IEEE ENGINEERING IN MEDICINE AND BIOLOGY. 0739-51 75/99/$10.0001999 ...... within the mathematics and physics facul- ties. His research interests are in ...
intelligent TeIem Critical-Care Patie A System for Providing Detailed, Red- Time Knowledge Available Any Time From Any location
S. Barro’, J. Presedo D. Castro, M. Fernandez-Delgado, S. Fraga, M. Lama, J. Vila Deportment of Electronics and Computer Science, University of Sontiago de Compostela, Spain
n recent years, we have witnessed an increasing process in the instrumentalization of different hospital units, especially those orientated toward the intensive care of patients. The increasing availability and sophistication of sensors for obtaining the most varied parameters and signals from a patient and the increase in computing capacity, in storage, and in representation of information from microprocessor-based instrumentation is a two-sided coin for medical staff and patients in their care. The availabilityof more and more reliable information on the physical state and evolution of the patient is always desirable, but this information needs to be assimilated and evaluated by the staff and then adequately integrated into the decision-making processes, otherwise its value is negligible. In this respect, the role of intelligent systems with regard to supporting medical staff in these tasks is becoming ever-more important [ l]. Another complementary evolutionary path that monitoring systems have followed involves the improvement in the storage capacity of information and its recovery and presentation to the user. In this sense, the evolution and improvement of the user interface in monitoring systems has been constant [2,3,4], and at the same time inter-equipment connectivity has increased. This last aspect is highly valued, and at present all bedside monitor manufacturers offer the possibility of connecting their equipment to a local area network. This allows the storage of generated information into a central server, according to a client-server architecture, in such a manner that visualization of any patient monitoring information can be carried out from any system connected to the network. This possibility of patient monitoring from a remote location is reaching levels that only a few years ago were unthought IEEE ENGINEERING IN MEDICINE AND BIOLOGY
of, going hand in hand with the information and communications technology. At the present time, a major aim is that the user of monitoring data should have integral access to the information at any moment, from any place, and in the best form possible. The spread of the internet and the appearance of web technology are factors that are helping to achieve this objective. Sutil+ is an intelligent monitoring system that is under development within the Intelligent Systems Group of the University of Santiago de Compostela. The aim of this system is to attempt to advance the two lines mentioned above. That is, to endow the system with more intelligent behavior and to make the presentation of information to the user as flexible as possible, in space, time, and form. In this article, we present two telemonitoring solutions that enable access to information resulting from the monitoring of patients in coronary care units (CCUs), independent of the location of the system user. We concentrate on aspects of data acquisition and storage and, above all, interaction with the user. We will not consider intelligent decision making, which has been already dealt with in other publications [ 5 ] .
Distributed Monitoring Environment The different components that make up the monitoring environment that we
are designing include the following and are shown in Fig. 1. Commercial Monitoring Systems %til+ is an experimental monitoring system. As such, it needs to co-exist with commercialbedside monitors that function in the CCU. More specifically, the different physiological variables processed by Sutil+ are obtained by means of a local network dialogue between our system and the different bedside monitors in use. In this 0739-51 75/99/$10.0001999 E IEE
