Technology drives the answer as the equipment and techniques used for sleep recordings become more sophisticated

For the last several decades, the polygraphs routinely used in sleep laboratories were much like electroencephalographs (EEGs) but provided several additional channels with analog-based amplifiers. These instruments, still considered today to be the “gold standard” in sleep recording equipment, produced paper recordings that had to be scored and interpreted in the configuration in which they were recorded. Recently, computer-based polygraphs have been increasingly used by sleep laboratories—either as a replacement for the analog-based equipment or as an addition to existing analog equipment. Until recently, recording of sleep data was severely limited by the recording equipment (analog amplifiers and paper tracings). With the development of digital recording equipment, the limitations experienced in the traditional setting have all but disappeared. As digital acquisition, processing, and recording become the new standard for evaluating sleep/wake data, clinicians must adapt the skills they have acquired in the analog world to the technology-driven world of sleep diagnostics.

Digital Technology
Data recording refers specifically to data acquisition, data processing (filtering), display, and storage. In the field of sleep diagnostics, a wide range of methods for data recording exists. Data recording ranges from the traditional paper-based, attended study in a hospital to a digital-based, automated, and unattended study in the home setting.

In digital-based signal acquisition, the analog signal generated by the transducer in each of the channels is digitized immediately after it is picked up from the electrode or sensor. Critical to this process is the sampling rate of the signals. If the sampling rate is insufficient, waveforms become distorted. Accordingly, scoring often becomes difficult and may produce errors with interpretation of the data. Digital processing of the data occurs following data acquisition and primarily involves filtering unwanted signals using digital-based filters in order to reduce interference and distortions, thereby optimizing signal quality. Unlike during analog acquisition, digital sleep/wake data is acquired in its “raw” form and stored immediately. Digital filtering thus becomes a process that can be subsequently manipulated in a variety of ways, both online as well as once the recording has ended. In contrast, sleep/wake data that is recorded using paper-based analog equipment can be manipulated only in real time. Once it is recorded with specific filter settings, the sleep professional cannot go back in time to change the filter settings, and thus cannot manipulate the data that has been previously recorded.

Data storage has traditionally been a problem for most sleep laboratories. Saving records for any length of time produces monumental storage requirements. Each standard polysomnogram is approximately 800 screens or, in the case of paper studies, 800 sheets. By the end of the year, a modest-sized sleep laboratory with four beds could produce a ton (metric) of paper studies. Digital storage refers to the data being stored as data files on either the computer hard drive, removable disks, such as compact disks (CDs), or optical disk. With the advent of computers and the various methods of data storage, that annual volume of information illustrated previously could be as small as an average music collection at home of approximately 120 CDs. Digital-based polysomnography consequently offers the ability to archive large amounts of sleep/wake data onto hard drives, CD drives, or optical disks.1

Paper-based sleep recordings have achieved their gold standard status primarily due to the high-quality resolution of the recorded data on the paper tracings. In digital-based polysomnography, the digitized bioelectrical signals, traditionally displayed as pen tracing on paper, is immediately stored magnetically or optically onto the recording media. The scoring and reviewing of these tracings on a computer screen require a display resolution sufficient enough to clearly visualize waveforms. Display resolution is dependent on the characteristics of the computer, display monitor, and the software used for acquisition and display of the sleep/wake data. Low resolutions on computer displays have been commonly associated with problems in assessing EEG data. There is also increasing difficulty with recognizing higher frequency waveforms, such as spindles and seizures. Display resolution has historically been the primary disadvantage of digital-based sleep systems over analog-based sleep systems. In recent years, however, technological advances in display resolution, coupled with lower costs associated with larger computer monitors, have diminished the disadvantage.

Digital Sleep Software
Technology drives not just the digital amplifiers, filters, and signal processors; it also drives the software used to analyze the data that has been recorded. Current sleep software systems are equipped with intuitive user interfaces and flexible configurations that are able to handle a broad spectrum of analysis protocols and acquisition techniques. These systems allow customization of fundamental aspects of the sleep recording to meet the laboratory’s individual requirements. Customized menus control the active features of the program and the layout of the displays to highlight data that the technologist and physician consider critical to analysis. The sleep software programs streamline the steps in the sleep study process using a Windows-based user interface. Among key features in a sleep software program is the ability to provide assistance with downloading data, analyzing recordings, and creating reports. With this interface, data displays can be split horizontally or vertically to view different time periods or make other data comparisons. Many sleep software programs utilize recording and device managers that keep the data and connected devices organized and readily available to the operator. The majority of features and functions necessary to operate most sleep software are commonly contained in a single display and accessed using the mouse.

