Portal de Periódicos da CAPES

Tele-EEG in epilepsy: review and initial experience with software to enable EEG review over a telephone link

2003; Elsevier BV; Volume: 12; Issue: 2 Linguagem: Inglês

10.1016/s1059131102002297

1532-2688

David Holder, J.K. Cameron, C.D. Binnie,

Atomic and Subatomic Physics Research

The EEG investigation of epilepsy is a discipline in which the use of telemedicine of likely to be of particular value, on account of the following characteristics:1.In most countries, EEG facilities are widely dispersed.2.Patients may require investigation at locations remote from an EEG laboratory, for instance in the operating theatre or ITU, or in the home, school, or workplace to monitor seizures in a natural environment.3.Expertise in EEG interpretation is not widely available and particular knowledge of the EEG in epilepsy tends to be confined to special centres.4.Immediate interpretation may be required, for instance in status epilepticus, during electrocorticography and often to direct the conduct of intensive monitoring.5.Electroencephalography is a small discipline and its practitioners often work alone or in small groups; audit, clinical governance, research, and continuing medical education, therefore, require collaboration and joint review of records between centres.6.EEGs are available as digitised signals, which can easily be transmitted by telecommunications systems. Remote reporting of EEGs can be broadly divided into two functions—routine reporting of several or many records, and urgent reporting in individual cases. The routine reporting can usually brook a delay of several days, but urgent reports usually need to be made within hours. Routine EEGs for reporting can be transmitted on paper, microfilm, using recent PC-based media such as CD, or by the postal system or courier. Urgent records may be sent rapidly by fax1.Gibbs E.L Gibbs T The facsimile transmission of electroencephalograms.Clinical Electroencephalography. 1970; 1: 171-175Google Scholar, although it is usually only practicable to send a small proportion of the record. These methods are all practicable and are, to our knowledge, used by departments of Clinical Neurophysiology in the UK. However, the routine methods all require time, expense and delay for practical implementation, and faxing for urgent reports is limited in that remontaging is not possible, and it is time consuming if more than a few pages are transmitted. Both routine and urgent reporting could be easily accomplished, with a minimum cost in time or expense, by electronic transmission of the digitised signals over telephone or other data communication lines. This is not trivial as the volumes of data are considerable: typically 40–100 MB per hour or 1 GB per hour with video. Three approaches may be adopted:1.Off-line transmission of EEGs as complete files. This technically simple approach presents surprising difficulty unless dedicated wide bandwidth facilities are available. An attempt to download overnight half a dozen EEG files each of some 50 MB by the Internet over standard phone lines is likely to fail, as there is a high probability of an irrecoverable interruption of transmission during the period of several hours required for this procedure.2.On-line synchronous live EEG transmission is the preferred method for supervised EEG monitoring, but is beyond the capacity of the public telephone network, using the standard data formats of commercial EEG recorders without some form of data compression.3.Rapid off-line EEG transmission for live review. Electroencephalographers commonly scan EEGs visually at up to two 10 second pages per second, pausing from time to time to examine features of interest. To interpret a recently recorded EEG conveniently without tedious delays, the reviewer needs a display that can be updated at 20 times the original recording speed. The requirements of all three approaches will doubtless be fully satisfied in due course by the universal introduction of wideband networks for telephony, video and other applications. In the meantime, a system that can support all the applications envisaged earlier must function over the telephony links generally available, i.e. the public telephone network. At present, there are three principal transmission speeds available in the non-specialist commercial environment available to clinical departments—(1) analogue telephone line and a modem: 28.8–56 kbits per second, (2) integrated services digital network (ISDN): 64–128 kbits per second, (3) asynchronous digital subscriber line (ADSL) 512–3000 kbits per second. The amount available to a public consumer will depend on commercial decisions concerning the number of users who share a line. (These are rates in the UK; they may vary in other countries.) An ethernet or fibreoptic physical link usually transmits at 10 or 100 Mbits per second. For comparison, the information in an EEG may be calculated. Assuming 21 channels, sampled at 240 Hz, review at two pages of 10 seconds per screen each second, and 12 bits per sample results in a need to transmit about 1.2 million bits per second. It is clear that EEG may be easily reviewed at full speed over an ethernet or fibreoptic network, but full transmission of data will not be possible over a slow ADSL, or full speed ISDN or analogue telephone line (Fig. 1). Biotelemetry, including EEG, was employed in the first manned space flights of the 1960s and Ray et al.2.Ray C.D Bickford R.G Walter W.G Rémond A Experiences with telemetry of biomedical data by telephone cable and satellite, domestic and international.Medical and Biological Engineering. 