How AI can help us become more precise

Text: Annika Lund, written in Swedish for the magazine Medicinsk Vetenskap no 4/2021.

What is AI’s role in precision medicine?

“One area where it is used is in handling large volumes of information, i.e., Big Data. Omics, such as genomics, proteomics and metabolomics, generate enormous volumes of data. A self-learning system that handles this sort of data can see patterns that a human brain is not be able to pick up. It can provide us with important new knowledge, for example clustering patients in various ways. This might identify a new subgroup that has some important factor in common, which is relevant for diagnosis or treatment. Monitoring seriously ill patients also generates large volumes of data that need handling, and AI can help with that. But the use that AI is best known for is in image processing, for example mammograms. Here, AI techniques have trained on old images and learned to analyse tumours alongside radiologists. So we can be more confident about the results and it saves radiologists some time.”

Tell us about one of your own projects.

“One of my projects involves working with psychiatrists and psychologists to investigate whether AI can improve internet-based CBT. We want to find out whether AI can help the treating psychologist pick up more information from the system. This could include noting that a patient is using the system late at night. That could be a sign of sleeping problems, which is clinically relevant.”

You’ve been working on a strategic project called AI @ KI. What have you observed?

“For one day a week, I’ve been visiting different parts of the health care sector where Karolinska Institutet researchers work to get an idea of how AI is currently being used. I’ve also visited interdisciplinary units such as SciLifeLab to ask the same question. I’ve observed that AI is already being used in an impressive number of ways, but in order to do more, which is something many people would like, we need more people with AI skills.”