Periodic Reporting for period 2 - MediCoDe (The Medial Frontal Cortex in Cognitive Control and Decision Making: Anatomy, Connectivity, Representations, Causal Contributions)
Berichtszeitraum: 2023-08-01 bis 2025-01-31
Cognitive control enables humans to flexibly adapt their behavior for goal achievement. Despite intensive research we still miss an overarching understanding of cognitive control and its main underlying structure, the posterior medial frontal cortex (pMFC). This is due to insufficient consideration of pMFC neuroanatomy, its subregions and individual variability; low sensitivity of conventional group-level studies; diffuse use of multiple methods and paradigms in disparate studies impeding differentiation of general cognitive-control principles from study idiosyncrasies; sparsity of causal evidence in humans.
This project proposes a radical shift in tackling these issues by two novel approaches:
A) Within-subject multimodal dense sampling and representational modeling considering individual neuroanatomy will provide a fine-grained spatiotemporal mapping of representations of cognitive-control and decision-making variables to connectivity-based pMFC subregions and enable discriminating between competing theories. The rationale is that representations of latent constructs of cognitive control must generalize across tasks and contexts. In multiple sessions behavior, fMRI, EEG, eye movements and autonomic reactions are recorded while participants perform tasks enabling to estimate cognitive control variables (CCV) from behavior and computational models. Representational similarity analysis of CCVs across tasks and modalities in a regression framework will identify signatures of core theoretical components of cognitive control.
B) A novel stimulation method using low-intensity transcranial focused ultrasound stimulation (TUS) noninvasively modulates neuronal activity in deep brain structures with excellent spatial resolution. TUS combined with EEG and fMRI will causally test the necessity of CCV representations in pMFC and its network for adaptive behavior.
The project will open up new research avenues to address individual differences and pathological changes in cognitive control.
This project proposes a radical shift in tackling these issues by two novel approaches:
A) Within-subject multimodal dense sampling and representational modeling considering individual neuroanatomy will provide a fine-grained spatiotemporal mapping of representations of cognitive-control and decision-making variables to connectivity-based pMFC subregions and enable discriminating between competing theories. The rationale is that representations of latent constructs of cognitive control must generalize across tasks and contexts. In multiple sessions behavior, fMRI, EEG, eye movements and autonomic reactions are recorded while participants perform tasks enabling to estimate cognitive control variables (CCV) from behavior and computational models. Representational similarity analysis of CCVs across tasks and modalities in a regression framework will identify signatures of core theoretical components of cognitive control.
B) A novel stimulation method using low-intensity transcranial focused ultrasound stimulation (TUS) noninvasively modulates neuronal activity in deep brain structures with excellent spatial resolution. TUS combined with EEG and fMRI will causally test the necessity of CCV representations in pMFC and its network for adaptive behavior.
The project will open up new research avenues to address individual differences and pathological changes in cognitive control.
WP A1: A new processing pipeline for connectivity-based parcellation of the posterior medial frontal cortex (pMFC) has been developed. In the human connectome project data set participants with unequivocal gyrification patterns (two groups: with/without prominent paracingulate sulcus) were identified and individual seed regions for tractography were determined. Preliminary results converge on a rather stable number of nine connectivity-based clusters (regions), which show remarkably similar patterns within each configuration group. Analyses are ongoing and are planned to result in a manuscript towards the end of 2025.
WP A2: The technical set up for the dense-sampling approach was established including the use of individualized head cases to minimize head movement and optimize signal-to-noise ratio of the fMRI data. Based on these pilot experiments, a dense sampling was performed and finished for a first participant (15 recording sessions). Based on the results, adjustments of the dense-sampling protocol have been made. Based on the individual anatomical configuration five more participants were recruited for continuation of WP A2. Individual head cases aiming to minimize head movement in the scanners are currently being produced. Technical issues with eye tracking in the MRI scanner were solved. Recordings will start in spring 2025.
WP B1/2: In a first TUS study, participants underwent TUS for 120 s in the target region (pMFC), an active control region (posterior cingulate cortex), or sham stimulation in three separate sessions in counterbalanced order. TUS targeting is done with neuronavigation based on individual head models and the individual results of a preceding fMRI session localizing error-related brain activity. Within 15 min after sonication, the ultrasound gel was removed and an EEG cap with 16 electrodes was applied. Participants performed a speeded flanker task while EEG was recorded. Statistical analysis showed that task performance was better after sonication of the pMFC compared to the sham and control sessions. Surprisingly, no TUS effect on EEG correlates of performance monitoring (ERN, Pe) or response conflict (N2) were found. Analyses of preresponse changes in the EEG dynamics are ongoing. A manuscript is in preparation. One reason for the relatively small effect of TUS may be that relatively little stimulation energy was used, compared with studies in non-human primates. An updated modeling procedure with optimized individual head models has been implemented, which will allow to model the stimulation focus and the acoustic/mechanic and thermic effects better. This will enable us to modify the stimulation protocol in the follow-up study and to increase the energy conveyed to the target area while keeping all parameters well within the safety limits. In the follow-up study five TUS sessions within the same participants are planned. Three target areas are based on prior knowledge about the performance monitoring network. As active control regions serve frontal white matter and lateral ventricle. Preceding MRI sessions are ongoing (T1, PETRA, fMRI). The actual TUS sessions are planned to be started in April 2025.
