Structured reasoning for attendings · Active teaching for residents

Real-time radiology reasoning, during dictation

Speak your findings and get a ranked differential with rationale before you finish the case. No tab switching.

Free during beta · Private by default: dictated audio and transcripts are never stored

You dictate
“Hypervascular liver lesion in a cirrhotic l
RadReason asks
Which feature best fits the lesion?
washoutcapsulediffusion restrictionhepatobiliary retentionnot sure
Differential updates
HCC rises substantially — arterial enhancement plus washout in cirrhosis strongly favors hepatocellular origin.
Modes

One tool, across the whole read.

Prime before you open the case. Quick Diff and Learn Mode while you work through it. Ask whenever a reference question comes up. Pearls keep what you want to remember.

Speed

Quick Diff

Dictate, get a ranked differential, move on. A second perspective for routine reads without breaking flow.

  • Instant ranked differential with rationale
  • No questions, no interruptions
How Quick Diff works →
Reasoning

Learn Mode

Targeted discriminator questions refine your differential as you dictate, the way an attending would push back on rounds.

  • You answer, the differential updates
  • 1–3 teaching pearls per completed case
How Learn Mode works →
Pre-read

Prime

Give it the indication before you open the images and get a short approach brief. Commit to your pre-test probability, then read.

  • The question, the mimics, and the review areas readers skip
  • Never sees your images — it teaches the approach
How Prime works →
Reference

Ask

A question comes up mid-read, you ask by voice or text and get a direct answer with key points and pitfalls in seconds.

  • Grounded in LI-RADS, Bosniak, Fleischner, O-RADS
  • Save any answer as a Pearl, one tap
How Ask works →

Most AI tools give answers. RadReason helps you work through the case — a reasoning scaffold, not a replacement for radiologist judgment.

Mic-first, built for dictationPrivate by default — nothing storedDictate in English, French, Spanish, Portuguese, or GreekModality-agnostic
More detail

How RadReason fits your workflow

What RadReason helps with

RadReason is an AI radiology assistant designed to support imaging interpretation in the moment, not after the read is done. It listens to your dictated findings and surfaces structured differential diagnoses while you are still in front of the case.

Speak the imaging findings you actually see — RadReason maps them to ranked, reasoned differentials. The output is structured, traceable, and meant to surface possibilities you may not have prioritized, never to make the diagnosis for you.

  • Generate differential diagnoses from imaging findings
  • Support real-time decision-making during radiology dictation
  • Refine complex cases with targeted discriminator questions in Learn Mode
  • Ask radiology reference questions by voice — criteria, pitfalls, and differentials mid-read
  • Collect teaching pearls from cases and Ask queries, browsable in your Pearls tab
  • Strengthen structured diagnostic reasoning
For residents and fellows

Learn Mode turns passive reading into active reasoning: you articulate a differential, answer discriminator questions, and see how each answer shifts the ranking. Completed cases generate short teaching pearls you can review later on the Pearls page — turning passive reading into active reasoning for residents, fellows, and any radiologist who wants the workflow tool to double as a teaching tool.

For attendings and practicing radiologists

Quick Diff gives a structured, traceable second perspective on routine reads without breaking flow. The output is meant to surface possibilities you may not have prioritized, never to make the diagnosis for you.

RadReason is shaped around radiology workflow specifics: modality-agnostic, mic-first, and structured around how dictation actually happens. It is a radiology reasoning assistant — not a general medical AI tool, and not a replacement for radiologist judgment.

Frequently asked questions

Is anything I dictate stored?

No. Dictated audio, transcripts, and Ask queries are processed in memory and never stored.

Is this a diagnostic tool?

No. RadReason is for educational and decision-support use only, not a substitute for clinical judgment.