Universität zu Köln

Käsbauer, Anne-Sophie ORCID: 0000-0003-3757-385X (2020). Modulation of behavior and brain activity by probabilistic inference. PhD thesis, Universität zu Köln.

Preview

PDF Diss_final_kaesbauer_2021_seitenzahlzentriert.pdf - Accepted Version Bereitstellung unter der CC-Lizenz: Creative Commons Attribution Non-commercial No Derivatives. Download (5MB) | Preview

Abstract

Expectancies and beliefs about upcoming sensory events encoded by the brain play a crucial role in shaping our perception. Therefore, stimulus detection and processing can be facilitated by prior beliefs about the stimulus’ location or its features. These beliefs are rapidly generated by former observations/experience of the individual. Bayesian principles can evidently be used to describe this probabilistic inference. The present thesis aimed to characterize the mechanisms underlying probabilistic inference in the healthy and the lesioned human brain. In healthy participants, probabilistic inference in the context of attentional deployment has already been described with the help of computational models, and the underlying neural mechanisms have been explored with functional neuroimaging (Dombert, Kuhns, et al., 2016; Kuhns et al., 2017; Vossel et al., 2015). However, it is not known how the resting-state network architecture of the brain relates to this process and how the lesioned brain performs probabilistic inference. To investigate these questions, two experiments have been conducted using modified versions of a Posner-cueing paradigm. In this context, probabilistic inference describes the ability to infer changing probabilities about the validity of a cue and the updating process of the belief about them. By manipulating the percentage of cue validity (%CV) (i.e., the proportion of valid and invalid trials) over the time course of an experiment, the participants had to infer the actual cue validity level (i.e., the probability that the cue will be valid in a given trial), so that probabilistic inference could be assessed. In Experiment 1, a modified location-cueing paradigm with block-wise changes of the %CV and true and false prior information about the %CV before each block was employed in healthy young participants. A Rescorla-Wagner model was used to characterize probabilistic inference. Moreover, resting-state fMRI was recorded before and after the task and a seedbased correlation analysis was used to define the resting-state functional connectivity (rsFC) of the right temporo-parietal junction (rTPJ). Correlations of each behavioral parameter with the rsFC before the task, as well as with changes in rsFC after the task, were assessed in a ROI-based approach. It was observed that higher intrahemispheric rsFC between rTPJ and IPS before the task was associated with slower probabilistic inference after false priors. Furthermore, increased interhemispheric rsFC between rTPJ and lTPJ after the task was related to relatively faster probabilistic inference in false blocks. Both findings support previous research and highlight that not only resting-state connectivity per se is relevant for cognitive functions but also that cognitive processing during a task can change connectivity patterns afterwards in a performance-dependent manner. In Experiment 2, probabilistic inference in stroke patients was investigated to assess a hypothesized relationship with the spatial neglect syndrome (Experiment 2a) as well as commonalities and distinctions between probabilistic inference in different cognitive subsystems (Experiment 2b). Three modified versions of the Posner-cueing task with different cue types were used to investigate spatial attention (location cues), feature-based attention (color cues) and motor-intention (motor-response cues). In contrast to Experiment 1, no prior information about the %CV was provided and probabilistic inference was operationalized by assessing the impact of the %CV manipulation on RTs by means of regression analyses as well as by asking participants to explicitly estimate the %CV. Furthermore, patients were screened for the neglect syndrome using a diverse neuropsychological test battery. Lesion-symptom mapping (VLSM) as well as lesion-network mapping was performed on the relevant behavioral parameters. The results indicated that patients’ probabilistic inference abilities across domains were not per se impaired. However, by trend it was found that some right hemisphere damaged patients exhibited difficulties using their knowledge to adapt their behavior in contralesional space as indicated by a reduced modulation of RTs by %CV in invalid contralesional trials in the spatial attention domain. However, there was no strong evidence for impairments of probabilistic inference being related to the neglect syndrome. Moreover, the correlation of the two probabilistic inference parameters (invalid contralesional %CV regression weight & averaged explicit %CV estimate) within domains revealed no significant relationship between the both, stating them as independent components of probabilistic inference, which was further supported by the VLSM results. However, the correlations across domains revealed some commonalities, which were also in line with the VLSM results. Thus, our data suggests that the neural implementations for probabilistic inference seem to be dedicated to domain-specific subsystems, which share some common nodes. Consequently, the present thesis provides novel insights into the computational mechanisms of probabilistic inference in the healthy and lesioned brain. The work thereby enables future studies to transfer the gained knowledge from basic research of healthy participants and patients to clinical applications.

Item Type:	Thesis (PhD thesis)
Creators:	Creators Email ORCID ORCID Put Code Käsbauer, Anne-Sophie a.kaesbauer@fz-juelich.de orcid.org/0000-0003-3757-385X UNSPECIFIED
URN:	urn:nbn:de:hbz:38-520228
Date:	December 2020
Language:	English
Faculty:	Faculty of Human Sciences
Divisions:	Faculty of Human Sciences > Department Psychologie
Subjects:	Psychology Medical sciences Medicine
Uncontrolled Keywords:	Keywords Language probabilistic inference English spatial attention English stroke English voxel-based lesion symptom mapping English lesion-network mapping English computational modelling English resting-state functional connectivity English
Date of oral exam:	15 April 2021
Referee:	Name Academic Title Vossel, Simone Jun.-Prof. Dr. Weidner, Ralph PD Dr.
Refereed:	Yes
URI:	http://kups.ub.uni-koeln.de/id/eprint/52022