JUIy/AUgUSt
1999
CCU Local Area Network
6 HosDital Local
Remote User Interface 1
Remote User Interface M
1. Monitoring-environment components.
way we avoid, on one hand, having to directly acquire the physiological variables from the patient, and at the same time, we ensure that our system does not alter the normal procedures and operations of the CCU. Evidently,for Sutil+ to be capable of capturing the data that flow through the local area network of the CCU, it is necessary to know their format and to develop the driver charged with their acquisition. At this time, our system is capable of dialogue with bedside monitors manufactured by Marquette [6], and we hope to develop future drivers for dialogue with monitors from other manufacturers. Intelligent Monitoring Systems (Sutil+) The function of intelligent monitoring systems is high-level monitoring and surveillance of patient physical state over time. Within the system are diverse software modules under a modular and flexible architecture. This system is based on a general-purpose computer capable of supporting the Unix System V Release 4 (SVR4) operating system. Specifically, July/August 1999
its functionality has been tested on IBM PC-compatible computers as well as on Sun Workstations, both running the Solaris operating system, which is SVR4 compatible. The minimum requirements for a system with these characteristics are a Pentium I1 processor at 233MHz, 64 MBytes RAM, and a 4-Gbyte hard disk. A CD-ROM and a tape are also recommended. Central Web Server This system has the task of centralizing all monitoring information generated in the CCU into one single location for storage and presentation, serving as a union nexus with the different user interfaces and monitoring systems active at each moment. In order to facilitate access to information, a web server is installed on this system that acts as an information/application server. The operating system of this computer, as for the different monitoring systems, is SVR4, and the minimum requirements will depend on the number of possible clients that will be simultaneously connected to it. IEEE ENGINEERING IN MEDICINE AND BIOLOGY
Integral Access Interface (IAI) Any normal computer with the capacity to access the internet and with a Java-compatible browser has access to the IAI. By means of this module, the physician will be able to remotely access all data associated with the monitoring of a patient, using an application developed in the Java programming language executed by the Web browser after connection to the central web server. In this case, the minimum requirements for a computer to act as an IAI are significantly lower than for a monitoring system computer or central web server. Both desktop computers and laptops are used as well as different hardware architectures (PCs, Macintosh, workstations, etc.), thanks to the independence of specific hardware requirements achieved by using Java. With respect to the communication channel for connection to the central web server (modem, ISDN, etc.), a 28.4K baud bandwidth or higher is recommended, while graphical resolution should be 800 x 600 pixels or higher. Personal Alert Interface (PAT) The PA1 is a palm-top computer-based user interface that alerts the user at all times of an alarm in the CCU. Due to its reduced dimensions and its limitations with regard to calculation and memory, this interface has important restrictions with regard to its capacity for the presentation of information. In any case, the most noteworthy point is its role, which is complementary to the IAI, as will be discussed in greater detail below.
Signal Acquisition and Processing of Signals and Parameters Sutil+ is based on the signals and physiological parameters suppliedby the bedside monitors in CCUs. Although it is not the aim of this article, it is worthwhile to briefly point out some aspects of its software architecture (see [5] for a fairly detailed description of a recent version), made up basically of the following elements. Information Access Driver This software module interacts with the different bedside monitors and with recovering information given by them. In the case of wanting to change the monitor or system supplying the signals andphysical parameters to Sutil+, one only need carry out modifications in this module, leaving all the others unaltered. 81
Arrhythmia Database format [7], which is widely accepted for the storage of one-dimensional biomedical signals. The acquisition driver also allows files in this format to be read, thus facilitating access to database registers or off-line access to registers resulting from the monitoring of patients by means of Sutil+.
These bedside monitors may be configured in different manners or may change their configuration over time, as the consequence of new needs in the monitoring of a patient. The driver and the rest of the algorithms in our monitoring system are capable of detecting changes in configuration and adapting themselves to them in the appropriate manner. The only requirement is that, at all times, there should be an active electrocardiograph (ECG) derivation, otherwise it will not be possible to identify each cardiac cycle, which would hinder a great majority of the monitoring algorithms from functioning properly. Our system stores information in a format compatible with the MIT-BIH
Principal Module Blocks Signal and parameter processing by Sutil+ is carried out in increasing abstraction levels, distributed in two principal modules: the low level monitoring module and the perception specialist (Fig. 2). We now briefly describe the principal module blocks: Sample by sample processing level. In this processing level are found all those tasks that should be executed with each sample of physiological signals acquired. Among these tasks are those charged with filtering the different signals, with the detection of noise, and the detection of heart beats from the ECG signal. Beat-by-beat level. In this level are found all those tasks that must be executed as a consequence of the detection of a new beat i n the sample-by-sample processing level. Here, on the ECG signal, we go on to the delineation of a new beat, its morphological categorization, and the determination of its origin of activation (normal, supraventricular, or ventricular), as well as the extraction of all those characteristics thought to be opportune (height and width of the different waves of the beat, deviation of the ST segment,
