Fons van der Sommen: “Trust is key for succesful use of AI”

Van der Sommen is an associate professor in the Video Coding & Architectures research group at the department of Electrical Engineering. His specialty is developing smart algorithms for image recognition, which are able to detect diseases, especially cancer, at an early stage, in order to increase the chances of patient survival. In doing so, he uses different types of medical images, such as CT scans and endoscopy videos.

Cancer in the ceiling?

How important trust is, is shown by an anecdote Van der Sommen relates in his conversation with Lieven Scheire.

"When we wanted to test our cameras, we took a picture of the ceiling. The AI system concluded that cancer cells were present there with 90 percent certainty. You can imagine what the reactions were of the specialists present. They lost all faith in our system, but to us this error came us no surprise. After all, we had only trained our algorithms on images of tissue. It's not at all surprising that they make mistakes when they have to analyze a ceiling."

The researcher emphasizes how important it is to really talk with the people who will be using his tools. "What exactly do they need, and where does it go wrong?".

For that, he likes to work with experts in the field of Human Technology Interaction at TU/e, who know about UX (user experience) and UI (user interface). "That's also where the strength of our new AI institute EAISI lies. It encourages you to collaborate with other disciplines. That's the only way we can make AI a success."

Listen to the podcast

In the latest episode of our science podcast Sound of Science, you can hear all about Van der Sommen's research and the great importance of collaboration with other scientists and users.

You can listen to the podcast in the Spotify player below. Or go to your favorite podcast platform and search for 'Sound of Science'. Subscribe to the channel to be automatically notified of new episodes!

Comments or suggestions? Email us at nieuws@tue.nl.