Artificial intelligence refers to computer systems that can learn patterns from large amounts of data. Medicine is one of the most promising areas for this technology, and one that calls for the most care: used well, it can speed up diagnosis and lighten clinicians’ workload; used badly, it can harm patients.
Medical imaging
Imaging is where AI is most widely used in medicine today. Systems that flag suspicious findings on X-ray, CT, MRI and mammography images, prioritise emergencies such as brain bleeds, or screen retinal photographs for diabetic retinopathy are in clinical use. The U.S. Food and Drug Administration’s (FDA) list of AI-enabled medical devices has passed a thousand products, the large majority of them in radiology.
Drug discovery and AlphaFold
Knowing a protein’s three-dimensional structure is crucial for understanding disease and developing drugs. AlphaFold, developed by Google DeepMind, broke new ground by predicting protein structures from their amino acid sequence with high accuracy. Half of the 2024 Nobel Prize in Chemistry went to Demis Hassabis and John Jumper for this work; the other half went to David Baker for his work on protein design. AI is also used to screen candidate drug molecules and design clinical trials.
Everyday clinical tasks
Systems that transcribe patient consultations and draft notes, early-warning tools that estimate risk, and applications that ease appointment and paperwork flows are less visible but increasingly common uses of AI in hospitals.
Limits and risks
- Data bias: A model trained on one patient group may perform worse for patients of a different age, sex or ethnic background.
- Privacy: Health data is among the most sensitive personal data.
- Overreliance: Clinicians or patients accepting a system’s suggestion without question can lead to errors.
- Accountability: Who is responsible when something goes wrong is still debated.
In guidance published in 2021, the World Health Organization recommends that AI for health protect human autonomy, be transparent and explainable, and be designed inclusively. The European Union’s AI Act also treats AI systems for medical purposes as high-risk and imposes extra obligations.
A note for patients
General-purpose chatbots can help you understand health questions, but they do not diagnose and can make mistakes. See a health professional about your symptoms, and read the privacy terms of apps you share your health data with.
In which fields are approved AI medical devices used?
The US Food and Drug Administration (FDA) keeps a public list of the AI-enabled medical devices it has authorised for marketing. As of 2025 the number of products on this list had passed one thousand, and the distribution is quite uneven:
- Radiology far ahead: About three quarters of the devices on the list are aimed at reviewing X-ray, CT, MRI and mammography images.
- Cardiology in second place: Applications such as ECG interpretation and detection of heart rhythm disorders stand out.
- Other fields: There are also approved tools in neurology, ophthalmology, pathology and gastroenterology, but in smaller numbers.
A tool's approval does not mean it will perform equally well in every hospital and every patient group; approval is granted for a specific intended use and specific conditions.
How is the reliability of a health AI evaluated?
When researchers and hospitals evaluate a new tool, they look at a few basic questions:
- External validation: Was the tool tested on data from different devices and patient groups outside the hospital where it was developed?
- Prospective study: Was it tried not only on past data but within real clinical workflow, and if possible in randomised trials?
- Transparent reporting: The CONSORT-AI and SPIRIT-AI guidelines published in 2020 aim to standardise how clinical trials involving AI are reported.
- Continuous monitoring: After deployment, is it tracked whether performance declines over time?
In 2024 the World Health Organization also published separate guidance on the use of large multi-modal models (including chat-based systems) in health, offering more than 40 recommendations to governments, developers and health institutions.
This article is for general information and is not a substitute for medical advice. Consult your doctor about decisions concerning your health.
Sources and further reading
- World Health Organization (WHO), “Ethics and governance of artificial intelligence for health” (2021).
- U.S. Food and Drug Administration, “Artificial Intelligence-Enabled Medical Devices” list.
- Nobel Prize Outreach, “The Nobel Prize in Chemistry 2024” press release.
- Jumper, J. et al., “Highly accurate protein structure prediction with AlphaFold”, Nature (2021).
- U.S. FDA – Artificial Intelligence-Enabled Medical Devices (list)
- Liu X. et al. – CONSORT-AI extension, Nature Medicine (2020)
- WHO – Ethics and governance of AI for health: Guidance on large multi-modal models (2024)
Leave a comment
Your email address will not be published. Required fields are marked *