Large-scale brain attractor dynamics in pain: a precision neuroimaging study in experimental pain

For the full preregistration text, please refer to the attached document. Information processing in the brain is both distributed across multiple regions and hierarchical in nature. Complex, higher-order cognitive and affective functioning and multifaceted experiences, such as pain, are often thought to be represented and governed by higher-order “integrator regions” at the top of this hierarchy, which have access to all information relevant to these complex processes. However, exactly these functions are the ones that often defy empirical mapping to specific brain regions. According to a recently introduced theoretical framework [1,2], large-scale integrative information processing at the apex of the computational hierarchy is achieved through the attractor dynamics of the entire system, rather than through dedicated integrator regions (Figure 1). In this framework, large-scale brain dynamics are organized around attractor states that - through a mathematical duality with simpler, hierarchical systems (Figure 2) - can be assigned to precise computations (from Bayesian causal integration to the computations of prediction errors). Previous studies [1,2] have demonstrated that attractor states and the corresponding dynamics can be reconstructed from functional brain connectivity (as measured by fMRI). Based on the topology of the functional connectome, the model establishes an energy level for any arbitrary activation patterns and determines a “trajectory of least action” towards one of the finite number of attractor states, that minimize this energy (Figure 1). Initial evidence [1,2] suggests that attractor networks yield a promising phenomenological computational model for large-scale brain dynamics. A wider applicability of this framework requires the validation of its key assumption, which states that attractor states are (approximately) orthogonal to each other within a single individual. This assumption not only guarantees optimal computational properties but it also allows us to make use of the aforementioned mathematical duality (Figure 2) to understand the computations performed by the attractors and to deploy a quick and robust analytical technique to reconstruct attractor dynamics from less data. In the present project, we aim to reconstruct and investigate large-scale attractor states in a precision neuroimaging dataset of n=1 participant, including a total of 33 scanning sessions, consisting of resting state scans, experimental heat runs, as well as tonic experimental pain (using capsaicin). We aim to harness the high power of this dataset and accurately reconstruct a large number of individual attractors. We will verify the orthogonality assumption and identify attractors that integrate information related to pain experience. By applying multiple preprocessing pipelines, we will evaluate the robustness of reconstructed attractors to various methodological choices. Finally, we aim to characterize changes in the energy landscape of the attractor network during sustained pain experience and sleep, focusing on dwell-times on the basins of attraction and the dynamics of transitioning between different states (e.g. the cost of optimal paths through the energy landscape). These metrics may prove useful in a wide variety of research scenarios, in pain research and beyond. [1] Englert R, Kincses B, Kotikalapudi R, Gallitto G, Li J, Hoffschlag K, Woo CW, Wager TD, Timmann D, Bingel U, Spisak T. Connectome-Based Attractor Dynamics Underlie Brain Activity in Rest, Task, and Disease. eLife. 2024 Sep 5;13. [2] Spisak T., Ghost in the machine: information integration and causal inference in the brain via large-scale attractor dynamics, https://pni-lab.github.io/ghost-in-the-machine