The Bayesian Brain: From Clinical Prediction Models of Pain to High-Field Laminar Imaging of Interoception

Abstract:

The Bayesian Brain hypothesis describes perception as the brain’s unconscious inference on states of the world. This dissertation had two goals to apply that idea: First, to use it in clinical context to predict the treatment outcome of chronic pain. Second, to perform an experiment to detect the layered signature of a Bayesian computational model in the interoceptive cortex (insula).

At the beginning, I examine the relation between treatment expectations of patients with chronic pain and the treatment outcome. I find an association, which allows to predict the outcome out- of-sample. Notably, this is the first time, that the clinical response to treatment can be predicted out-of-sample on the basis of treatment expectations in the field of pain.

In the next two chapters, I perform methodological developments as preparations for the experiment. First, I adapt methods to check the consistency of computational Bayesian inference pipelines to models of behaviour. Second, I improve the robustness of the reconstruction of functional MR images, acquired with non-Cartesian readout trajectories. This allows the imaging of cortical laminae. Based on these improvements, I develop an experiment, which combines a pain learning task with laminar fMRI.

Finally, I perform the first laminar analysis in interoception. In a heartbeat attention task, I find layered differences in signal in both hemispheres of the dorsal dysgranular insula.

Chapter 3: Individual treatment expectations predict clinical outcome after lumbar injections against low back pain

Chapter 4: Simulation-based calibration for behavioural models

Chapter 5: Reconstruction of high resolution fMRI images with spiral readout trajectories to study laminar brain function

Chapter 7: Layer analysis of interoceptive heartbeat attention task

Citation

Müller-Schrader, Matthias. „The Bayesian Brain: From Clinical Prediction Models of Pain To High-Field Laminar Imaging of Interoception“. PhD Thesis, University of Zürich and ETH Zürich, 2025. https://doi.org/10.5167/uzh-270865.

@phdthesis{muller-schraderBayesianBrainClinical2025,
  title = {The {{Bayesian Brain}}: {{From Clinical Prediction Models}} of {{Pain To High-Field Laminar Imaging}} of {{Interoception}}},
  author = {{M{\"u}ller-Schrader}, Matthias},
  year = {2025},
  address = {Z{\"u}rich},
  school = {University of Z{\"u}rich and ETH Z{\"u}rich},
}