| Abstract |
Electroencephalograms are used as routine examination in the diagnosis of neurological diseases. The aim of this work was to create a computer-aided diagnostic system for clinical EEGs. This was achieved by a hybrid combination of different methods. First, spectral properties of the EEG signals are extracted by a new method based on autoregressive spectra of short time intervals. EEGs are particularly important for the diagnosis and control of epilepsy. To detect spike-wave complexes which are typical for epilepsy, a pattern recognition method known as Active Shape Models was adapted to EEG signals. Finally, a rule-based expert system gathers all pieces of information computed by the former methods and infers facts about physiological and pathological properties of the EEG. The output comprises pseudo-verbal phrases, colored markers of the ongoing EEG and 3D maps, both highlighting important spatio-temporal activities. |