Teacher guide

Everything you need to run a 15 to 20 minute session on how AI can help doctors read X-ray, CT and MRI images, for students aged 11 to 14.

Learning goals

By the end of the session, students can:

  • describe how X-ray, CT and MRI make pictures;
  • explain what an AI helper does, and why a person checks it;
  • discuss privacy and fairness in medical AI.

Students look at a real, de-identified scan and guess what they see. Then they watch an AI describe it, twice: in kid words and in doctor words. Finally they compare the AI’s answer with the answer key that came with the image. The comparison is the point. AI is a fast, tireless helper that can be wrong, and a doctor always decides.

Curriculum links. Light and the electromagnetic spectrum; body systems; computing and digital literacy; ethics. These are kept general so they fit any school system.

You will need. A screen or projector and a laptop with the demo (from the download, or on the website). For the warm-up, a torch. For the follow-up, printed worksheets.

Words to know

Say the everyday word first, then the medical word with its meaning.

X-ray
A picture made with invisible rays. Bone blocks more of the rays than soft parts do, so bones show up clearly.
CT
Short for computed tomography. A machine takes many X-rays from all around the body, and a computer builds them into slices.
MRI
Short for magnetic resonance imaging. It uses a strong magnet and radio waves, not X-rays, to show soft parts such as the brain.
Scan
Any picture of the inside of the body made by a machine.
Slice
One thin layer picture from a CT or MRI, like one slice from a loaf of bread.
Fracture
A broken bone.
Pneumonia
An infection in the lungs. It can show up as a cloudy white patch on a chest X-ray. Doctors can treat it.
Radiologist
A doctor who is an expert at reading scans.
AI
Short for artificial intelligence. A computer program that learned patterns from a huge number of examples, so it can do things like describe a picture.
Model
The AI program itself, after it has learned. The screen always names the model that gave an answer.
Prompt
The instructions and question we give an AI. “Under the hood” in the demo shows them.
Data
The examples an AI learns from, such as lots of text and pictures.
Privacy
Keeping personal information, like your medical pictures, to yourself and the people who care for you.
Bias
When an AI works better for some people than for others, for example because the examples it learned from did not include people like them.

If a student asks about a CT “window”

A CT scanner measures how strongly each tiny block of the body stops X-rays. A window chooses which range of those values becomes black-to-white: a brain window shows soft brain tissue, a lung window shows the air-filled lungs. Same scan, different window, different picture.

Before the session: shadow play

Five minutes with a torch. Solid things block light and make shadows. That is the idea behind an X-ray.

  1. Dim the lights and shine the torch at a wall or a sheet.
  2. Hold a hand, a key and a pencil in the beam. Ask: “What do you see? Which things make the darkest shadows?”
  3. Try a sheet of thin paper or a clear plastic ruler. Some things let light through.
  4. Say: “An X-ray works a bit like this. Bone blocks the invisible rays, so it leaves a shadow. Soft parts let more rays through.”

Teacher note. On an X-ray picture the blocked parts look white, so bone is white. It is the shadow idea shown the other way round, like a photo negative.

The session

Case numbers are the number keys in the demo, 1 to 8. Nothing advances by itself: you press the keys, and students stay in charge of their guesses.