WP A2: The technical set up for the dense-sampling approach was established including the use of individualized head cases to minimize head movement and optimize signal-to-noise ratio of the fMRI data. Based on these pilot experiments, a dense sampling was performed and finished for a first participant (15 recording sessions). Based on the results, adjustments of the dense-sampling protocol have been made. Based on the individual anatomical configuration five more participants were recruited for continuation of WP A2. Individual head cases aiming to minimize head movement in the scanners are currently being produced. Technical issues with eye tracking in the MRI scanner were solved. Recordings will start in spring 2025.
WP B1/2: In a first TUS study, participants underwent TUS for 120 s in the target region (pMFC), an active control region (posterior cingulate cortex), or sham stimulation in three separate sessions in counterbalanced order. TUS targeting is done with neuronavigation based on individual head models and the individual results of a preceding fMRI session localizing error-related brain activity. Within 15 min after sonication, the ultrasound gel was removed and an EEG cap with 16 electrodes was applied. Participants performed a speeded flanker task while EEG was recorded. Statistical analysis showed that task performance was better after sonication of the pMFC compared to the sham and control sessions. Surprisingly, no TUS effect on EEG correlates of performance monitoring (ERN, Pe) or response conflict (N2) were found. Analyses of preresponse changes in the EEG dynamics are ongoing. A manuscript is in preparation. One reason for the relatively small effect of TUS may be that relatively little stimulation energy was used, compared with studies in non-human primates. An updated modeling procedure with optimized individual head models has been implemented, which will allow to model the stimulation focus and the acoustic/mechanic and thermic effects better. This will enable us to modify the stimulation protocol in the follow-up study and to increase the energy conveyed to the target area while keeping all parameters well within the safety limits. In the follow-up study five TUS sessions within the same participants are planned. Three target areas are based on prior knowledge about the performance monitoring network. As active control regions serve frontal white matter and lateral ventricle. Preceding MRI sessions are ongoing (T1, PETRA, fMRI). The actual TUS sessions are planned to be started in April 2025.
The medial frontal cortex shows a high interindividual variability of gyrification. Previous work has suggested that macroanatomical features such as the presence or absence of a paracingulate sulcus are associated with variability of the size and localization of cytoarchitectural fields, with subtle behavioral differences and with different vulnerability for mental disorders. However, the quantification of gyrification patterns is mainly based on a protocol that has not been changed for nearly two decades. It may be too coarse to capture the large variability found in the population. Therefore, we are currently developing a novel morphometric approach taking into account not only the sulcal length but more spatial and volumetric parameters to enable a more comprehensive characterization of anatomical features. Importantly, these features will be linked to connectivity patterns to enable a comparison of subfields across individuals and to interpret the localization of neural activity.
Traditionally, neuroimaging studies have been carried out in samples of multiple participants by normalizing individual brains to a standard “average” brain. However, this approach does not allow precise anatomical localization and ignores interindividual variability described above. It is therefore not suited to study representations of cognitive control variables which are assumed to be represented in individual patterns that are strongly influenced by the underlying individual anatomy. Here, we will use dense sampling within individual participants. The multivariate data analysis techniques that we develop and apply during the project will open up new avenues to study the representational geometry of representations of cognitive control (and other) variables and to link these to the individual neuroanatomy.
We furthermore apply TUS to systematically map the causal contributions of brain regions found to represent cognitive control variables in the neuroimaging studies described above. With individualized simulations and prior fMRI localizer experiments we optimize targeting and intensity of the stimulation. In addition, the use of active control regions and auditory masking will diminish unspecific acoustic effects of TUS.
Traditionally, neuroimaging studies have been carried out in samples of multiple participants by normalizing individual brains to a standard “average” brain. However, this approach does not allow precise anatomical localization and ignores interindividual variability described above. It is therefore not suited to study representations of cognitive control variables which are assumed to be represented in individual patterns that are strongly influenced by the underlying individual anatomy. Here, we will use dense sampling within individual participants. The multivariate data analysis techniques that we develop and apply during the project will open up new avenues to study the representational geometry of representations of cognitive control (and other) variables and to link these to the individual neuroanatomy.
We furthermore apply TUS to systematically map the causal contributions of brain regions found to represent cognitive control variables in the neuroimaging studies described above. With individualized simulations and prior fMRI localizer experiments we optimize targeting and intensity of the stimulation. In addition, the use of active control regions and auditory masking will diminish unspecific acoustic effects of TUS.