Novel approaches to neuroimaging
Machine learning
In recent years machine learning, the study and development of data analysis tools capable of extracting patterns from datasets, has become increasingly important in neuroscientific research. With regard to the motor system, new research has shown that machine learning tools can be successfully applied to fMRI data for the classification of fine motor movements, and that the obtained results enable insights into the mechanisms underlying motor control and learning. Our goal is to broaden this research direction by applying suitable machine learning methods not only to the analysis of neuroimaging data from the motor system, but to integrated sensorimotor imaging data.
Computational models
One of the issues plaguing Cognitive Neuroscience is the use of nebulous terms such as ‘attention’, ‘reward’ and ‘learning and memory’. Without precise definitions it is not possible to accurately investigate these meta-cognitive phenomena. However, by converting these terms into mathematical equations we are able to more precisely define, and as such investigate, these processes. Here, we implement existing computational models such as Rescorla-Wagner and the Hierarchical Gaussian Filter in order to more precisely investigate brain-behavior interactions.
Along with computational models of behavior we also use computational models as tools for advanced data analysis. This includes classical statistical techniques, machine learning models such as support vector machines, nearest neighbor classifiers and deep learning networks to classify neuroimaging data. We are currently implementing these methods in the study of sensorimotor integration, and the classification of psychiatric disorders based on resting state connectivity. We will also develop suitable computational models for the analysis of other types of neuroimaging data.
Multi-modal imaging
Multi-modal imaging combines two or more imaging techniques, which allows the integration of the strength of individual modalities, while overcoming their limitations. For example, by combining fMRI with EEG we benefit from both the high spatial resolution of fMRI and high temporal resolution of EEG, resulting in a better estimate of brain connectivity.