heart rate, etc.). In this processing level, parameters of other types of signals are also determined, such as systolic, diastolic, and mean pressure in each of the invasive pressure derivations. Basal-to-basal processing. This is the processing level with the task of constructing the basal beat, the result of the average throughout constant time intervals (usually 10 sec) of all those normal beats included in a condition of normal rhythm. Its basic function is to show the temporal evolution of normal morphology, thus making it possible to visually analyze the evolution of determined abnormalities, such as the appearance of ischemic episodes. Perception. The perception specialist [8] is charged with carrying out high-level tasks within Sutil+, generating data of a qualitative nature and specific a l m s . It is implemented by means of a blackboard architecture [9], which facilitates modularity and flexibility in carrying out functions. It operates on the data that it continually receives from the low-level monitoring subsystem, via the communications module (CM), and it applies a series of processing agents, the results of which are redistributed in an increasing abstraction-level architecture. The different agents are the linguistic filtering agent [lo], which operates on the patient’s numeric variables in order to give an interpretation of the signal in linguistic terms; the episode detector agent, which detects the different types of significant
I
I
Basal Level
I Beat Level I I 1Sample I
I I
Building of Basal Beat Feature Extraction MorphologicalCharacterization Determinationof Origin of Activation Filtering Noise Detection Beat Detection
I
I
Information Access Driver
------__
Blackboard +Episode
_ _ _ - _ _ - - I
I
I+
-1
I/
-1
I1
Detection
Linguistic Filter
I
I
____---_--------Perception Specialist
I
Bed Monitor
2. Software architecture of Sutil+. a2
IEEE ENGINEERING IN MEDICINE AND BIOLOGY
July/August 1999
%til+
A i s t of Patients Patient Selection
Information
[ 7
Visualization of Data Received
t -
User intervention
Change in Interface
L State Real Time => Off-Line Off-Line => Off-Line Off-Line => Real Time
Sending of Request
Sending of Reply to Request
Bridge
J- Reply to Request
Request
3. State diagram corresponding to the execution of the IAI, from the activation of
the interface until functioning under normal conditions. episodeson the signals (e.g., ischemia episodes [ 111); and the temporal profile detection agent, which carries out the identificationof curves in the evolution of the signal that are clinically relevant to the physician.
User Access to Information As we have pointed out, the CM acts as a mechanism of interaction between the different perception elements and the low-level monitoring subsystem. Moreover, it has the task of storingmonitoringinformation and managing access to it from the different user interfaces that are active. The communication mechanism chosen is based on the TCP/IP protocol, and more specifically on the sending of UDP packets by means of the use of sockets [13] according to a client-server architecture. Specifically, the CM acts as an information server, waiting for requests from the different software modules through a single UDP port. Nevertheless, user interfaces do not communicate directly with the CM, because, as shall be seen, all dialogue must necessarily pass through the central web server. More specifically, the different user (client) interfaces send packets requesting information to the central web server, where there is a process under way that acts as a bridge, only allowing authorized packets to pass into the local CCU network, and directing them to the appropriate moniJuly/August 1999
toring system according to the information present. As a consequence of the reception of a request by a monitoring system (Sutil+ module), this is interpreted by its CM and resolved in an appropriate m a n n e r . O n c e all the information required is obtained, the monitoring system remits a packet with this information to the central web server, which in turn sends it to the interface that had originally requested the information. There are various reasons for introducing the central web server. On the one hand, in this way it is possible to focus all the external consultations of data coming from the CCU into one point. This offers
more security against outside interference, as there is only one system that could be vulnerable (the central web server), and not all the elements of the network. On the other hand, the different user interfaces do not need to know which monitoring systems are activated at the time of establishing communication, and what their IP addresses are. Instead, it is left to the central web server to indicate which monitoring systems the different interfaces can connect to, suitably directing the packets received to each of the active monitoring systems. At present, the central web server acts as a mere information bridge and as a notebook that allows the consultation of very simple information, such as, for example, knowing which monitoring systems are active. Nevertheless, we are currently working on the incorporation of a database that will permit the suitable management and storing of information generated in the CCU in such a manner that consultations may be made to this database via a web browser. As we have already mentioned, our system currently permits access to the monitoring information of Marquette instruments. The dialogue between this equipment and %til+ is also realized via UDP, with the communication protocol being imposed by the manufacturer in this case. It should be pointed out that information loss could come about, in which case the data access driver will note this situation, which will be reflected both at visualization and storage levels. We now concentrate on those aspects that are more typical of user interaction, currently designed around the two paths of interaction, with objectives that are practically complementary: IAI and PAI.
