«PEKKA TIIHONEN Novel Portable Devices for Recording Sleep Apnea and Evaluating Altered Consciousness Doctoral dissertation To be presented by ...»
3.2.2 Electroencephalography The human nervous system can be divided into two parts: the central and peripheral nervous systems. Anatomically, the central nervous system (CNS) consists of the cerebrum, cerebellum, brainstem and spinal cord (Heimer 1995). The cerebrum has two hemispheres which are interconnected by the corpus callosum. Both hemispheres have four general subdivisions: the frontal, temporal, parietal and occipital lobes. The basic element of the nervous system is the neuron (Niedermeyer 1993). Neurons transmit information to other neurons through synapses in a chemical form. Post synaptic potentials (PSPs) in thousands of synchronously active pyramidal cells (e.g. group A in layer 5 in Figure 3.1) of the cerebral cortex produce electrical potentials which are recordable on the scalp with an amplifier through appropriate electrodes. Different information is processed in different functional areas of the cerebral cortex, and if the Figure 3.1: A simplified picture of the origin of EEG and its recording with scalp electrodes and a preamplifier. The polarity of surface EEG depends on the location of the synaptic activity within cortex. A Group A receives excitatory input from thalamus in layer 4 causing a positive potential in Electrode A and in contrast Group B receives an excitatory input from the contralateral cortex via corpus callosum in layer 2 causing a more negative potential in Electrode B.
active cortex is small and perpendicular to the scalp surface, there may be only a minor voltage difference that is seen at the inputs of the amplifier. Cerebrospinal fluid and the scalp are much better conductors than the skull. These three layers attenuate, filter and Kuopio University Publications C. Natural and Environmental Sciences 261: 1 - 79 (2009)
- 37 Tiihonen, P.: Novel Portable Devices for Recording Sleep Apnea and Evaluating Altered Consciousness smear the potentials measurable on the scalp. The amplitude of a typical scalp EEGsignal varies from 10 to 100 PV and when measured from subdural electrodes it varies from 10 to 20 mV (Aurlien 2004). The main frequency content of the scalp EEG-signal lies between 0.16 and 70 Hz (Nuwer 1998). A typical example of spontaneous EEG is seen in Figure 3.2. Conventionally, EEG frequency bands have been defined as described in Table 3.2.
Figure 3.2: Example of 8 seconds of spontaneous EEG recorded from a healthy subject with eyes closed while awake.
Electrodes Fz, Cz, Pz, and Oz were placed according to the international 10-20 system referred to the right mastoid.
Table 3.2: The definitions of frequency bands of EEG (Niedermeyer 1993, Rampil 1998).
Clinical electroencephalography devices should comply with the general standards for recording of the digital electroencephalogram (Nuwer 1998). However, these standards should not restrict research use of EEG, or its use outside EEG laboratories and thus exceptions can be accepted. For example, the standard suggests that at least 24 and preferably 32 EEG channels should be used, but normal bispectral (BIS) or entropy monitors use only two or a maximum of four frontal channels. However, most of the recommended values are universally valid. The minimum digital sampling rate is 200 Hz and the corresponding anti-aliasing filter prior to digital sampling is 70 Hz (roll-off rate at least 12 dB/octave). Digitization must use a resolution of at least 12 bits and must be able to resolve the EEG-signal down to 0.5 PV. Preamplifier input impedances should be more than 100 M: and interchannel cross-talk less than 1% of signal Kuopio University Publications C. Natural and Environmental Sciences 261: 1 - 79 (2009)
- 38 Measurement of Lowered Consciousness amplitude. The common mode rejection ratio (CMRR) should be at least 110 dB and noise should be less than 0.5 PVRMS (RMS is root mean square). In the recording, the
contact impedances should be less than 5 k:.
Basically EEG artefacts are voltage changes that do not originate from the patient’s brain. External technical artefacts can arise from other electrical equipment and a qualified EEG technician must be able to recognise when the EEG is distorted due to technical reasons. Physiological artefacts (eye movements, muscular activity etc.) are generated from the patient him/herself and they can be sometimes used as signs of consciousness level. In intensive care units, artefacts due to other equipments are generally more serious than those encountered in laboratories designed for clinical EEG and ERP recordings (Jordan 1999, Young 1999).
