🩺 Real World AI · Medical Diagnosis

Medical AI Decision Lab

Select symptoms, watch Bayesian probability update live, tune detection thresholds, and discover why a false negative in medicine can cost a life — the highest-stakes AI application!

🩺 Symptom Checker

📊 Bayesian Update

⚠️ False Positive Lab

🔬 X-Ray Analyser

🏆 Patient Cases

How AI Diagnoses Disease

🩺

Symptom Input

Patient describes symptoms. AI cross-references against a database of millions of clinical cases to rank possible conditions.

📊

Bayesian Inference

Each new symptom updates the probability of each diagnosis. More symptoms = narrower, more confident diagnosis.

⚖️

Threshold Setting

The decision threshold balances false positives (unnecessary treatment) vs false negatives (missing real disease).

🔬

Medical Imaging

CNNs analyse X-rays, MRIs, and retinal scans — often surpassing human radiologists in detection accuracy.

🩺

Wizzy the AI Tutor

Welcome to the AI Doctor lab! 🏥 Click symptoms to add them to the patient profile. The AI runs a differential diagnosis — it ranks all possible conditions by probability. Add more symptoms and watch the probabilities sharpen. More clues = more accurate diagnosis!

Step 1 — Symptom Checker

Click symptoms to add:

Selected: None yet

Differential Diagnosis

Select symptoms to see diagnosis

The AI matches your symptom pattern to thousands of clinical cases.

🩺

Wizzy the AI Tutor

Watch Bayes' theorem in action! Each symptom is a new piece of evidence that updates the probability of each diagnosis. A symptom that's rare in healthy people but common in patients with Disease X dramatically increases the P(Disease X). This is exactly how AI doctors reason!

Step 2 — Live Bayesian Probability Update

Add symptoms one by one and watch the probabilities shift:

📊 Probability Updates

Add symptoms to see how each one updates the diagnosis probabilities!

🩺

Wizzy the AI Tutor

This is the most important lesson in medical AI! 🏥 Drag the detection threshold. Lower it = catch more real cases (high recall) but also alarm more healthy people (low precision). Raise it = fewer false alarms but miss real patients. In cancer screening, which error is worse?

Step 3 — The False Positive Trade-off

🎚️ Detection Threshold

Threshold: 50%

—

Precision

—

Recall (Sensitivity)

—

False Positives

—

False Negatives

Drag the threshold to see how it affects precision and recall.

⚖️ The Medical Dilemma:
A missed cancer (false negative) = patient doesn't get treatment they need.
A false alarm (false positive) = unnecessary surgery, stress, cost.
Which is worse? It depends on the disease!

🩺

Wizzy the AI Tutor

AI can analyse chest X-rays and spot pneumonia markers that humans might miss! 🔬 Click on the X-ray image to hover over regions — the AI highlights areas of concern. Different shades represent density — denser tissue appears brighter. Infected lungs show cloudiness called "consolidation".

Step 4 — Chest X-Ray AI Analyser

Move mouse over the X-ray to inspect regions

🔬 AI Analysis

Classification—

Confidence—

Scan typeChest PA

Select an X-ray type above to see AI analysis.

🩺

Wizzy the AI Tutor

🏥 Final challenge! Read each patient's symptoms and make your own diagnosis, then compare with the AI's answer. Can you beat the AI's accuracy? Remember: the AI is not perfect — it's a decision support tool. The final call always belongs to the human doctor!

Step 5 — Diagnose 5 Patients

0/5 cases

Select a patient to begin diagnosis.

🩺

Medical AI Badge!

You mastered Bayesian diagnosis, precision-recall trade-offs, and medical image analysis!

Optional. Stays on this device only — not sent to WhizzStep.

🩺 WhizzStep AI Lab

Activity completion card for

Student Name

has explored Medical AI — Diagnosis & Ethics

This records completion of a browser activity only. It is not an accredited certificate or proof of mastery.

Medical AI Expert

Bayesian Thinker

Diagnostic AI

whizzstep.in

Key Concepts Mastered

Bayesian Inference

📊 Updating Beliefs

P(Disease|Symptoms) ∝ P(Symptoms|Disease) × P(Disease). Each new symptom multiplies the prior probability.

Precision vs Recall

⚖️ The Trade-off

Precision = of all positives flagged, how many were real? Recall = of all real cases, how many did we catch?

False Negatives

⚠️ Missing Disease

In cancer screening, a false negative is catastrophic — the patient doesn't get treatment. So recall must be very high.

DeepMind Health

🏆 Real Deployment

DeepMind's AI detects 50+ eye diseases from retinal scans with 94% accuracy — better than most ophthalmologists.

Decision Support

🤝 AI + Doctor

AI doesn't replace doctors — it flags cases for human review. The final diagnosis is always a human's responsibility.

Data Privacy

🔒 Medical Ethics

Medical AI needs patient data to train. Who owns it? Who can access it? HIPAA and GDPR try to answer this.

About this lab

AI-20 High-Stakes AI and Human Oversight Classes 9-12 18 min Applied

Learning objective: Explore a simplified educational decision-support scenario and the role of human oversight in high-stakes AI use.

What this simplifies: This is not diagnosis or medical advice; it is a simplified educational scenario about human oversight of AI recommendations.

Privacy: No learner input leaves the device.

Teacher prompt: Ask the class why this simulation might mislead someone who takes it too literally.

Reflect: What is one thing this activity showed you that you did not expect?

← Deepfake Verification Lab Back to All Labs Intent Chatbot Builder →

Bring this learning to my school →