The latest generation of software programs allows the sleep traces to be rereferenced to provide more information from each study. This particular feature allows practitioners to alter events according to their expert opinions and adapt the software’s parameters and functions to meet their needs. Many of the sleep program settings, such as the time and amplitude axes, can be completely customized prior to a study or during subsequent analyses of the same data. Sleep-related events, such as apneas, arousals, limb movements, and electrocardiograph (ECG) abnormalities, can be marked on the display screen in real time or during subsequent analysis. Sleep software programs provide the user with the ability to manipulate each of the tracings by dragging and clicking, rescaling, and editing in various ways. The latest generation of sleep programs also allows for the creation of recording and analysis using multiple templates that can be saved to view sleep/wake data according to individual users’ preferences. This also provides a platform to allow scoring comparisons between operators. In addition, most sleep software programs commonly provide the user with the ability to create their own customized reports and to export those reports in different file formats, such as MS-Word and PDF.2

Where Digital Is Going and Analog is Not
Where is the field of polysomnographic technology headed? Wherever the destination, one can, with certainty, state that it will continue to be technology-driven. Some envision a sleep laboratory that uses the Internet to monitor patients sleeping at home. In this context, the number of patients tested on any given night will very likely be limited primarily by the amount of equipment available, and not the bed space in a sleep laboratory or even the number of monitoring technicians.3

The latest technological improvements in digital polysomnography involve advances in automation and telemonitoring. Recent technological advancements have allowed the ability for sleep software programs to provide automatic analysis of sleep/wake data, through the application of computerized algorithms for sleep, respiratory events, movement events, and arousals. Much of the recent research consists of the evaluation of computer scoring against a standard with assessment of reliability, validity, or both.1

The basic requirements for the development of automatic scoring programs for any bioelectric signal are the presence of well-defined detection criteria. A study4 analyzed more than 23,000 epochs for four mice that yielded an overall agreement of 94% between two human scorers and the program. The scoring algorithm matched the human consensus best for wakefulness and non-rapid eye movement (NREM) sleep, but was lower for REM sleep and theta-dominated wakefulness. Perhaps this is due to the similar waveform characteristics and staging criteria of REM and wake. Investigative studies have not always resulted in positive outcomes. For example, a study5 suggested that a computerized sleep system, when allowed to score automatically without editing from the technologist, resulted in substantial errors that could lead to interpretative problems. The autoscore system often scored sleep stages incorrectly, failed to recognize many apneas/hypopneas and arousals, and overscored periodic limb movements (PLMs). However, the authors suggested that individualizing the default parameter settings would have improved the autoscore results. The epoch-by-epoch analysis yielded some interesting results, the most prominent of which was the tendency to score more wake time and lighter sleep on the edited computer analysis than on paper with only a 75.7% overall agreement. Finally, this study suggested that scoring can be more quickly accomplished by editing a computer record than by scoring completely on paper. The technician in the study required an average of 172.6±9.9 minutes to score a paper record and only 79.7±4.8 minutes to edit the previously autoscored computer record. The investigators stated that the reduced scoring time on computer should not be interpreted as necessarily indicating that a computerized sleep system reduces the total cost of polysomnography.

Another study by Doman et al6 concluded that the digital processing of sleep signals saves computer operator, polysomnographic technologist, and computer time. It also saves resources such as polysomnographic paper and FM tape. The digital signals lend themselves to a large array of analysis techniques and result in improved signal quality. Automated REM and delta-wave detection via digital processing correlates highly with visual counts of REM and delta waves. A study by Kayed et al7 concluded that automatic detection and analysis of periodic movements in sleep can provide an objective method for the study of several aspects of this disease that are still not yet fully understood. The events missed by the computer and scored visually were those with durations that did not exactly correspond to the specified duration of 0.5 to 5 sec.

