Glossary
What is AI in healthcare
In short
AI in healthcare means programs that help doctors read scans and test results, keep records, monitor patients and take work off the reception desk.
AI in industries · AI in healthcare
MMOOZZGG.com editorsChecked: 1 min read
For a child
Imagine you are building a tall tower from a construction set, and your grandmother watches to see if it wobbles. Doctors also have a careful helper. It looks at scans and notices small spots that are easy to miss. But the doctor still treats and talks to the patient, because the doctor knows the person and is responsible for them.
In plain words
A doctor spends a lot of time on work that does not need their experience: filling in records and searching the medical history for data. And where experience is needed, for example when reading an X-ray or a CT scan, a tired doctor can miss a small detail. The program analyzes scans and marks suspicious areas, turns the doctor’s speech during an appointment into the text of a report and reminds the patient about medicines. For example, a radiologist opens a study where the system has already circled an area in the lung, and looks at it more closely. The doctor writes the final conclusion.
For business
For a clinic there are two different zones. The first is non-medical tasks such as booking, answering patient questions and transcribing appointments. They launch faster, and the effect shows in the load on the reception desk and in doctors’ time. The second is help with diagnosis and treatment. Such a program falls under the definition of a medical device: the law directly includes special software in it, and only registered devices may be used (323-FZ, Russia’s healthcare law, Article 38). The price depends on the zone, the integration with the medical information system and where data is stored. The personal data law puts health information in a special category, so lawyers check the processing before launch. A sensible path: start with an administrative task and count the hours saved.
For a specialist
Main directions: medical image analysis with convolutional networks and transformers for detection and segmentation, clinical decision support systems (CDSS), clinical text processing (clinical NLP), speech recognition for records, and forecasting deterioration from monitoring data. Quality is judged not only by sensitivity and specificity on retrospective data but also in prospective studies with patient outcomes, because a model loses accuracy on another clinic’s data (distribution shift). In its guidance of 28 June 2021, the World Health Organization set out six agreed principles for the ethics and governance of AI in health. In Russia, medical software falls under the definition of a medical device in Article 38 of Law 323-FZ and needs state registration.
Example
A chain of clinics connected appointment transcription: the doctor talks to the patient, and a draft report appears in the medical system. The doctor edits and signs it instead of typing the text after the shift.
Related terms
Questions and answers
Can AI make a diagnosis instead of a doctor?
Sources
Where companies use it
Not sure whether your company needs this?
AI solutions map