Voice dictation for radiology reporting: a practical guide

What is voice dictation in radiology reporting?

Voice dictation in radiology reporting is the practice of speaking a report aloud and having software convert the speech into editable text. The radiologist reviews the images, speaks their observations and conclusions, and then checks and signs the resulting document.

Historically, dictation was recorded on tape or digital audio and typed up by a transcriptionist. Many departments now use speech recognition, where the text appears on screen as you speak. Some setups still mix the two, with a typist checking recognised text before it returns to the radiologist.

Whichever method is used, the principle stays the same: the spoken words are the source, and the written report is a faithful record of them.

• Traditional dictation: audio is recorded and typed by a person later.
• Speech recognition: software converts speech to text in real time.
• Hybrid workflows: recognised text is checked by an editor before sign off.
• Structuring tools: dictated text is organised into report sections for review.

Why do radiologists dictate rather than type?

Radiologists dictate because it lets them keep their eyes on the images while they describe them. Typing pulls attention down to the keyboard, while speaking allows continuous viewing, scrolling and measuring at the same time as the report takes shape.

Dictation also suits the way many radiologists think through a study. Talking through the anatomy in a set order is a familiar habit from training, and it maps neatly onto a spoken report. For long or complex studies, speaking is often more comfortable than typing many paragraphs.

That said, typing still has its place. Short reports, small corrections and situations where speaking aloud is not appropriate are often easier with a keyboard. A good workflow lets you switch between the two freely.

• Eyes stay on the images rather than the keyboard.
• Hands remain free for scrolling, windowing and measuring.
• Spoken review follows a natural, systematic order.
• Typing remains available for quick edits and quiet settings.

How does browser-based speech recognition work?

Browser-based speech recognition uses a speech service built into the web browser to turn microphone audio into text. The web page asks for microphone permission, listens while you speak, and receives the recognised words back from the browser to display on screen.

In some browsers, recognition is handled by an online service run by the browser vendor, which means audio may be sent to that vendor for processing. Support also varies between browsers. At the time of writing, recent versions of Chrome, Edge and Safari offer speech recognition to web pages, while Firefox does not. Typing works in every browser.

Because the service is general purpose, it may not know every medical term. Unusual eponyms, drug names and abbreviations can be misheard, so careful review matters even more.

• Allow microphone access when the browser asks.
• Use a supported browser such as Chrome, Edge or Safari.
• Expect occasional errors with rare medical terms.
• Fall back to typing whenever dictation is not available.

How should you set up your workspace for dictation?

A good dictation setup starts with a quiet space and a decent microphone positioned at a consistent distance from your mouth. Background noise, echo and a microphone that moves around are among the most common causes of misrecognised words.

A headset or a desk microphone with a fixed stand keeps a constant distance from your mouth, which keeps recognition steady. Keep the volume of nearby speakers low, close doors where you can, and avoid dictating while others are talking close by.

It also helps to check your browser settings so the correct microphone is selected. Many reading rooms have several audio devices connected, and recognition quality drops sharply if the browser is listening to the wrong one.

• Choose a quiet room or a quiet corner of the reading room.
• Use a headset or fixed desk microphone where possible.
• Keep the microphone at a steady distance.
• Confirm the correct input device in browser settings.

How do you dictate clearly so the text needs fewer corrections?

Dictate in complete, steady phrases at a natural pace, and speak punctuation and section names clearly. Recognition engines work best with flowing sentences rather than isolated words, and they struggle with mumbled or very rushed speech.

Plan the report briefly before you start speaking. A moment spent deciding the order of your findings often saves several corrections later. Many radiologists follow the same sequence for each type of study, which keeps both the dictation and the finished report consistent.

When you need a specific term, say it fully rather than relying on a local abbreviation. Abbreviations that are obvious to a colleague may be unclear to a referrer and are easily misheard by software.

• Speak in full sentences at an even pace.
• Name each section aloud, for example indication, findings and conclusion.
• Follow a consistent order for each type of study.
• Prefer full terms over local abbreviations.
• Pause briefly between sections to keep the text tidy.

How do you correct dictated text efficiently?

The most efficient way to correct dictated text is to fix small errors with short, specific commands or quick keyboard edits, and to re-dictate only when a whole sentence is wrong. Trying to repair a badly recognised paragraph word by word is usually slower than saying it again.

In Medicate, you can type or say literal commands such as "replace moderate with mild" or "delete small", and you can add a section name, for example "in findings", to limit where the change applies. These commands only swap or remove the exact words you name; they never generate new wording.

Whatever tool you use, read the corrected sentence again afterwards. A replacement can sometimes change grammar or meaning in a way you did not intend.

• Use replace and delete commands for single words or short phrases.
• Limit a correction to one section when the word appears more than once.
• Re-dictate a sentence when most of it is wrong.
• Reread every corrected sentence in context.

What should you check before signing a dictated report?

Before signing, check that every statement in the report matches what you intended to say, with particular attention to laterality, measurements, negations and the conclusion. These are the details where a single misheard word can change the meaning of a report.

Read the report as the referrer will read it, from top to bottom. Make sure the indication reflects the clinical question, the findings are in a sensible order, and the conclusion answers the question that was asked. Compare the draft with your original dictation if anything looks unfamiliar.

Medicate highlights words such as left, right, bilateral and numbers that appear in the structured draft but not in your transcript, so you can check them quickly. The physician still reviews and signs every report.

• Left, right and bilateral are correct throughout.
• Every measurement and unit matches what you dictated.
• Negations such as "no" and "not" have not been dropped.
• The conclusion answers the clinical question.
• The referrer and patient reference are correct.

Where does a tool like Medicate fit into this workflow?

Medicate fits between dictation and sign off: you dictate in the browser, it organises your words into Indication, Findings and Conclusion, and you then review, correct and sign. It reformats your dictation only and is designed never to add findings, measurements or laterality you did not say.

The workflow is a simple wizard with four steps: Dictate, Structure, Review and Sign. You can go back at any point without losing what you entered, choose the referring physician from your own directory, add your signature and download the report as a PDF or Word file with your letterhead. Nothing is emailed by the app.

Medicate structures dictated text; the physician reviews and signs every report. It is not a diagnostic tool.

Frequently asked questions