Imaging the Serotonin system

Project "Multimodal brain imaging of the serotonin system" funded by the Danish Council for Independent Research | Medical Sciences

Over the last decades, an immense progress in mapping the human brain structure and function has been made. Much less has been done to map the brain neurochemical signaling systems. Serotonin (5-HT) is an important modulator in, e.g., mood, appetite and diurnal regulation and is involved in frequent neuropsychiatric disorders. With this computationally advanced project, we will utilize unique existing molecular, functional and structural brain imaging data from healthy individuals to generate a high-resolution atlas of the human brain 5-HT system and to examine the functional connectivity between the 5-HT "control center" and corresponding brain regions. The 5-HT functional brain connectivity will be examined in winter depressed, in their symptomatic and asymptomatic phases. The outcome of this project will be to provide open-access tools and a highresolution normative database atlas, and novel insight in the brain 5-HT functional connectivity patterns will be gained.


Link to the publicly available serotonin atlas:

Project "Exploring a multimodal serotonin atlas with the help of machine learning methods" funded by the Lundbeck foundation

Neuroimaging is an interdisciplinary field covering areas of biology, chemistry, computer science, mathematics and medicine and has contributed immensely to our understanding of the nervous system. In the proposed project, we want to explore how machine learning techniques can be applied to robustly identify variations in the human brain as determined by molecular and structural brain imaging obtained with functional Positron Emission Tomography (PET) and structural Magnetic Resonance Imaging (MRI). Specifically, we want to utilize a unique multi-modal data set of the serotonin neurotransmitter system that has been collected at the Center for Integrated Molecular Brain Imaging (Cimbi). It consists of approx. 200 high-resolution brain PET scans of healthy volunteers that target 4 different serotonin neuroreceptors (5-HT1A, 5-HT1B, 5-HT2A, 5-HT4) as well as the serotonin transporter (SERT), paired with the corresponding MR scans. This high dimensional neuroimaging data will be examined by machine learning tools such as clustering methods or multivariate regression and utilized to build a population-based atlas of the serotonin system that can be used as a reference and for constructing biomarkers of brain disease.

In (a) we show the results of a univariate two-tailed t-test of patients having a smaller cortical thickness than controls. This needs to be corrected for multiple comparisons.In (b) the results of a cluster-wise multiple comparison correction as done with FreeSurfer and a voxel-wise threshold of 0.01 and a cluster wise threshold of 0.01 are displayed. Only two regions, the insula and the temporal cortex, survive.On the contrary, a multivariate feature selection shown in (c) reliably identifies additional regions that were found in the intersection of feature selections performed over 5 independent subsets. This case demonstrates how the multivariate analysis can add information to a univariate analysis.