Short term fellowships (1999-2003).
Final of Jane Neumann
Results:
The purpose of the visit was to lay the foundations for the application of neural network modelling methods to fMRI data, an area of research which requires knowledge of both the nature of fMRI data and the methods and tools provided for connectionist modelling.
Our work focused on the interaction of modelling and imaging work for investigating the functional neuroanatomy of attention. Selective attention enables us to concentrate on specific locations or specific features in our environment. Recently reported event-related fMRI experiments indicate that specific brain areas are involved in the control of attentional resource allocation, whereas other brain areas change their activation levels in response to these changed attentional foci. For example, increased activation in the left frontopolar cortex was observed when subjects had to reallocate attentional resources, either between visual dimensions such as colour and motion or between locations (Pollmann, 2001; Pollmann et al, 2000).
While existing connectionist models for selective attention tasks are able to simulate the observed behaviour and response times of the subjects (see e.g. Mozer & Sitton 1998)), they do not take into account increased activation levels of specific brain areas as described above. We developed first ideas for a connectionist model of the tasks carried out in the event-related fMRI experiments. The different functionalities of particular brain areas is captured by the modular architecture of the model and the interaction between the network modules.
Bibliography
- Mozer, M.C., Sitton, M. 1998. Computational modeling of spatial attention. In: Pashler, H. (Ed). Attention, pp 341-388. East Sussex, UK: Psychology Press.
- Pollmann, S. 2001. Switching between Dimensions, Locations, and Responses: The Role of the Left Frontopolar Cortex. NeuroImage 14, pp118-124.
- Pollmann, S., Weidner, R., Mueller, H.J., von Cramon, D.Y. 2000. A Fronto-Posterior Network Involved in Visual Dimension Changes. Journal of Cognitive Neuroscience 12:3, pp 480-494.
Collaborative and/or training aspects:
The research visit was extremely useful as it allowed me to be introduced to the imaging techniques used at the Clinic of Cognitive Neurology and the particular neuro-scientific goals of Prof. Pollmann's research group. I was able to familiarise myself with the specifics of fMRI data and related data analysis tools and gained a better understanding of how imaging data can be incorporated into connectionist models. A number of very valuable seminars and discussions with leading researchers in the field of cognitive neuroscience gave me a deeper insight into the research issues and methods available in what is for me a new research area. In addition, I was given the chance to give a talk to researchers in the host institute introducing the general concepts of connectionist networks and the particular aspects of connectionist architectures relevant to models of cognition.
Expected future impact of visit:
Our future collaboration will aim for the development of a fully functional connectionist model of the event-related fMRI experiments described above. Such a model should not only lead to a better understanding of the functionality of the particular brain areas involved, but also allow predictions that potentially form the basis for further experimental imaging work. Moreover, taking up a research position at the Max Planck Institute for Cognitive Neuroscience in Leipzig, I will be actively involved in the further development of LIPSIA (Leipzig Image Processing and Statistical Inference Algorithms), the software package for the analysis of experimentally obtained fMRI data. We intend to augment the existing mechanisms with connectionist methods, thereby extending the functionality of the software to enable the connectionist modelling of experimentally observed fMRI data.