The 20-minute plan

TimeWhat happensStudents do
0:00–2:00 Welcome, then the hand X-ray (case 3). Ask: “Who has had an X-ray? Today an AI will look at real scans with us, and we’ll check whether it’s right.” Show of hands. Count the bones (there are 27).
2:00–4:00 The three doors. X-ray: shadows, because bone blocks the rays and air doesn’t. CT: many X-rays from all around, made into slices like a loaf of bread. MRI: a giant magnet and radio waves, no X-rays. Guess which machine is the loud one (the MRI).
4:00–7:00 Case 1, a normal chest. Ask “Is anything wrong here?” Then Analyze. A good AI doesn’t invent problems. Reveal: normal. Vote: normal or not.
7:00–10:00 Case 2, pneumonia. “Compare it with the last one.” Analyze, show the box, Reveal. White and cloudy can mean fluid from an infection. Doctors also listen to your chest and ask how you feel. Vote. Spot the difference.
10:00–12:30 Case 4, the wrist. “Race the AI: who can spot the break first?” Analyze, then Reveal to see both boxes. Point at the screen.
12:30–15:00 Case 7, a brain MRI (or case 5, a chest CT). Slices again. Tell them the company that makes this AI says it isn’t designed to read complex CT and MRI scans. That’s why hospitals use special, tested tools, and why doctors check. Name the brain parts.
15:00–17:00 Under the hood. Show the instructions the AI was given, which model answered, how long it took and about what it cost. Ask about tokens and cost.
17:00–18:30 Wrap-up, three big ideas: AI is a helper; doctors decide; medical pictures are private. An AI can work less well for people unlike the ones in the data it learned from. –
18:30–20:00 Questions. Use the suggested question buttons on screen. Ask.

The 10-minute version

Use cases 1, 4 and 7, then Under the hood and the wrap-up. A suggested split:

TimeWhat happens
0:00–1:00Welcome, and the three machines in one sentence each.
1:00–3:30Case 1, the normal chest: guess, Analyze, Reveal.
3:30–6:00Case 4, the wrist: race the AI to the break.
6:00–8:00Case 7, the brain MRI, and the AI maker’s own limit.
8:00–9:00Under the hood.
9:00–10:00Wrap-up and one question.

Two cases are hidden until you turn them on. A head CT with a bleed and a brain MRI with a bright lump can worry some students. Preview them first. Press P in the demo for the presenter panel, then choose which cases are shown. Serious findings are always framed around how doctors help.

Presenter keys

Every key also has a button on screen.

KeyWhat it does
→ or SpaceNext step: analyze, then reveal, then questions
←Previous step
1 to 8Open case 1 to 8
AAnalyze
RReveal the answer key
KSwitch between kid words and doctor words
HShow or hide the “about here” boxes
UUnder the hood
LPresenter extras, unlocked with the presenter passcode (it asks once)
IInstant: skip the animation
+ − 0Zoom the image in, out, or reset
FFull screen
PPresenter panel: animation speed and hidden cases
EscClose a panel, or cancel a running analysis

Questions students ask

Will AI replace doctors?

No. AI helps doctors notice things and work faster, but the doctor talks to you, examines you and decides. In this demo the AI even lists what it can’t tell.

Is the AI always right?

No. It can miss things or see things that aren’t there. That’s why we check every answer against the answer key, and you saw that it isn’t perfect.

Do hospitals already use AI?

Yes. By the end of 2025 the US FDA had authorized more than a thousand AI-enabled medical devices, and about three out of four of them work with medical images. That is about 1,450 devices, 76% in radiology, as reported from the FDA’s list. Check the FDA’s own list before you quote a number.

Does an X-ray hurt? Is it dangerous?

It doesn’t hurt. It uses a tiny dose of radiation, so doctors take X-rays only when they help. MRI uses no X-rays at all.

Why is the MRI so loud?

Its magnets switch on and off very fast, which makes loud knocking sounds. People wear ear protection.

Whose scans are these?

Real people’s, shared for teaching and research with names and personal details removed.

Can I put my own X-ray in?

No. Medical pictures are private, and this demo doesn’t check anyone’s health. Ask your doctor about your own scans.

How does the AI know?

It learned patterns from a huge amount of text and pictures. We gave it instructions and the image, and it wrote an answer. It isn’t looking anything up in a list of answers.

How much does it cost?

A few cents for each scan in this demo.

Worksheet

A one-page follow-up for after the session. The button prints only the worksheet below, not the rest of this guide.

Answers. Question 1: X-ray is C, CT is B, MRI is A. Question 2: the crack across the upper bone. Questions 3 and 4 have no single right answer; look for reasons, such as “a doctor should check” or “the AI can be wrong”. The break in question 2 is a simple drawing. For a real one, print the wrist X-ray from the demo.

