Electroencephalography is a medical diagnosis technique. It consists in measuring the biopotentials produced by the upper layers of the brain at various standardized places on the skull. Since the biopotentials produced by the upper parts of the brain have an amplitude of about one microvolt, the measurements performed by an EEG are exposed to many risks. Moreover, since the present tendency is measure those signals over periods of several hours, or even several days, human analysis of the recording becomes extremely long and difficult. The use of signal analysis techniques for the help of paroxysm detection with clinical interest within the electroencephalogram becomes therefore almost essential. However the performance of many automatic detection algorithms becomes significantly degraded by the presence of interference: the quality of the recordings is therefore fundamental. This thesis explores the benefits that electronics and signal processing could bring to electroencephalography, aiming at improving the signal quality and semi-automating the data processing. These two aspects are interdependent because the performance of any semi-automation of the data processing depends on the quality of the acquired signal. Special attention is focused on the interaction between these two goals and attaining the optimal hardware/software pair. This thesis offers an overview of the medical electroencephalographic acquisition chain and also of its possible improvements. The conclusions of this work may be extended to some other cases of biological signal amplification such as the electrocardiogram (ECG) and the electromyogram (EMG). Moreover, such a generalization would be easier, because their signals have a wider amplitude and are therefore more resistant toward interference.

Electroencephalography is a medical diagnosis technique. It consists in measuring the biopotentials produced by the upper layers of the brain at various standardized places on the skull.<p><p>Since the biopotentials produced by the upper parts of the brain have an amplitude of about one microvolt, the measurements performed by an EEG are exposed to many risks.<p><p>Moreover, since the present tendency is measure those signals over periods of several hours, or even several days, human analysis of the recording becomes extremely long and difficult. The use of signal analysis techniques for the help of paroxysm detection with clinical interest within the electroencephalogram becomes therefore almost essential. However the performance of many automatic detection algorithms becomes significantly degraded by the presence of interference: the quality of the recordings is therefore fundamental. <p><p>This thesis explores the benefits that electronics and signal processing could bring to electroencephalography, aiming at improving the signal quality and semi-automating the data processing.<p><p>These two aspects are interdependent because the performance of any semi-automation of the data processing depends on the quality of the acquired signal. Special attention is focused on the interaction between these two goals and attaining the optimal hardware/software pair. <p><p>This thesis offers an overview of the medical electroencephalographic acquisition chain and also of its possible improvements.<p> <p>The conclusions of this work may be extended to some other cases of biological signal amplification such as the electrocardiogram (ECG) and the electromyogram (EMG). Moreover, such a generalization would be easier, because their signals have a wider amplitude and are therefore more resistant toward interference.<p> / Doctorat en sciences appliquées / info:eu-repo/semantics/nonPublished