Your real-time radiology reasoning partner.

1. 1

   Prime the approach, before you open the case

   • Enter the indication and exam type, by voice or text.
   • You get a short approach brief: the primary target, what else presents this way when the history is incomplete, and the review areas that get skipped.
   • Prime never sees the images. It reasons about the clinical question, the read stays yours.
   • Optional on every case. Skip straight to findings when you don't need it.
2. 2

   Dictate your findings

   • Speak naturally during your read.
   • Works alongside your existing dictation setup.
3. 3

   Choose your mode

   Quick Diff

   • Instant ranked differential.
   • Designed for fast workflow use.
   • No questions asked of you — but you can push back on the result.

   Learn Mode

   • Guided reasoning with targeted discriminator questions.
   • Includes board-style teaching questions.
   • Designed for deeper thinking and learning.
4. 4

   Ask, whenever a question comes up

   • A reference question mid-read — imaging criteria, staging, follow-up, pitfalls — asked by voice or text without leaving the case.
   • Answers are structured: direct answer, key points, pitfalls, and a calibrated confidence level.
   • Grounded in the current versions of the criteria systems. Save any answer as a Pearl in one tap.
5. 5

   Refine your reasoning

   • In Learn Mode, targeted discriminator questions help you narrow the differential.
   • In Quick Diff, Talk back lets you answer the result: disagree with the ranking, add a finding you spotted later, or ask what you are missing. You get a direct reply and a strip showing exactly what moved and why.
   • It holds its position when you are wrong, and says what evidence would change its mind.
   • Externalize, refine, and stress-test your reasoning — not just receive an answer.

• Active reasoning, not passive lookup — built to externalize the differential, not just generate one.
• Targeted discriminator questions — narrow the differential the way experienced radiologists actually reason.
• Workflow-native learning — deliberate practice during real cases, mic-first and modality-agnostic.
• Attending-style reasoning support, not replacement of radiologist judgment.

RadReason sits next to your worklist, not in front of it. Reach for it when a case actually warrants a second perspective — ignore it the rest of the time.

• Use alongside your existing dictation software.
• Toggle on for complex or uncertain cases.
• Skip it for routine findings.
• Acts like a second set of eyes when you need it.

Learn Mode asks you these questions unprompted. Quick Diff does not — but you can still push back on its differential with Talk back once you have it.

RadReason is a dictation sidecar for radiology — it runs alongside dictation and reasoning, never in front of it. It is not autonomous and augments, rather than replaces, the radiologist. The radiologist remains the interpreter of record on every case.

• Reads do not start in RadReason — they start in your dictation system. RadReason listens alongside it.
• RadReason does not produce a report and does not finalize a diagnosis. It returns structured reasoning the radiologist can use, edit, or ignore.
• There is no autonomous mode. Every output is generated only from what the radiologist explicitly dictates into a case.
• Final interpretation, recommendation, and report wording always remain with the radiologist.