................................................. ........
.-.-.-..
...................
1
,;
I I , - _ . -
4. Identification phase in the IAI. IEEE ENGINEERING IN MEDICINEAND BIOLOGY
83
by means of the IAI, which basically consists of the following steps (see Fig. 3): Identification to the system phase. In this initial phase, the user connects to the central web server via a Java supported browser. This results in the execution of an applet (a Java application executed in a browser) and the user is invited to introduce hisher identification data into the system (Fig. 4). Once the information has been introduced, it is sent to the web server for validation. Patient selection phase. Once the user has entered the system, helshe is shown a list of the patients admitted to the CCU who at that moment are being monitored (Fig. 5). In this phase, the user has to select the patient. From this moment on, a logical link is established in the central web server between the user interface and the monitoring system. Monitoring phase. Once the user has been identified and has selected the patient which helshe wishes to observe, a series of windows appears, in Integral Access Interface accordance with a profile established The integral access interface has been by the user, that presents information developed using web technology. Given relative to the patient. It is during this the dynamic character of the information phase that the user has access to all to be presented (real-time visualization), the monitoring information, both we have chosen Java for the development real-time as well as that generated of this task [14, 1.51. Our choice is based throughout the entire patient monion the fact that this type of technology is toring process. simple to introduce and to use (via a Finalization Phase. This phase is Java-supported web browser), since the very similar to the user identification only software necessary in order to have phase in that heishe must once again access to information is the browser, and, identify himselfherself in order to in the case of modifying the application, it request the termination of the visualis only necessary to modify information ization process. residing in the central web server. Let us now go back to the monitoring We now describe the interactionprocess between the user and our monitoring system phase in order to give a more detailed explanation. Once the user has selected the patient that helshe wishes to monitor, the user profile in the central web server is ........-. ............._........................._. ._ "'.':i i i consulted. This specifies the information ;......................... -.,si L................................................ i z t z Cr~cio!;es to be presented initially. For example, Fig. SaIc.cchns dispcSitiv.5 : . ......................................... .....................$ 6 shows a possible configuration of the IAI during the monitoring of a patient. In iR osA ._.RO xtz I.R A M 0N_.E 0 x5 the upper central section, four physiological signals are shown in real time, which at all times are selectable by the user from the set of signals being monitored. As the signals move from left to right, annotations derived from the processing of %til+ are superimposed (detection and classification of beats and other significant events, such as ischemic episodes, etc.). Other windows shown in Fig. 6 show the temporal evolution of parame5. Patient selection phase in the IAI. ~
84
IEEE ENGINEERING IN MEDICINE AND BIOLOGY
ters obtained from the monitored signals (left) and the alarms occurring since the onset of monitoring (upper right), of which it is possible to consult their description and certain related information. With regards to the way the interface operates, we should discriminate between requests relative to the real-time presentation of information and requirements relative to the presentation of past or off-line information. A real-time request is produced each time the scroll bar associated with the time axis of one of the windows is scrolled to the extreme right. This request is sent to the central web server, such that every time that the central web server receives information in real-time from one of the monitoring systems, it consults whether there is any active interface subscribed to the system. If so, it determines what real-time information is required, with the aim of sending it out. In this case, the central web server assumes the task of maintaining the updated real-time information required in the interface until it is disactivated, or until it enters into a state of off-line information visualization. An off-line information request will be produced whenever the user moves the scroll bar associatedwith the time axis of a window to any position other than one at its extreme right. Consequently, a request is emitted that is received by the central web server, which in turn transmits it to the corresponding monitoring system. The monitoring system will cany out the search for the information required, which could either be in the memory or on disk (stored in MIT format [7]), and will respond to the request, passing the information to the central web server, which will in turn send it to the corresponding