André Marquand has been appointed professor of Computational Psychiatry at Radboud university medical center / Radboud University. He applies artificial intelligence techniques to large datasets containing different types of data. Thereby, he can predict how psychiatric disorders will develop in individual patients. This approach also contributes to the development of improved therapies.
Psychiatric disorders such as ADHD, autism, schizophrenia and depression have many forms and causes. But we still lack knowledge on how they originate and progress. Lead researcher André Marquand of the Donders Institute for Brain, Cognition and Behaviour uses artificial intelligence (AI) to recognize patterns in large datasets containing different types of data, such as brain scans and genetic data as well as behavioral and environmental factors. This approach allows to identify differences between individuals, predict the disease course of individual patients, and discover leads for personalized therapy.
Ocean of data
Last year, Marquand published a study in which he and colleagues analyzed brain scans of 60,000 people. Using these data, they created an atlas of the brain. Brain scans of psychiatric patients can now be compared with this atlas to reveal abnormalities. Recently, he also started using data from smartphones. Together with colleagues from the Department of Psychiatry, he spent one year collecting this type of data from one hundred patients suffering from depression. Marquand explains: ‘How many WhatsApp messages you send per day, how often you move, how much your screen is on; these data say a lot about your behavior. We swim in an ocean of data, but it is difficult to recognize patterns in this ocean. A computer does this much better. That is why we use AI methods to make sense of this wealth of data.’
For many people, AI is a ‘black box’ that is difficult to grasp. That's why Marquand uses techniques that are relatively simple. ‘We need to be able to understand it and explain it to others’, he says. ‘That increases confidence in the results we generate and provides others with the chance to check them. But naturally, we also want to make good predictions, and more complex methods can be better at that. That remains an important tradeoff in our work.’ Ultimately, this should lead to better treatments for psychiatric conditions. ‘By getting a better picture of differences between patients, we can apply personalized treatment in the future’, Marquand said. ‘In the field of oncology, they are already very advanced with such a ‘precision medicine’ approach; we need to bring that to the field of psychiatry as well.’
Marquand (Vevey, Switzerland, 1977) studied computer technology and psychology at the University of Canterbury in New Zealand. He received his doctorate from King's College London on Machine Learning techniques to analyze MRI images (dissertation title: ‘Probabilistic Machine Learning Methods for Multivariate Prediction of Magnetic Resonance Images’). Subsequently, he was a postdoc and associate professor in that institute. In 2014, he joined Radboudumc, first as assistant professor and from 2019 onwards as associate professor. He obtained several grants, including a VIDI (2015), Wellcome Trust ‘Digital Innovator’ grant (2020), Horizon Europe grant (2020) and NIH R01 grants from the U.S. government (2021, 2022).
Marquand has been appointed professor of Computational Psychiatry for a five-year term, effective from December 1, 2022. His chair is funded for fifty percent by a Consolidator Grant from the European Research Council (awarded in 2020).