Our research focuses on understanding how the brain shapes behavior. Rather than operating through isolated regions or small neuron clusters, the brain relies on complex interactions within its networks. Additionally, the brain is dynamic, processing and integrating the temporal histories of these network interactions, known as brain dynamics. In psychiatric disorders, brain network interactions and dynamics deviate from optimal functioning, and treatments can be viewed as attempts to restore these networks to a more optimal state.
Our research program addresses this through two main directions: (1) investigating brain network interactions and dynamics, and (2) quantitatively analyzing behavior using computational methods.
To study brain network interactions and dynamics, we use two approaches: (1) identifying patterns of network interactions and dyanmics, and (2) testing causal roles by modulating these interactions. For the first approach, we analyze functional neuroimaging data using network analysis methods to better understand brain network functions (e.g., Hein, Morishima et al., Science, 2016; Nakataki et al., Neuropsychopharmacology, 2016; Ishida et al., Frontiers in Psychiatry, 2021). Recently, we have incorporated temporal dynamics into our analyses, revealing altered brain network dynamics in individuals at clinical high risk for psychosis (Kindler, Schizophrenia Bulletin Open, 2024). For the second approach, we are developing new frameworks to modulate brain network interactions using non-invasive brain stimulation techniques (e.g., Fehér et al., Frontiers in Human Neuroscience, 2017; Brain Connectivity, 2021; Zimmermann, in revision). We also study psychiatric disorder psychopathology by combining sophisticated cognitive tasks with brain stimulation (Koshikawa et al., Psychiatry and Clinical Neurosciences, 2022).
In our computational approach to quantifying behavior, we leverage advancements in machine learning, including deep learning and natural language processing (NLP). These techniques extend beyond simple classification, allowing us to quantify complex behaviors—such as thoughts and emotions—with precision and reproducibility. In psychiatry, symptom evaluation is traditionally subjective and prone to inter-rater variability. Computational methods help address this issue by providing more objective measures. Our initial study combining NLP with neuroimaging revealed significant individual differences in accessing semantic information, which are linked to cerebellar grey matter volume (Zengaffinen, 2023). We are also investigating acoustic analysis to quantify complex emotions, such as embarrassment, a key emotion in social anxiety (Sipka, under review).