interface. As can be seen, within the IAI, those windows that show temporal information have some special characteristics that differentiate them from the others. One of the peculiarities of these windows is that they possess a synchronization button. Its function is as follows: If at a given moment we observe a significant event in one of the windows, by pressing the button the rest of the windows showing temporal information are synchronized at the temporal instant corresponding to the event. Thus, the pressing of this button is the equivalent to the sending of multiple off-line requests by each of the windows that need to be synchronized. Many of the IAI windows have menus for the selection of the set of signals (or parameters) to be shown (see the window July/August 1999
that shows the temporal evolution of parameters in Fig. 6). We consider it important to note that these menus are generated in a dynamic manner, as at each moment a bedside monitor may be reconfigured. This change is detected by the information access driver of the monitoring system, which informs the central web server, and this, in turn, informs the different user interfaces so that they modify the state of their menus. Lastly, there are other research groups and manufacturers working on solutions in the same line as we present here. At a commercial level, there is at least one monitoring-equipment manufacturer with a solution that i s similar to ours. More specifically, we refer to the Ultraview Web Source product from SpaceLabs (http://www.spacelabs.com). This product is also based on Java technology and also allows both real-time and off-line access to information. Nevertheless, our proposal is intended to have an advantage over this solution. On the one hand, our solution is independent of the bedside monitor used, being adaptable to new equipment by simply modifying the information access driver in the appropriate manner. On the other hand, throughout the development of our monitoring system, we have paid special attention to attaining levels of competence in patient monitoring that could be classified as “intelligent,” something that commercial equipment generally lacks. Other research groups working in experimental monitoring systems have developed web-based interfaces for information access, such as the case of Simon-Web (http://www .vuse. vanderbilt .edu/-simod) or the IMI project (http://gray.lcs.mit .edu/imidemo/) , even if the information access available is generally either not real time (Simon-Web), or, in the case that it is, aimed at demonstrating the possibilities afforded by the web (IMI project). Personal Alert Interface By allowing remote access, the IAI enables the specialist to have total access to the monitoring information of any patient, even when not physically in the CCU. Nevertheless, the need to have adequate equipment available, and to pay special attention to the information given, does not cover the function of alerting the physician to important eventualities that are produced in relation to the different patients being monitored. At present, there July/August 1999
are commercial solutions based on the use of pagers, which also touch on this line [16], but they have the inconvenience of only being able to provide very limited alphanumeric information. Recently, commercial graphic solutions have appeared in what can be seen as an improvement in the performance of pagers. We refer to the 1MPACT.wf system by Marquette (http://www.mei.com). This system is capable of receiving alarm calls in text format. Associated with each alarm call are two screens that can be accessed by pressing a button and that show a segment of six seconds corresponding to the different ECG signals monitored. This solution is very interesting, even though it suffers from a number of problems. The first, and in our opinion the most important, is that we are dealing with nonstandard electronics, due to which the price of this type of device could be very high. Furthermore, it has severe limitations with regard to screen resolution (100 x 60 pixels) as well as communication capabilities, which are limited. More specifically, this device is only a receiver, having a memory capable of storing the last 50 alarm calls received. Thus, the user who carries this device is not able to converse with the monitoring system; for example, in order to carry out a consultation with regard to alarms prior to those that are stored in the memory. With the aim of addressing the problems noted, we use a PDA in the form of a
Through remote
“palm-top PC.” From among the multiple types of PDAs available, we have chosen one based on the WindowsCE operating system from Microsoft. The reasons for our choice of a device with these characteristics are as follows: They are devices that fall within the category of consumer electronics, thus they are reasonably priced and can be expected to be even lower in cost. They are small and can be carried by the user at all times. Nevertheless.
.
....... - - . . . . I I. .... - . - ._,
:. ... ... : . . . :
. . . .
. ... .. .. ......, . . . . .. .. .. .. .. . . . . . .. ’ :
.._
.
,
,
.I..
~
.
.,
............
. . . . . . . . : I
...
. . . . . . .
i
.“.
.
......
. . . .
. . .