1965; 3: 169-177Google Scholar describe its use over standard telephone and satellite links. Bennett and Gardener3.Bennett D.R Gardner R.M A model for the telephone transmission of bioelectric information.Journal of the American Medical Association. 1970; 29: 404-408Google Scholar transmitted six channel EEGs over two parallel voice lines from isolated small communities for remote interpretation, but noted that this greatly increased the cost of the clinical service. By 1974, Schear et al.4.Schear H.E Rowe W.J Pori J.R Telephonic transmission of electroencephalograms.Clinical Electroencephalography. 1974; 5: 24-30Google Scholar had established a system for transmitting eight channel EEGs with a bandwidth up to 65 Hz, over a single telephone line from 31 sites in the USA and Canada to a centre in San Francisco and had used this for remote review of some 6000 EEGs. The cost was four times that of a conventional in-house EEG service but this was offset by the avoidance of the need to transport patients to the nearest EEG laboratory. Only one application to epileptology was cited, the investigation of non-convulsive status epilepticus. The combination of local radio telemetry in the home, with a telephone link to a central laboratory was used in a four channel version by Hanley et al.5.Hanley J Zweilig J.R Kato R.T Adey W.R Rovner L.D Combined telephone and radiotelemetry of the EEG.Electroencephalography and Clinical Neurophysiology. 1969; 26: 323-324Abstract Full Text PDF PubMed Scopus (8) Google Scholar, and in an eight channel configuration by Rosekind et al.6.Rosekind M.R Coates T.J Thoresen C.E Telephone transmission of all-night polysomnographic data from subjects’ homes.The Journal of Nervous and Mental Disease. 1978; 166: 438-441Crossref PubMed Scopus (7) Google Scholar for polysomnography. All of these were analogue systems and the signals were subject to degradation by noise on the telephone lines. An eight channel digital (PCM) system with an effective 30 Hz bandwidth was developed specifically for home monitoring of epilepsy by Kamp7.Van der Weide H Kamp A Long-term supervised domicilliary EEG monitoring in epileptic patients employing radio telemetry and telephone telemetry. I. Telephone telemetry system.Electroencephalography and Clinical Neurophysiology. 1984; 57: 581-583Abstract Full Text PDF PubMed Scopus (4) Google Scholar, 8.Kamp A Long-term supervised domicilliary EEG monitoring in epileptic patients employing radio telemetry and telephone telemetry. II. Radio telemetry system.Electroencephalography and Clinical Neurophysiology. 1984; 57: 584-586Abstract Full Text PDF PubMed Scopus (4) Google Scholar. This application was one for which this technology was particularly suited. The local telemetry gave the patient freedom of movement without the encumbrance of cables. The display of the signals in the base laboratory allowed continuous supervision and hence quality control and feedback by telephone, to ask carers to rectify technical faults such as displaced electrodes, and to describe any clinical events accompanying salient EEG features. Somewhat surprisingly, there do not appear to have been any published accounts of EEG transmission over telephone lines more recently. Most major manufacturers now provide the ability for digital EEG transmission over fast ethernet or fibreoptic networks9.Loula P Rauhala E Erkinjuntti M Raty E Hirvonen K Hakkinen V Distributed clinical neurophysiology.Journal of Telemedicine and Telecare. 1997; 3: 89-95Crossref PubMed Scopus (19) Google Scholar, but we are not aware of any published results to transmit EEG over slower telephonic links. A proprietary system, ‘TeleTrend’10.Vespa P.M Nenov V Nuwer M.R Continuous EEG monitoring in the intensive care unit: early findings and clinical efficacy.Journal of Clinical Neurophysiology. 1999; 16: 1-13Crossref PubMed Scopus (145) Google Scholar (www.nicolet.com), intended for long-term ITU monitoring, compresses EEG and permits individual pages to be downloaded over the Internet using a Web browser, but only individual pages at a time may be seen. In contrast, there have been several publications on methods for compression of the EEG (e.g. see References 11.Antoniol G Tonella P EEG data compression techniques.IEEE Transactions on Bio-medical Engineering. 