«PEKKA TIIHONEN Novel Portable Devices for Recording Sleep Apnea and Evaluating Altered Consciousness Doctoral dissertation To be presented by ...»
7.1 Novel low-cost ambulatory devices for screening of sleep apnea It has been estimated that 9 - 24% of the middle-aged population suffers from undiagnosed OSA (Young 1993, Young 1997). Since undiagnosed OSA increases significantly the risk of many cardiovascular diseases and mortality (Wilcox 1998, Shahar 2001, Marin 2005, Valham 2008), it is of great importance to develop an inexpensive and straightforward means for diagnosing OSA.
In Study I, a novel 8-channel Type 3 device (Venla) for screening of OSA was introduced. Prior to its in vivo application, the device was evaluated with extensive technical tests and was found to meet strict electromagnetic emission limits (Williams 1996). The first part of the clinical evaluation was arranged by making 19 recordings simultaneously with Venla and a clinical reference instrument (Embla) in a sleep laboratory.
Similarly, in the first part of Study II, a novel ambulatory Type 3 device (APV2) was evaluated by conducting 10 simultaneous measurements with this device and a clinical reference instrument (Embla). Both evaluated devices (Venla and APV2) showed similar diagnostic capabilities as the reference equipment in detecting sleep apnea. They also showed no statistical difference to the reference device in the evaluation of the apnea-hypopnea or oxygen desaturation indices. In addition, the AHI and ODI values determined with the devices and the reference instrument were highly linearly correlated.
The second part of the clinical evaluation in Study I was arranged by conducting 323 ambulatory home recordings of which 275 (193 males and 82 females) were diagnostically acceptable. The Venla device and the commercial reference device (Embletta) showed similar diagnostic capabilities in detecting sleep apnea. No statistical differences were found between the devices in the estimation of the apnea-hypopnea or oxygen desaturation indices.
Good technical reliability is a key issue for a sleep-recording device when screening large populations, since unsuccessful recordings increase economic costs and queuing times. In the second parts of Studies I and II, the clinical home recordings of Venla (N = 106), APV2 (N = 149) and Embletta (N = 169) were evaluated by comparing the diagnostic quality of recordings. The result was that 89.1, 90.0 and 77.2% of the recordings were technically perfect, 93.3, 96.0 and 80.8% were diagnostically acceptable and 6.7, 4.0 and 19.2% totally failed with Venla, APV2 and Embletta, respectively. The results indicate that APV2 and Venla are technically more reliable
than Embletta, a device which is widely used in sleep apnea diagnostics. Furthermore, when compared to the percentage of successful recording (76.6 – 94.4%) reported in the literature for current commercial devices (Whittle 1997, Portier 2000, Lloberes 2001, Gagnadoux 2002, Dingli 2003), the novel devices demonstrated good technical reliability.
The technical reliability of the recording depends on the transducers used. The ease of use is especially important, since in many cases the patient must be able to operate the equipment by themselves without technical support. Usually patients see the equipment and transducers for the first (and probably the last) time in their life and must attach the transducers to themselves according to written instructions. Thus, the transducers must be easy to attach and they must stay firmly in place regardless of movements throughout the night. For these practical reasons an oronasal thermistor and strain gauge respiration movement sensors were chosen for incorporation into the Venla device.
In addition, there is very limited possibilities to confirm that the transducer is working properly when it is set in its place. For example, when the oxygen saturation transducer is set, there is no way to confirm that it is working properly unless there is a digital display to show the saturation and pulse signals. Venla and APV2 both contain an LCD screen to help to confirm that the oxymeter is working properly. For the real time verification of airflow and respiration movements, there is a coarse graphical display of the signals on the LCD. Furthermore, the display showing the status of the transducers is handy when the transducers are cleaned and checked after recording, since mechanical or electronic problems can be immediately confirmed and corrected.
Several studies have shown that a nasal pressure cannula is a better transducer for detecting hypopneas than a thermistor (Norman 1997, Ballester 1998, Series 1999, Hernandez 2001, Teichtahl 2003, BaHammam 2004). This follows from the fact that the square root of the differential pressure derives directly the volume flow rate, (Equation (10) in Chapter 2.4.2). In contrast, a thermistor will produce a flat signal, regardless of airflow, if the temperature of exhaled airflow is same as the ambient temperature.
Furthermore, it must be noted that a thermistor records the time derivative of the airflow. However, a nasal cannula cannot record mouth breathing (Farre 2004). For this reason, Venla was equipped with an additional thermistor channel. In practice, a nasal cannula may sometimes irritate the sensitive skin around and inside the nostrils and the patient may remove it during the night. Thus, the additional thermistor may sometimes rescue the whole recording.
Venla and especially its successor, APV2, have been designed to be as simple and robust as possible to enable the operation of the devices with minimal or no training.
Moreover, since the recorded data are stored in a relatively small file (5 MB) in an encrypted format without any confidential patient information, the file may be safely emailed to the specialist sleep physician for analysis. These features, together with good technical durability and reliability, make the evaluated devices idelally suited for telemedical screening of OSA.
Both evaluated recording devices, Venla and APV2, are intended for recording adult people with suspected sleep apnea. They are not recommended for use with children or with some complicated special patients.
7.2 Automatic analysis of ambulatory recordings In Study III, the recordings made in Study I were re-analysed automatically with widely used commercial analysis software (Somnologica 3.2). The aim was to evaluate the agreement between manual and automatic analysis in Venla and Embletta recordings.
