Håkan Hall poster abstract

Data mining in schizophrenia in the Human Brain Informatics project

H. Hall, S. Arnborg, I. Agartz and G. Sedvall

Karolinska Institutet, Department of Clinical Neuroscience, Karolinska Hospital, Stockholm, Sweden

In the HUBIN (Human Brain Informatics) project a relational database is established at the Karolinska Institute on human brain data.

This study combines molecular genetic, phenotypic, brain imaging (MRI) and environmental data for schizophrenia patients and healthy control subjects resulting in a large number of clinical and biological variables.

The project is based on the assumption that cross-domain analyses using data mining approaches can provide new hypotheses much more efficiently if many different domains are investigated on the same subject population. Using a variety of data analysis methods a better understanding of brain structure and function in neuropsychiatric diseases may be achieved.

In this study detailed volumetric data from MRI studies are compared with molecular genetic data, disease state and other variables obtained. Several aids are available for deciding the most relevant projections: principal component analysis, dimensionality reduction, mixture identification, autoclass identification etc. The evaluation of these procedures, as well as results obtained, such as a correlation of regional brain volumes, serum enzyme levels and other variables between the different study groups, will be described.

View poster