AI in medicine

In short: AI helps in medicine with image diagnostics, drug discovery and documentation. It is meant to support doctors, not replace them. Strict rules apply to medical devices.

In more detail: Well-defined tasks with many examples work especially well, such as detecting abnormalities on X-rays or skin photos.

In detail

• Imaging: CNNs and vision transformers find tumours, retinal damage, fractures. Often as a “second opinion”.
• Language: summarise doctors’ letters, write dictations, search guidelines.
• AlphaFold: predicts the 3D shape of proteins from the amino-acid sequence. Nobel Prize in Chemistry 2024 (Hassabis, Jumper, Baker).
• Drugs: AI proposes molecules and ranks candidates, saving lab work.
• Risks: biased data (skin colour, sex), transferability to other devices/clinics, false trust (“automation bias”), privacy of sensitive health data.
• Approval: in the EU the Medical Device Regulation (MDR) and the EU AI Act (high risk) apply.
• For laypeople: chat assistants do not replace a diagnosis. For symptoms ask a doctor, see Using AI assistants sensibly and safely.

See also: Bias, fairness and data quality, Convolutional networks (CNN) for images