Pattern Classification Methods for Dementia Analysis
The information spaces currently covered by the GRK 1042 involve textual information, personal information management, multimedia and graphical objects, as well as dynamic processes. The goal of this research proposal is to extend the scope of the GRK to other large information spaces, namely in medical imaging. In the following paragraph an example of a large information space in medical imaging is presented which we intend to address in the future.
Large medical image databases make it possible to observe complex disease specific patterns such as in dementia. The main challenge in a real-world application is to develop a robust classification approach in the presence of noise and variations in the data due to different imaging protocols and scanner manufacturers. An analysis could benefit from reducing the dimensionality of the data (e.g., based on principal component analysis) prior to application of pattern classification and machine learning approaches for disease classification. In order to provide a robust classification, multiple biomarkers of dementia will be analyzed at the same time, e.g. patterns of dementia obtained from molecular imaging data, or shape and size biomarkers of particular structures extracted from anatomical image data such as MRI.
|Deutsche Forschungsgemeinschaft||501/09||GRK 1042/2||01.01.2009 – 30.06.2013|
|Period:||01.07.2010 – 30.06.2013|