University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Mirco Richter

Doctoral Student in the PhD program since 01.01.2010.


1. Jun.-Prof. Dr. Merhof
2. Prof. Dr. Oliver Deussen

organisational data

Room: Z 924
Tel.: +49 (0)7531 / 88-4636
Other Resources: Personal Page at the Workgroup

project description

Analysis of dementia with structural and functional image data
Alzheimer's disease (AD) and frontotemporal lobar degeneration (FTLD) are the most prominent types of the neurodegenerative disease dementia which affects the structure and function of the human brain. To improve their diagnosis, data acquired by magnetic resonance imaging (MRI), diffusion tensor imaging (DTI) and positron emission tomography (PET) is analyzed to extract features, align them in a common space and apply statistical analyses for group discrimination and identification of possible disease types.

In the first part of this work, the definition of the common space is addressed which is often accomplished by registration to an atlas image that labels main lobes and foldings as anatomically distinct regions. Statistics applied on average cortical thickness values might obscure information such as higher values in gyri than in sulci. To separate these out- and inward foldings, the segmented white matter (WM) was skeletonized and geodesic distances between feature points on the WM surface were used to establish a continuous pruning function that was thresholded to identify gyral and sulcal regions. With the separation of gyri or sulci, classification of dementia could be significantly improved as shown by logistic regression applied to cortical thickness data of different disease groups.

In the second part, complementary features from multiple modalities will be combined as improved discrimination between different types of dementia is expected. From images acquired in a collaborative study, features such as cortical thickness from MRI, WM integrity from DTI, and glucose uptake from PET will be collected. Due to the high variability of AD and FTLD, discrimination of their disease subtypes is a complicated task which requires accurate methods for artifact removal, segmentation and registration to get features in a common space. These will be analyzed by feature reduction methods such as random forests and principal component analysis and classification methods such as LR and support vector machines.

In the last part, an alternative approach to establish gyral correspondences between brains will be developed that is transformation- and template-free. The matching of cortical and subcortical structures will be simplified using different abstraction levels of the WM skeleton and will be guided by geometric and topological similarity features. Designed as an energy-minimization optimization problem, this approach is expected to provide equal or improved robustness as compared to vertex- or surface-based registration.


The following list of publications covers only those, which are or were published during participation at the Graduiertenkolleg / PhD program.

Articles in Journals


Conference Papers


curriculum vitae

Since 01/2010 PhD student, Visual Computing, University of Konstanz
03/2006 - 12/2009 Software developer, AFRA GmbH, Erlangen
10/1999 - 02/2006 Studies of Computer Science (Diplom), University of Erlangen