Dagmara Panas PhD

Dagmara Panas

Research Interests

The research interest that first drove me from Physics towards Computational Neuroscience was sleep. This area is still far beyond our full understanding and sports fascinating questions on all levels: from the macrostructure of EEG sleep stages to the cellular and molecular basis underlying (likely, but still hotly debated) memory consolidation. Especially fascinating to me was the intermediate level - the network dynamics underlying neural activity during sleep, which is what motivated my first research project (a Master's project with The University of Warsaw, under the supervision of Dr. Piotr Suffczynski): Statistical analysis of sleep spindles occurrences. Employing Matching Pursuit time-frequency analysis to extract spindles form EEG traces and Maximum Likelihood to estimate parameters of probability density distributions of spindle occurrences, I investigated whether two extant types of sleep spindles are generated by two distinct mechanisms. Mathematical models of thalamo-cortical interaction predict one type of spindles to occur randomly and another type of spindles to recur rhythmically in a deterministic manner, and probability distributions should reflect those two modes of generation. It transpired from the study that, while often topographically and morphologically distinct, sleep spindles could not reliably be distinguished by their statistical properties.

My interest in the behaviour of neural networks is strongly coupled with the interest in information processing and the emergent behavioural outcome, i.e. various aspects of cognition. In particular, I find the Bayesian approach in perception both useful and insightful - the idea of the brain functioning as a sub-optimal observer is simple, elegant, widely applicable and finds increasing support in studies. This, in turn, had motivated my Master's research project with the Doctoral Training Center, under the supervision of Dr Peggy Series: Specificity of prior expectations in visual statistical learning; a psychophysical study of the process of implicit acquisition of statistical knowledge. My aim was to assess whether the brain can detect statistical independence of features within simple stimuli and, if so, whether it uses the acquired information in a Bayesian manner. However, in our data it appeared that expectations acquisition was in competition with adaptation and this interfering effect rendered our results inconclusive.

While my interests cluster around various computational, modelling and statistical approaches in application to understanding human brain function, I was also keen to explore a more basic avenue of research. This has led me to the current doctoral project: Model-based analysis of stability in networks of neurons. Under the supervision of Dr. Matthias Hennig and co-supervised by Dr. Luca Berdondini I am investigating the spontaneous activity of cultured hippocampal neurons and the relationship between single-neuron and group stability. The results of this research are currently submitted and pending peer review.


Model-based analysis of stability in networks of neurons (PhD)