Significant differences in detection and classification of the severity of OSA were detected between the devices (Venla versus Embletta) when automatic analysis was used. However, manual analysis of the same data as in Study I detected no statistical difference between the devices. The results are in line with the current clinical guidelines for diagnosis of OSA (Collop 2007). The guidelines recommend a review of the raw data by a certified sleep specialist and manual scoring or manual editing of the automated scoring results by skilled personnel. The reliability of automated analysis has been criticised in several studies (Carrasco 1996, Cirignotta 2001, Fietze 2002, Yin 2005). However, the suitability of using analysis software with different recording devices has not been previously investigated. Many clinics use several different devices and analysis software. The possibility to use only one type of analysis software is tempting as it could bring significant economical and practical advantages. It might also allow the use of low-cost monitoring devices without the need to purchase expensive instrumentation-specific software.
In Study III, automatic analysis proposed a false negative diagnosis in a significant portion of patients having mild obstructive sleep apnea (65.4% with Venla and 11.4% with Embletta). In addition, the same trend of classification into a milder class was seen in patients with moderate and severe disease. The unreliability of diagnosing the mild disease with automatic analysis is a very important finding because mild OSA represents the most prevalent subgroup of OSA patients. This is especially important since the treatment of OSA can prevent its progression and the initiation of harmful cardiovascular consequences (Buchner 2007, Sahlman 2007).
When similar transducers were evaluated for measurement of airflow and oxygen saturation in Venla and Embletta only minor technical differences were determined between the biosignals recorded with these devices. However, in Study III significant differences were discovered in the reliability of automated analysis of these signals.
This indicates that even minor variations in the signal properties can alter the reliability of automated analysis. It is known that the technical quality of the ambulatory recordings varies significantly due to practical problems (Golpe 2002, Dingli 2003, Reichert 2003, Yin 2006). In Study III, the most common reasons for technical failures and for the exclusion of the recording from the study were loss of either airflow or oxygen saturation signals.
The main differences in the signal acquisition with Venla and Embletta are the sampling frequencies of the nasal pressure (8 versus 20 Hz), the oxygen saturation (1 versus 3 Hz), the thorax and abdomen movements (8 versus 10 Hz) and the body position (8 versus 10 Hz) signals. These differences may have contributed to the differences in the accuracy of the results of the automatic analysis.
Automatic and manual analyses were found to differ significantly in the sensitivity of detection of sleep apnea and in the classification of mixed and central apneas. This is an important finding as even though automatic detection of mixed and central apneas is commonly included in the analyses, its reliability has not been thoroughly investigated.
This is a clinically relevant issue, because large numbers of mixed and central events
are indicative of a more severe form of OSA, and misclassifications of these events may significantly delay the treatment of the patient.
The present findings are important because several ambulatory monitoring systems have been developed in order to optimize cost-effective screening of OSA. The automated analysis of the signals is widely applied since it is simple to perform, saves time and resources and is reliable in most cases. However, the present results show that the exclusion of OSA should never be based on automated analysis alone. The diagnosis of mild OSA should always be based on manual analysis of recordings, but if the automatic analysis points to moderate or severe obstructive disease, the suggestion may be accepted without further manual confirmation of the validity of the analysis. It is important that the classification of the type of the apnea further to obstructive, mixed or central type should always be done manually.
7.3 Applicability of the novel portable ERP device for brainfunction monitoring
In Study IV, a compact portable battery operated device for ERP, SEP, EEG and ECG measurements was designed, constructed and programmed for off-line monitoring of the level of consciousness or depth of sedation. The device is intended to be used in intensive care units and emergency rooms, which places special requirements for the robustness and ease of use of the instrumentation.
In the technical evaluation, the RMS noise amplitudes for the EEG and ECG channels were found to be at an acceptable level. Even though the RMS noise of the non-filtered signal (2.1 μV) exceeds the recommended maximum (0.5 μV) (Nuwer 1998), one can obtain high quality clinical EEG measurements with this device (Westeren-Punnonen 2005). The ERP measurements are based on averaging of hundreds of individual events which effectively reduces the RMS noise and improves the signal quality compared to spontaneous EEG. Band pass filtering reduces the RMS noise even further.
Modern equipment must not disturb or be disturbed by the other medical instruments. This is especially important with monitoring and life supporting equipment, which must fulfil the requirements of the International Special Committee for Radio Interference (CISPR 11 EN 55 011) (Williams 1996). For this reason, the Emma device was constructed to minimize spurious electromagnetic emission, even though this meant some compromises which slightly reduced its performance. In extensive EMC testing (Pohjois-Savo Polytechnic, Kuopio, Finland), the device was found to meet the strict emission limitations. The device was designed to be battery operated to help to minimize electromagnetic noise and to enhance electric safety. The estimated operating time of the device is 27 hours, which is well suited for clinically typical 24-hour monitoring periods in the ICU.
In addition to high quality of recorded EEG, the quality of the stimulus is very important in ERP recordings. The audio stimulus intensity was found to correspond accurately to the thumb wheel setting (Figure 6.3). In addition, the stimulus tones were found to be free of any imperfections such as glitches, hum or broadband audio noise.
This is important, because it is known that the audio stimulus intensity has a significant effect on the ERP responses (Adler 1989). In addition, the audio stimulus imperfections can distort the ERP responses and, thus, reduce the reliability of clinical measurements.
Kuopio University Publications C. Natural and Environmental Sciences 261: 1 - 79 (2009)
- 66 Discussion In addition to the technical tests, the device was successfully evaluated in the clinic.
The oddball paradigm with 85% standard (800 Hz) and 15% target (560 Hz) stimuli produced a clearly distinguishable ERP response (P300) to the target tones compared to the standard tones. In addition to ERP measurements, it was possible to undertake cortical SEP measurements producing distinct, artefact free responses (N19 and P22).
Additionally, the sharp stimulus artefact spikes just after the stimulus event indicate that the external event marking mechanism in the data acquisition section is precise.