AI Imaging Lab worksheet

Name
Date

1. Match each machine to how it makes its picture

Write the letter of the right idea in the box.

  • X-ray
  • CT
  • MRI
  • A. A giant magnet and radio waves. No X-rays.
  • B. Many X-rays from all around, put together as slices, like a loaf of bread.
  • C. Shadows: bone blocks the rays and air lets them through.

2. Circle the break

Two forearm bones are drawn below. One of them has a break. Circle it.

3. Would you trust the AI on its own? Why?

  • Yes
  • No
  • Not sure

4. Write one rule for using AI in a hospital

AI-assisted analysis — preliminary and not final. Findings must be reviewed and confirmed by a qualified medical expert.

Running the demo

Running the demo from the download

The download is one zip file. Unzip it, then open START-HERE.html. It runs on any computer, with no installation and no sign-in.

  • Windows: right-click the zip and choose Extract All. Open the new folder, then double-click START-HERE.html.
  • macOS: double-click the zip. Open the new folder, then double-click START-HERE.html.
  • Chromebook: open the zip in Files and copy the folder into My files. Then open START-HERE.html.

Unzip first. Opening the page from inside the zip runs it from a temporary folder, and images may not load.

Using the hosted demo

Open the demo on the website. It works for everyone, with no sign-in. Every answer is labelled with the model that gave it. A presenter passcode unlocks extra features from inside the demo.

The day before

  • Download the latest demo. Unzip it on the laptop and on a USB stick, and open it once to check that it starts.
  • Preview the two hidden cases and decide what to show.
  • Print the worksheets.
  • Install laptop updates and turn sleep off, so there is no surprise restart.
  • In the browser: a clean profile, notifications off, zoom at 125%. Press F for full screen.
  • If you have a presenter passcode, try it before the session.

Safety and privacy, in plain words

AI-assisted analysis — preliminary and not final. Findings must be reviewed and confirmed by a qualified medical expert. In plain words: the AI’s answers can be wrong, and a doctor always decides. This demo shows how AI might help doctors. It does not check anyone’s health. If students have questions about their own health, they should ask a parent, carer or doctor.

What this demo does not do

  • It gives no medical advice and does not support decisions about anyone’s care.
  • Nobody in the room is analysed. Every image comes from a public teaching collection.
  • The AI is told the images are for teaching and that nobody will be treated because of its answer.

Children’s privacy

  • No student accounts or logins. The only cookie is the presenter’s unlock cookie on the website.
  • No analytics, and no scripts, fonts or other services from outside the site.
  • Nothing a student says is stored. If a question is typed in, only the fact that a question was asked is logged, never its words.
  • No photos, audio or video of students.
  • Students never upload images, and the demo never stores a picture that is shown to it.
  • The downloaded demo makes no network requests.

The images

  • They come only from public, de-identified teaching and research collections whose licenses allow educational reuse. The credits page names every source.
  • Nobody tries to identify anyone in a collection. Every image is checked for names, dates or hospital marks in the picture, and these are cropped or masked.
  • If a student offers their own scan (“I’ve got my X-ray on my phone!”), the answer is a kind no: medical pictures are private.

Honest and kind

  • Every answer carries a badge naming the model that gave it, so students always know what they are looking at.
  • The answers are real AI answers. We never edit them, and misses stay in. A miss is a teaching moment.
  • The AI never sees the answer key, so a correct answer is earned.
  • Confidence is shown as a word, never as an invented percentage.
  • Every answer was reviewed for tone before it was included.

If something goes wrong

What happensWhat to do
The AI says something wrong or unkind in front of the class Say “this is exactly why doctors check”. Move on to the next case, and note the case and the time so it can be reviewed afterwards.
A presenter feature is slow or failing Follow the choice the screen offers, or carry on without it.
The website will not load Open the downloaded demo from your computer instead.
A student shares a personal medical image Don’t put it on screen. Explain kindly that medical pictures are private.

Image credits for every scan are on the credits page.