Lord et al8 looked at 26 records of sleep and breathing obtained with a portable monitoring system from elderly subjects that were scored by three raters with computer assistance to examine inter-rater reliability of scoring. Raters were a medical student, a nurse practitioner, and a family physician, all of whom had at least 1 month’s experience with the equipment. Significant agreement was observed for all variables, although agreement was better for variables describing breathing than for those describing sleep. The investigators concluded that inter-rater reliability of identifying and characterizing breathing disturbance during sleep as recorded by portable monitoring is high among trained raters using computer assistance.

A number of sleep laboratories are moving into the future utilizing telemonitoring as a means of acquiring sleep/wake data. Recent technological innovations have allowed sleep testing to expand beyond the laboratory’s walls. Studies are now routinely performed in hospital rooms, the intensive care unit, or remote locations. But because it is not always possible for a trained technologist to be present, these alternate site recordings often present significant challenges. However, both recent and future advances in technology have successfully paved the way for telemonitoring in sleep diagnostics to eventually become a viable alternative to laboratory-based testing. Some sleep experts argue that patients sleep at home in greater comfort, and an improved test can be provided if their sleep/wake patterns are monitored in their own homes via the Internet. As the use of cable modems and fiber optics becomes the mainstay, use of the relatively low data-transfer rates of the modem (the most common form of communication between patient homes and laboratories) will help diminish the limiting factors associated with telemonitoring.

Conclusion
As the field of sleep diagnostics has grown, the techniques and equipment used for sleep recordings have become more sophisticated. For more than three decades, sleep equipment has been based on analog technology and sleep/wake data has been recorded on paper. Analog-based equipment is still utilized today, primarily for research and teaching purposes. Analog-based equipment also continues to be used as a backup mechanism for digital recording equipment. Even with the continued use of analog, paper-based systems, the downside of analog is significant. Digital amplifiers and signal processors have provided many advantages not achieved with analog-based equipment. For example, digitized data can be manipulated to display a variety of montages, depending on the task to be accomplished. In addition to digital amplifiers and processors, several other technical advances have improved the diagnostic value of a polysomnogram. For instance, video recording can provide simultaneous analysis of physical behavior with polysomnographic findings. As technology advances the science of polysomnography, technologists and physicians alike must no longer debate which is better, analog vs digital, or which type of system will prevail. That debate is over. Digital-based sleep systems are here to stay. The burden is now on the technologist and the physician to make sure he or she has the education, skills, and competency required to function in a technology-driven profession.

Tom Smalling, MS, RRT, RPFT, RPSGT, is clinical assistant professor of health sciences at the School of Health Technology and Management at SUNY Stony Brook, NY, and is a member of Sleep Review’s Editorial Advisory Board; Russell Rozensky, RRT, CPFT, RPSGT, is a clinical instructor of health sciences at the School of Health Technology and Management at SUNY Stony Brook. He is also the technical supervisor of the Sleep Disorders Laboratory at John T. Mather Memorial Hospital, Port Jefferson, NY.

References
1. Hirshkowitz M, Moore CA. Issues in computerized polysomnography. Sleep. 1994;17:105-112.
2. Geyer JD, Carney PR, Payne TA, Aldrich MS. Atlas of Digital Polysomnography. Philadelphia: Lippincott Williams & Wilkins; 2000.
3. Proctor SM. Analog vs digital polysomnography: technology being overshadowed by need for telemonitoring. ADVANCE for Respiratory Care Practitioners. 2000;13:20-21.
4. Van Gelder RN, Edgar DM, Dement WC. Real-time automated sleep scoring: validation of a microcomputer-based system for mice. Sleep. 1991;14:48-55.
5. White DP, Gibb TJ. Evaluation of a computerized polysomnographic system. Sleep. 1998;21:188-196.
6. Doman J, Detka C, Hoffman T, et al. Automating the sleep laboratory: implementation and validation of digital recording and analysis. Int J Biomed Comput. 1995;38:277-290.
7. Kayed K, Roberts S, Davies WL. Computer detection and analysis of periodic movements in sleep. Sleep. 1990;13:253-261.
8. Lord S, Sawyer B, Pond D, et al. Interrater reliability of computer-assisted scoring of breathing during sleep. Sleep. 1989;12:550-558.