1997; 44: 105-114Crossref PubMed Scopus (138) Google Scholar, 12.Hinrichs H EEG data compression with source coding techniques.Journal of Biomedical Engineering. 1991; 13: 417-423Abstract Full Text PDF PubMed Scopus (8) Google Scholar), but the motivation for this appears to be to reduce storage requirements, and we are not aware of any work which has used these for telephonic transmission. Over the past 4 years, we have set out to develop software which would enable EEGs to be transmitted for routine review over a telephone link. The specification for this was that it could be used on IBM compatible personal computers (PCs) by inexpert users, at home or in remote clinics. Of the options mentioned earlier, it would provide interactive online review. The principle would be that a recorded digital EEG would be available on a remote server, running the BRIAN software. The user would also use BRIAN and connect transparently to the server. EEG records would then appear on the user screen, and the user could interactively change montage, gain, filtering and other parameters in exactly the same way as if the records were on the local machine. It could be linked to the server by any available method—analogue modem, ISDN, ADSL, Internet or local network, and the program would automatically make use of the available bandwidth. It would employ sophisticated compression techniques to permit transmission as close as possible to optimum review speeds. BRIAN is written in C++ and can run on any PC under Microsoft Windows 95, 98 or NT or under Linux. It was designed to be highly modular. The sections are constructed as a ‘pipe’. Data is pulled off the remote server, processed according to parameters set by the user, compressed, transmitted and then received by the user and displayed. The modular design is intended to permit a variety of different transmission options to be employed, and allow the incorporation of filters for different manufacturer formats and compression algorithms in a robust way. A description of the compression methods used is outside the scope of this article. The software is still under development. At the time of writing, the following speeds have been achieved (Fig. 1). These are expressed as the time taken to review a single page of EEG, which contains 10 seconds of recording, on an XGA screen (1024×768 pixels). The measurements were made using a server PC with a Pentium processor running at 500 MHz and a review PC with a Pentium at 133 MHz. The timings were compared with a commercial program, PC Anywhere v. 8.0 (Symantec software; www.symantec.com), which allows a PC to view the screen of a remote host PC. For an assessment of the efficiency of BRIAN, comparison was also made with the timings that were calculated from a knowledge of the degree of compression and taking into account data processing and display. Two cases were considered: if the EEG data were transmitted with 100% efficiency and maximum compression, or all data was transmitted without compression at all (Table 1). It may be seen that there is a fixed additional network overhead of about 0.5 seconds, so that the speed increase varies from about 10 times with a 28.8 kbaud modem to four times for dual line ISDN.Table 1Timing of review per page of 10 seconds of EEG for BRIAN with 28.8 or 128 kbaud modems (times in seconds).ModemCalculated time per page (no compression)Calculated time per page (with compression)Measured time per page28.8 kbaud211.72.2ISDN dual (128 kbaud)4.90.61.2 Open table in a new tab The figures given were collected with maximum compression for all the modems employed. This used in part an algorithm which was lossy, which enabled an additional compression by a factor of about three times. This is a parameter which can be set manually by the user. Clearly, there is a theoretical concern that clinically significant features in the EEG may be lost. This has not yet been formally tested, but the clinical judgement of the two electroencephalographer authors is that the loss of detail is negligible. Some examples are given in Fig. 2, for frontotemporal sharp waves, or the more difficult problem of notched delta waves. The reduction of reviewing speed produced by avoiding lossy compression is shown in Table 2.Fig. 2Comparison of EEG tracings made with or without lossy compression: (a) and (b) are of frontotemporal sharp activity without (a) or with (b) lossy compression; (c) and (d) are of notched posterior slow activity without (c) or with (d) lossy compression. Although there are slight differences with the different types of compression, the notable features are still evident.View Large Image Figure ViewerDownload (PPT)Table 2Speed per page of reviewed EEG (in seconds) with lossless or lossy compression with three different types of modem.Compression28.8 kbaudISDN 64 