Parameterisation comparison for the detection of panic disorder using time-frequency transforms and support vector machines

Dietl, H. and Weiss, S. (2004) Parameterisation comparison for the detection of panic disorder using time-frequency transforms and support vector machines. In: 2nd International Conference on Advances in Medical Signal and Information Processing, 2004-09-05 - 2004-09-09.

[img]
Preview
PDF
dietl04e.pdf
Accepted Author Manuscript

Download (203kB)| Preview

    Abstract

    In this paper we compare the effect of the parameterisation on the automatic detection of diseases based on biomedical data. Exemplarily, we study the analysis of event related brain potentials in patients suffering from panic disorder, whereby the data comprises responses to neutral and panic causing stimuli. This data is parameterised by time-frequency (TF) transforms, from which features are selected by a statistical test. The selected features represent the input to a support vector machine classifier yielding a detection rate for the TF parametrised data. This is compared with detection rates obtained for unparameterised time domain data