Inaugural Lecture by Kristoffer Hougaard Madsen

Programme:

• 15:00 – 16:00 Inaugural lecture titled “Before You Predict the Future, Understand the Present: Probabilistic Modelling and Causal Inference for Personalized Neuroimaging"
• 16:00 – 17:00 Reception

Abstract

Predictive artificial intelligence promises to forecast a patient's clinical future from a brain scan, but that ambition only holds if we can properly characterize and understand their current state.

This talk reviews work spanning probabilistic modelling of brain networks, causal inference through techniques based on non-invasive brain stimulation, and active machine learning for smarter data acquisition, tracing how these threads developed and what they revealed about characterizing individual brain states.

Building on these findings, I will make the case for utilizing mechanistic models with uncertainty quantification to build agency into radiological practice, where agentic workflows continuously plan optimal data collection and interventions. Better biomarkers and more efficient imaging start with understanding the individual brain in front of us, before we try to predict where it is headed.

Biography

Kristoffer Hougaard Madsen studied engineering physics at DTU and discovered functional neuroimaging during an exchange at Florida Atlantic University. He completed an MSc thesis on retinotopic mapping with fMRI in 2004 and a PhD on fMRI processing strategies, defended in 2008.

From 2009 to 2025 he was a senior researcher at the Danish Research Centre for Magnetic Resonance, leading its computational neuroimaging group, while from 2016 also holding an associate professorship at DTU Compute's Section for Cognitive Systems, working at the interface of clinical neuroimaging and computational method development.

Since 2025 he has continued this work as faculty at DTU Compute's Section for Visual Computing, in a position shared with the Department of Radiology at Zealand University Hospital.

His research combines probabilistic modelling, active machine learning, Bayesian optimization, and causal inference to understand brain networks, with a focus on non-invasive brain stimulation.

Teaching is a lasting passion: he has organized the integrative neuroimaging course at the Sino-Danish Center since 2012, alongside several PhD courses, and has run DTU's course on active machine learning and agency since 2018, extending his research directly into the classroom.