kbaudISDN 128 kbaudLossless5.22.41.6Lossy2.21.71.2 Open table in a new tab In practice, use of BRIAN was similar to review at a stand-alone workstation. At first, menu options are used to dial the remote computer; the desired file is selected from a drop down list which appears after 10 seconds or so. In clinical use, the referral was made by email, so the remote reporting clinician was advised of the patient details and file name. In review, all usual parameters, such as filter settings, montage, gain and timing per page, could be set interactively in the usual manner. At the time of writing, BRIAN has just been completed. We are about to undertake clinical trials. Preliminary experience with it is encouraging. The practice of the electroencephalographer authors is to review EEGs normally at about two pages per second, pausing to inspect any page with important features. BRIAN can achieve this only over a fast ethernet network at present. However, the speeds achieved with ISDN or a modem are acceptable, and the quality of the reviewed data appears to be high. In the opinion of the authors, even review at 2 seconds per page with a 28.8 kbaud modem is vastly preferable to having to drive into work to review an urgent record! The main difficulty has been the robustness of the software with different computers and modem connections. Programming communication code under Windows is technically demanding, and much development time has been spent on debugging idiosyncrasies in Windows which cause the program to crash. Field trials with the software will commence shortly. It will be used for reporting of records between hospitals, and of urgent records by clinicians at home. Reports will be compared with those made at the collection site, using review equipment provided by the manufacturer. The robustness of systems in practical use will be monitored and the effect of different types of compression on reporting conclusions. BRIAN has not yet been tested with ADSL, but it is very probable that maximum review speeds will be achieved without difficulty with this system. As yet, ADSL is only available in large cities in the UK, and is expensive—about £500 per year. As full bandwidth, ADSL is roughly equivalent to a local network, it is probable that commercial EEG reporting software, which works over a local network, will produce similar results. If this is the case, then BRIAN will offer no advantage over proprietary software. However, ADSL lines will be shared by multiple users in many commercial arrangements, so it is possible that the full bandwidth of several megabits per second may not be available. BRIAN may then still be useful. It is likely that ADSL, or similar high bandwidth connections, will become universally available over the next few years. However, it will require special connections and it is likely that it will be significantly more expensive than a normal analogue telephone connection. BRIAN could, of course, still be used then, but will offer no advantage over proprietary software. We anticipate that it could be useful, however, in situations where the expense of an ADSL connection is not justified—this could be to outlying hospital departments, or to the homes of reporting clinicians. In particular, it could be useful for reviewing of records over the Internet by users at home. Shortly, the British Society for Clinical Neurophysiology (http://www.bscn.org.uk) will commence an Internet discussion list, open to any professional involved in EEG reporting; users will be able to download BRIAN and review full EEG records posted to the group for discussion, using an ordinary analogue telephone connection. Ideally, it would be desirable to review records at full speed with a 28.8 kbaud modem. Inspection of Table 1 shows that this could in theory be possible if compression by a further factor of three could be achieved. At present, the code is limited by an overhead which appears to be about 0.5 seconds per screen. This is probably due to handshaking on the communications code and other fixed procedures in the software. Use of more lossy compression may be clinically acceptable, but it seems unlikely that substantial further savings in time will be possible. The present code has taken almost four man years to write, and we propose to use it with the present performance in clinical trials. Remote EEG review appears to be a desirable activity that many electroencephalographers would welcome. It is likely that it will become widely possible with the advent of fast communication links such as ADSL over the next few years. We present, for the first time, the BRIAN software which permits EEG review at almost normal reviewing speeds over normal analogue or ISDN telephone lines, by using sophisticated compression techniques. Initial experience is encouraging, but full clinical field trials are needed before it can be recommended for routine clinical use.