210
Biomedical Signal and Image Processing
When the normal brain pattern with standard frequency content is alternated by
periods of EEG signal that have an extremely low frequency content (flat EEG), this
is identified as burst suppression. Burst suppression has been linked to a diminished
metabolic function of the cortical neurons. This is usually an indication of deprived
oxygen supply to the brain. Additionally, in anesthesia monitoring, this burst suppression is often seen in relation to the dosage changes of various anesthetic drugs.
EEG monitoring to determine the consciousness of a patient predominantly focuses
on the spectral content of the EEG. This monitoring can take place either in the operating room (OR) during or after surgery, or in the intensive care unit (ICU) for a patient
that may appear to be in a coma. The existence of more significant higher frequency
contents in an EEG signal generally indicates a more alert state of consciousness.
10.7 PROCESSING AND FEATURE EXTRACTION OF EEG
In order to maximize the information extracted from the EEG signal, the signals
need to be analyzed and characteristic features must be revealed. Some typical methods used for processing and featuring extractions of an EEG are described next. We
start this section with the description of typical noise sources for EEG signal.
10.7.1 SOURCES OF NOISE ON EEG
Several features will provide signals that are much more powerful than the EEG
signals, such as muscle contractions and eye movement. The eye is a large dipole and
as such produces a large signal under ordinary conditions. The eyes move left to right
at approximately 10 times per second to prevent saturation of the rods and cones in
the retina. Both eyes move, by definition, in unison and will provide both left and
right eye signals in both left and right brain. This phenomenon can be eliminated
by placing the reference electrode on the nose, thus providing a signal cancellation
through electronic processing. Using differential amplification, the eye movement
will be expressed in all electrodes and will be rejected by the common mode rejection factor of the amplifier.
Another potential source of noise is electrode motion. Each electrode forms an electrochemical equilibrium with the skin of the head, and when the electrode is moved,
the equilibrium will need to be reestablished. This type of noise is easily identified
because of the magnitude of the motion artifacts but, at the same time, will require
considerable time to reestablish equilibrium. This motion artifact caused by the electrodes can also be compensated for using signal processing if the range of frequency
for the motion artifact is different from the measured EEG signal. For instance, since
this motion often has a frequency less than a few Hz, a low-pass filter can eliminate
most of the noise without affecting the high-frequency waves in the EEG.
Another frequency artifact observed in the EEG is due to the temperaturedependent nature of this signal. Generally, a slowing of the EEG activity is observed
during hypothermia, when the body temperature drops below 35°C. This does not
necessarily mean the patient has suffered any brain damage. This is why it is often
recommended that EEG signals are taken in a room with a fixed temperature and
illumination settings.
Biomedical Signal and Image Processing
When the normal brain pattern with standard frequency content is alternated by
periods of EEG signal that have an extremely low frequency content (flat EEG), this
is identified as burst suppression. Burst suppression has been linked to a diminished
metabolic function of the cortical neurons. This is usually an indication of deprived
oxygen supply to the brain. Additionally, in anesthesia monitoring, this burst suppression is often seen in relation to the dosage changes of various anesthetic drugs.
EEG monitoring to determine the consciousness of a patient predominantly focuses
on the spectral content of the EEG. This monitoring can take place either in the operating room (OR) during or after surgery, or in the intensive care unit (ICU) for a patient
that may appear to be in a coma. The existence of more significant higher frequency
contents in an EEG signal generally indicates a more alert state of consciousness.
10.7 PROCESSING AND FEATURE EXTRACTION OF EEG
In order to maximize the information extracted from the EEG signal, the signals
need to be analyzed and characteristic features must be revealed. Some typical methods used for processing and featuring extractions of an EEG are described next. We
start this section with the description of typical noise sources for EEG signal.
10.7.1 SOURCES OF NOISE ON EEG
Several features will provide signals that are much more powerful than the EEG
signals, such as muscle contractions and eye movement. The eye is a large dipole and
as such produces a large signal under ordinary conditions. The eyes move left to right
at approximately 10 times per second to prevent saturation of the rods and cones in
the retina. Both eyes move, by definition, in unison and will provide both left and
right eye signals in both left and right brain. This phenomenon can be eliminated
by placing the reference electrode on the nose, thus providing a signal cancellation
through electronic processing. Using differential amplification, the eye movement
will be expressed in all electrodes and will be rejected by the common mode rejection factor of the amplifier.
Another potential source of noise is electrode motion. Each electrode forms an electrochemical equilibrium with the skin of the head, and when the electrode is moved,
the equilibrium will need to be reestablished. This type of noise is easily identified
because of the magnitude of the motion artifacts but, at the same time, will require
considerable time to reestablish equilibrium. This motion artifact caused by the electrodes can also be compensated for using signal processing if the range of frequency
for the motion artifact is different from the measured EEG signal. For instance, since
this motion often has a frequency less than a few Hz, a low-pass filter can eliminate
most of the noise without affecting the high-frequency waves in the EEG.
Another frequency artifact observed in the EEG is due to the temperaturedependent nature of this signal. Generally, a slowing of the EEG activity is observed
during hypothermia, when the body temperature drops below 35°C. This does not
necessarily mean the patient has suffered any brain damage. This is why it is often
recommended that EEG signals are taken in a room with a fixed temperature and
illumination settings.