Statistical learning of pain signals in the brain

Statistical learning of pain signals in the brain

Pain often fluctuates over time. We don’t know why, but we know that pain fluctuations strongly affect how well pain can be managed in everyday life.

Recent work in our group suggests that the human brain can learn and control the temporal dynamics of pain. Our approach combines computational models of sequence learning with behavioural and neuroimaging methods.

Specifically, we showed that somatosensory regions in the brain encode probabilistic predictions of the frequency of noxious inputs (Mancini et al., Nature Commun 2022). As predicted by Bayesian theory, cortical sensory responses to noxious inputs are inversely related to the confidence of probabilistic predictions of pain. The smaller the response, the higher the confidence of the posterior, and vice versa (Mulders et al. PNAS 2023). These probabilistic learning processes modulate pain perception (Onysk et al. eLife 2023). Now we are investigating whether statistical learning is affected in chronic pain. Watch this space.

Supported by

Neural dynamics of homeostatic control

Neural dynamics of homeostatic control

Homeostasis refers to the ability of an organism to adjust its internal environment to maintain a stable equilibrium. We use control theory and simulations with artificial neural networks to understand how neural populations in the midbrain contribute to homeostatic control and endogenous pain regulation.

This work is relevant to chronic pain conditions, which often involve dysfunctional endogenous pain control. The implications of this research could provide targets for the pharmacological treatment of pain, and a computational model would provide predictions about the behaviour of these circuits, which could then become avenues for further experimental and pharmacological research.

Computational methods for Neuroimaging

Computational methods for Neuroimaging

We develop computational methods for the analyses of human neuroimaging data.

For example, we have developed Fourier-based methods for the study of cortical topographic representations in humans, imaged with brain fMRI (Mancini et al. 2012; Mancini et al. 2019).

Recently, we are contributing tools for the analyses of high-density diffuse optical tomography data.