6. Example of the interaction with the user through the IAI. IEEE ENGINEERING IN MEDICINE AND BIOLOGY
85
their graphic capacity is acceptable (the device that we use has a resolution of 320 x 240 pixels), and in time, solutions with greater resolution and color representation will appear. Their multimedia capabilities are also of interest, as they are able to receive and transmit sound. At a communication level they enable the set-up of bidirectional communication. A channel can be established in multiple forms: by means of direct connection between the PDA and the computer with which it is to converse, or via modem. However, in the near future the appearance of wireless communication cards (via radio frequency) compatible with WindowsCE in Compact Flash format is expected, which is a solution that is the most appropriate for the application we propose (http://www.proxim.com). The application development tools for this type of device are the same as those used for developing applications under Windows951 WindowsNT, which markedly reduces the development t h e if one is familiar with these tools. The manner in which the PAI operates is very similar to the IAI, except that the information that it receives is limited only to that which is associated with the occurrence of alarms, not permitting the visualization of the signal or parameters in real time. In the same manner as in the IAI, the PA1 has a user-identification phase in which the user is allowed to select the set of patients for which he wants to receive alarms, a monitoring phase, and a termination phase. We now go back to concentrate on the monitoring phase, which we consider to be of most interest. Once the user has identified herself to the system and has selected the set of patients that she wishes to follow with the PA1 (the default setting is to select all those in the unit), a logical link is set up in the central web server between the set of monitoring systems selected and the corresponding interface (this link has different characteristics from those of an IAI link). More precisely, the only information that is sent in real time is that associated with the occurrence of an alarm. When this occurs, together with the alarm, the last 20 sec and the last 20 min prior to the occurrence of the alarm are sent, corresponding to the two signals and the two parameters, respectively, identified in the user profile. In any case, the user is able to 86
7. An example of the interaction with the user by means of the PAL
~
8. Example of menus for signal and parameter selection in the PAI. request any other signal or parameter within the available set. At all times the user has access to the alarms that have previously occurred. When one of them is selected, a request is prodrlced with the aim of showing the informhtion corresponding to that alarm along with the set of associated signals and paraheters. For signals, a window of 30 sec is sent (20 sec before the alarm and 10 sec after), and for parameters, 30 min (20 min before the alarm and 10 min after). Id spite of the PDA being capable of receibing signal segments of 30 sec and paraheter segments of 30 min, it is not IEEE ENGINEERING IN MEDICINE AND BIOLOGY
possible to visualize much information 01 the device, due to low resolution of the screen. At each moment in the PDA screen, 2.8 sec of signals and 2.7 min of parameters are shown, in an area of the screen with a width of 210 pixels. In order to explore the entire segment received, there is a horizontal scroll bar in the lower area of the user interface. With regard to vertical resolution, each signal is represented in an area with a height of 70 pixels, and in which the scale factor applied depends on the information shown. As an example of the PAI, Fig. 7 (left-hand section) shows the consultation July/Augusi 1999
of a segment with signals (electrocardiographic derivations MLII and V1) associated with an alarm. In the same manner, on the right-hand side of Fig. 7 a parameter segment is shown (heart rate and deviation of the ST segment in the MLII derivation) that is also associated with the same alarm. Lastly, in the same manner as in the case of the IAI,the generation of menus in the PAI is carried out dynamically, for the same reasons as in the former case. As an example,in Fig. 8 (left-hand side) the state of one of the menus is shown for when signals are being visualized (in this case only the patient’s ECG I and II derivations are being monitored), while in the right-hand side of the same figurethe state of the same menu when we are visualizing parameters appears.
Discussion and Future Work Medical care, particularly the care of critical patients in hospital environments, demands detailed, real-time knowledge of any event that may significantly affect the patient’s condition. At the same time, it requires the evaluation of an enormous quantity of information associated with the temporal monitoring of the patient. New information and communication technologies are headed toward the attainment of these aims, supporting the implementation of network-connected, distributed systems and immediate remote access to information from multiple platforms. Nevertheless, technological support needs to be accompanied by the development of a suitable content, which, in the case of patient monitoring, should enable us to reach levels of competence July/August 1999
and performance worthy of being considered “intelligent.” All these considerations define the setting in which we wish to situate Sutil+ and its different user-system interfaces. The system described in this work is already operative as an experimental system in the General Hospital of Elche (Alicante, Spain), a collaborator in its development, connected to a network of Marquette bedside monitors installed in its CCU. The present limitation with regard to the use of the PA1 should be pointed out, due to the unavailability up until the present of wireless communications for this type of device (although as mentioned, commercialization is imminent). Thus, for the moment, the clinical staff must connect to the system via modem and telephone line. Nevertheless, and in spite of the problems, both the IAI and the PA1 are being frequently used by the CCU staff in the hospital, and the user feedback is very promising with regard to the possible routine use of the system once it is satisfactorily completed. Evidently, our short-term future work will involve improving the performance and the possibilities of the system with regard to both intelligent monitoring and telemonitoring and remote access to monitoring information. To this aim, various paths of advancement have been pointed out throughoutthis work. In the longer term, we are looking at the possibility of giving support to the monitoring of patients dischargedfrom the CCU, and who are in other hospital departments, health centers, or even in their own homes, following the line of increasing interest in the field [17, 18, 191. For this, aspects of the development of the present system will no doubt be of use.
