes its debut at JFR 2026

From 8 to 11 October, at the Palais des Congrès de Paris, THERA SOFT presented Voxia on its stand at the Journées Francophones de Radiologie, the annual French-speaking radiology congress. Voxia is an assistant that controls the report editor by voice, structures reports according to reference standards and follows up patients after the exam.

The right report. The right follow-up. The launch film shown for JFR 2026.

For its first public presentation, Voxia was shown where it lives: in the TsXcare report editor. No separate software, no extension to install. The radiologist holds down Ctrl+Space or the back button of their SpeechMike, waits for the beep, speaks and releases: Voxia carries out the command.

What visitors saw on the stand

  • A patient file opened by voice: “Start dictation for the file of…”, then the exam’s template report loaded with a single sentence.
  • Diagrams that fill themselves in: the PI-RADS prostate map, with a sector marked and labelled by voice, then inserted into the report.
  • Mammography in a few clicks: a normal screening in four clicks, insertion included, with the ACR BI-RADS v2025 standard.
  • Proofreading before signing: laterality, organ, terminology, and comparison with previous reports.
  • Patient follow-up: recommended follow-up examinations become milestones, and those that have not taken place are raised as alerts.

Six structured reporting modules

Voxia comes with six modules that follow current reference standards: mammography (BI-RADS), prostate MRI (PI-RADS v2.1), thyroid ultrasound (EU-TIRADS and ACR TI-RADS), pulmonary nodules (Fleischner 2017 and Lung-RADS v2022), cystic renal masses (Bosniak 2019) and adnexal masses (O-RADS). Each one opens by voice, is filled in by clicking on a diagram, and checks that the report is complete before insertion.

In-house speech recognition

Voxia’s speech recognition is developed by THERA SOFT. It runs on dedicated servers with HDS-certified health data hosting: no audio or dictated text is logged, and no data is entrusted to a third-party dictation service.

The radiologist remains the author

A dictated sentence is never mistaken for a command. Voxia does not analyse images and makes no diagnosis; any text it writes remains flagged until the radiologist validates it. The classification support functions shown on the stand were presented as a prototype, not CE marked: they are going through qualification as a medical device before any commercial release.

What’s next?

Voxia is entering a pilot phase at a first imaging centre, before being gradually opened up to TsXcare users. Would you like to see it on your own reports? Write to us at contact@therasoft.fr.