EEG is irreplaceable in diagnosing epilepsy or determining the level of sedation when a status epilepticus patient is being given with general anaesthesia. Recording of EEG has an important role in investigation of encephalitis, Creutzfeld-Jakob disease or paroxysmal events with unknown etiology and when diagnosing brain death. While the unconscious patient may be investigated with other methods, EEG can provide additional diagnostic information and help in estimating the prognosis.
3.2.3 Auditory evoked potentials
Sensory stimuli produce time-locked electrical activity that can be recorded on the scalp. These evoked potentials can be classified according to the applied stimulus into three main categories: somatosensory (SEP), visual (VEP) or auditory (AEP) evoked potentials. Evoked potential recordings can be used to localise lesions, or monitor a sensory pathway during surgical procedures. The functioning of the auditory pathway (Figure 3.3) can be measured with evoked potentials which are usually divided into three subcategories according to their latencies. Short-latency potentials (below 10 ms) originate from the brainstem and the recording of these potentials is called brainstem auditory evoked potential (BAEP) recording. When the latencies are between 10 ms and 50 ms, the test is called middle-latency evoked potential (MLAEP) recording. The first negative peak (around 20 ms) of MLAEP may be of thalamic origin and the later peaks probably represent cortical phenomena (Pockett 1999). Both BAEP and MLAEP use broadband clicks as stimuli at a rate of about 10 Hz. Longer latency evoked potentials (LLAEP) (over 50 ms) form their own category and they are commonly named as event-related potentials (ERPs) (Goodin 1994). The stimulation for ERPs is usually done with beep-like sounds (i.e. sine-wave tones).
Comparison of Figures 3.2 and 3.4 show that the baseline EEG amplitude is much larger than the ERP responses. Thus, signal averaging is needed in order to obtain clean representative ERP curves. This is accomplished by summing at least several tens of original single responses and dividing the sum by the number of responses. This averaging operation can be done in real time or by means of an offline analysis after the recording. Offline analysis has the advantage that artefacts can be rejected mathematically or by visual selection. The basic assumptions of signal averaging are that the response to a sensory stimulus does not change during stimulation (i.e. it is deterministic) and that the background EEG activity is stochastic (a zero-mean process with variance that does not change during the recording). These assumptions are not always valid, since habituation may affect the responses during the recording (Ritter 1968).
Figure 3.3: Simplified schematic presentation of the auditory pathway, adapted from Felten (2003) and Martin (2003).
Transversal sections of medulla oblongata and midbrain and a coronal section of the cerebrum are shown along the route of the auditory pathway.
The ERPs measure the functional state of the cortex. They have exogenous components determined mainly by the physical properties of the stimuli (frequency, intensity and duration) and endogenous components determined by the task requirements and instructions, the psychological relevance of the stimulus and the subjects’ psychological state (Donchin 1978). Three main components can be seen in auditory ERP measurements: a negative peak around 100 ms (N100) (Figure 3.4), a mismatch negativity (MMN) response around 100 - 200 ms (Figure 3.5), and a positive peak around 300 ms (P300) (Sutton 1965), (Figure 6a in Study IV). N100 and MMN responses are of exogenous origin and P300 is of endogenous origin (Donchin 1983).
Kuopio University Publications C. Natural and Environmental Sciences 261: 1 - 79 (2009)
- 40 Measurement of Lowered Consciousness The N100 response is best seen in the central region (i.e. near the Cz electrode when referred to M1 or M2). It reflects mainly the activation of the auditory cortices in the temporal lobe and it is a sum of several functionally distinct and temporally overlapping subcomponents (Vaughan 1970, Wolpaw 1975, Hari 1982, Näätänen 1987). The MMN response appears only after an infrequent stimulus that differs from the ongoing sound stream of frequent stimuli. It is considered to indicate the presence of a memory system that stores the physical characteristics of the frequent stimulus (Näätänen 1978, Näätänen 1992). It has been suggested that frontal brain areas play an important role in the generation of MMN (Giard 1990, Alho 1994, Picton 2000). When the subject attends to infrequent target tones, the P300 response is generated. It is suggested to reflect higher brain functions related to cognitive stimulus evaluation, memory and attention. The P300 wave has its maximum at the parieto-occipital electrodes (Polich 1999). A smaller P300 fronto-central response is also generated by an unattended situation (Squires 1975, Plourde 1993).
