EEG Acquisition And Automatic Seizure Detection System

EEG is the electrical signal produced by the neurons in brain during their activities. It can be recorded by measuring the voltage between any two electrons placed on or beneath the scalp. EEG can be used to diagnose diseases of brain, and it is also an important tool for cognitive science and psychophysiology. A portable EEG acquisition device plays a significant role in nursery, academic research and brain-computer interface system. Epilepsy is a brain disorder which annoys a certain amount of people. It causes both physical and psychological damage to patients. Real-time seizure warning system could provide a great help.As mentioned above, this thesis focuses on three main aspects:the first is to implement a design of portable EEG acquisition system, which can be applied to various research fields and products; the second is to explore a better algorithm used in automatic seizure detection, which has a good performance and meanwhile a lower complexity; the third is to implement the real-time seizure warning system, making the algorithm to reality.Firstly, this thesis gives a design of EEG acquisition system. It can be divided into two parts:an acquisition device and a software application on host computer. The acquisition device is implemented by two methods, one based on TGAM module and the other ADS1299. The former can only collect a single channel, but it has a simpler circuit and a good resistance to noise. The latter has more channels and more precise A/D converters. The latter uses digital comb filters to restrict noises of 50Hz and 100Hz. The circuits of these two methods are both in smaller scales, which is a significant feature to be portable. The application on host computer is used to show the EEG wave, the spectrum and to record data. It is designed and realized in the object-oriented paradigm, which makes each module separate and independent. It benefits a lot when to reuse these codes and to add a new module. It is written with Qt, which makes it easier to transplant to other platforms or operation systems.Then this thesis gives a fully evaluation of the Generic Osorio-Frei Algorithm (GOFA) and describes the analysis in detail. To well suit to real application environment and overcome the over-sensitivity of artifacts in the algorithm, a new parameter Dterm is added and a new multi-channel discrimination rule is introduced into this study. This algorithm can get a better sensitivity and a lower detection delay with a low complexity.Finally, the automatic seizure detection algorithm is combined with the EEG acquisition system to form a real-time seizure warning system. With a good software structure, the automatic seizure detection module can be easily added. During the evaluation, a virtual data source is used instead of the real EEG acquisition device. The result shows that the system gives a warning at the right time as same as the simulation of the algorithm, and that the system completes the calculation just in time.