In diagnosis
Models trained on large collections of labelled images can flag suspicious areas on X-rays, scans and eye photographs, prioritise urgent cases in a queue, or measure features consistently. Regulators in many countries review such software as a medical device before it can be sold for clinical use, and the clinician stays responsible for the decision.
In drug discovery
Developing a medicine takes years and many failures. AI is used to predict the three-dimensional shape of proteins, to screen huge numbers of possible molecules for ones likely to work, and to suggest new designs. These steps can shorten early research, but a candidate still has to pass laboratory tests and clinical trials that only people and patients can complete.
Limits and care
A model is only as good as the data it learned from. If that data under-represents some groups, the tool can work worse for them. Patient data must be handled under strict privacy rules, and results should be checked in the setting where the tool will actually be used.
Where it is used
- Reading images Flagging scans and photographs for a specialist to review.
- Triage Putting urgent cases at the front of a queue.
- Research Predicting protein structures and screening candidate molecules.
- Hospital operations Forecasting demand for beds and staff.
AI as an aid compared with AI as the decision-maker
In current practice, approved tools support a clinician, who stays responsible. Each is validated for a specific task, such as one type of scan, and is not a general doctor.
Problems arise when a tool is used outside the setting or the population it was tested on, which is why regulators review each one as a medical device.
Key terms
- Radiology
- The medical field that reads imaging such as X-rays and scans.
- Protein structure prediction
- Estimating the three-dimensional shape of a protein from its sequence.
- Clinical trial
- A study in people that tests whether a treatment is safe and works.
- Software as a medical device
- Software that is itself regulated as a device because of its medical purpose.
Common questions
Can AI diagnose disease on its own?
- Approved tools act as aids under clinical supervision. A qualified clinician makes the diagnosis.
Does AI invent new drugs?
- It helps with early stages such as finding and ranking candidates. Safety and effectiveness still have to be shown in clinical trials.
Is patient data safe when AI is used?
- It depends on how the system is built and governed. Good practice includes de-identifying data, restricting access, and following health privacy law.
Sources and further reading
Independent pages we checked while writing this guide. They are not Reinnder products, and Reinnder is not affiliated with them.
Reinnder’s angle
How Reinnder looks at aI in diagnosis and drug discovery
Reinnder Health is the planned home for clinical technology, and any medical device would have to meet the rules of its market before use.
This guide explains the technology in general terms. It is not advice, and it does not describe a Reinnder product on sale.