Figure 3.4: Passive oddball ERP responses to standard tones from a subject who was a free diver and who volunteered to undergo a hypoxia exercise.
The recording (authorized by the local ethics committee) was done with the device developed in Study IV. During hypoksia N100 component is delayed.
It has been shown that deepening of the level of sedation increases the latency but decreases amplitude of the N100 (Yppärilä 2002a). Thus, these measures may be used as measures of the depth of sedation. In a recent study (Westeren-Punnonen 2005), the auditory ERPs were found to recover faster than the EEG after ventricular fibrillation.
This suggests that ERPs could have potential advantage over EEG in the prognosis of the recovery of brain functions after trauma or surgery.
In addition to monitoring the level of consciousness, ERPs have shown promice in investigations of disorders in which consciousness is impaired. For example, prolonged latencies of P300 have been found in Alzheimer’s disease (Polich 1990). The lengthening of the P300 latency is also found in healthy individuals who are at an
increased risk of acquiring Alzheimer’s disease, suggesting the possibility for early screening for this disease (Saito 2001). A decrease in the amplitude of the P300 has been found in Parkinson’s disease (O'Donnell 1987), schizophrenia (O'Donnell 1999) and depression (Gordon 1986). Furthermore, it has been shown that easily distractible adolescents have enhanced frontal and reduced parietal P300 amplitude compared to non-distractible adolescents (Määttä 2005).
Figure 3.5: Passive oddball ERP responses to deviant tones from a subject who was a free diver and who volunteered to undergo a hypoxia exercise.
During hypoksia MMN response clearly delays.
Technical requirements for ERP recording devices are similar to those of EEG (Nuwer 1998) and evoked potential (Nuwer 1994) devices. In addition, the stimulator of an ERP recording device must be able to produce different auditory stimulus sequences at a volume level adjusted according to the hearing level of the patient. The time stamps of stimulus events should be precisely handled with respect to incoming EEG in order to provide accurate averaging. For most ERP purposes, an A/D converter using 12 bits (4096 values) is sufficient, provided that the incoming signal is amplified properly (about 10000 times) to range over at least 8 bits (Picton 2000). The artefacts to be avoided in ERP recordings are virtually the same as those described in Chapter 3.2.2 for EEG recordings.
Sleep apnea syndrome is a highly underdiagnosed disease with potentially serious consequences on both the health of the individual and on national health care budgets.
The underdiagnosis is partly attributable to the lack of affordable and reliable devices suitable for telemedical diagnostics of the disease. The event-related potentials have shown good promise for diagnosis of depth of sedation and level of consciousness.
However, there is scarcity of portable devices that would enable the measurement of event-related potentials in intensive care units. The main aim of this thesis was to
respond to this demand for appropriate diagnostic devices. The specific objectives were:
1. To design, construct and evaluate a compact, technically reliable and easy-to-use eight-channel (Type 3) device (Venla) suitable for diagnostics of sleep apnea.
2. To investigate the diagnostic and technical reliability of a portable seven-channel (Type 3) recording device (APV2) suitable for clinical use in hospital and at home.
3. To investigate the suitability of analysis software primarily designed for one specific make of portable monitoring device for use with other devices and to evaluate the reliability of automatic analysis in detection of mixed and central apneas.
4. To design, construct and evaluate a compact, portable and automatic five-channel auditory ERP recording device (Emma) suitable for clinical use in intensive care units and emergency rooms.
This thesis consists of four independent studies (I – IV). In this section, the materials and methods used in the studies are summarized. A summary of the study design is presented in Table 5.1.
Table 5.1: Materials and methods used in Studies I – IV.