Serge Rombouts: 'AI is learning from brain scans and helping find a diagnosis'
Serge Rombouts is a physicist whose PhD thesis was about functional MRI (fMRI). This visualises activity in regions of the brain. The appealing images of glowing brain regions that emerge from the computer are the result of calculations. According to Rombouts, this isn’t proper artificial intelligence. ‘We’ll only have that when the computer can issue advice or make a decision.’ That is precisely what Rombouts wants to achieve. He works with partners including Erasmus Medical Center, which, like Leiden, is part of Medical Delta.
Much of Rombout’s work is devoted to the early identification of brain disorders such as dementia. ‘We are working to train computers to recognise a disease from advanced scans of brain networks. These computers would combine the scans of the structure, function and activity of one patient and compare these with those of other people. On the basis of a decision model that they would create by learning from all this data, these computers might in future be able to contribute to the diagnosis.’
Collaboration is key
Collaboration between different researchers is an important key to success, Rombouts says. ‘When I started in Leiden, I spoke to Professor Joop van Gerven from the Centre for Human Drug Research. He wanted to use fMRI to see the effects of medication on the brain, but it was too complex and restrictive having people do tests in the scanner. I could also see the problem and spent years working on fMRI scans of brain networks at rest, so without tests. A fantastic moment, for both of us. We embarked on a completely new line of research using fMRI to discover the effects of medication on the brain.’
Rombouts also works with Erasmus Medical Center. ‘People from families with inherited frontotemporal dementia are being studied there, and they undergo various MRI scans in Leiden each year. PhD candidate Rogier Feis recently showed that the disease is visible in the brain to an intelligent system shortly before the symptoms arise.’ (f)MRI is not yet able to contribute to the diagnosis very long before the first symptoms occur. ‘As the disease progresses, the computer starts to recognise more on MRI images.’
Text: Rianne Lindhout
Photo: Patricia Nauta