Acknowledgments This work has been financed by CICyT and the Xunta de Galicia under projects 1FD97-0813 and XUGA 20602A97. Senin Barro Ameneiro was born in A Coruiia, Spain in 1962. He received his B.Sc. and Ph.D. (with honors) in physics from the University of Santiago de Compostela, Spain, in 1985 and 1988, respectively. He is a full professor of computer science and head of the Department of Electronics and Computer Science at the University of Santiago de Compostela. IEEE ENGINEERING IN MEDICINE AND BIOLOGY
Before joining this university in 1989, he was an associate lecturer at the Faculty of Informatics of the University of A Coruiia, Spain. His research focuses on signal and knowledge processing (mainly in medical domains), mobile robotics, intelligent fuzzy systems, and artificial neural networks (architectures, applications, and biological modeling). He is the editor of three books and author of over 100 scientific papers in these fields. Professor Barro is a member of the Spanish societies AEIA and AEPIA and the international societies AAAI, ACM, EUSFLAT (of which he is a member of the management board), IEEE, and INNS. Jesu’s Maria Rodriguez Presedo was born in Santiago de Compostela, Spain, in 1967. He received his B.Sc. and Ph.D. in physics from the University of Santiago de Compostela, Spain, in 1989 and 1994, respectively. He is an associate lecturer of computer science at the University of Santiago de Compostela,teaching undergraduateand graduate courses related to computer science, digital control and robotics, and digital signal processing of biomedical signals within the mathematics and physics faculties. His research interests are in the field of digital signal processing of biomedical signals, specifically, intelligent monitoring of patients with ischemic cardiopathies in real time and telemedicine.
Daniel Castro Pereiro was born in A Coruiia, Spain in 1976. He received his B.Sc. in Physics from the University of Santiago de Compostela in 1998. He is a Pre-Doctoral research student in the Department of Electronics and Computer Science, and Ph.D. candidate at this university. His research fields are telemedicineand digital signal processing. Manuel Ferna’ndez Delgado was born in A Coruiia, Spain, in 1971. He received his B.Sc. and Ph.D. in physics from the University of Santiago de Compostela in 1994 and 1999, respectively. His research fields are neuronal computing and intelligent monitoring of physiological signals, mainly ECG. 87
Santiago Fraga Castro was born in Santiago de Compostela, Spain, in 1968. He received the B.Sc. in physics in June 1991fromthe University of Santiago de Compostela. Currently he is a Ph.D. candidate in the Department of Electronics and Computer Science at this university. His research fields are intelligent monitoring of physiological signals and fuzzy logic applications. Manuel Lama Penin was born in Ourense, Spain, in 1971. He received his B.Sc. in physics from the University of Santiago de Compostela in 1994. He is a Pre-Doctoral research student in the Department of Electronics and Computer Science and Ph.D. candidate at this University. His research interests include knowledge acquisition and processing in medical domains and intelligent systems. Jose Antonio Vila Sobrino was born in 1968 in Ourense, Spain. He received the B.Sc. and Ph.D. in physics from the University of Santiago de Compostela in 1991 and 1997. Since November 1992 he has
been an assistant lecturer in the Mathematics and Physics Faculties of this university, where he is teaching undergraduate courses related to computer science and digital signal processing. His research interest is in digital signal processing of biomedical signals and telemonitoring of patients. Address for Correspondence: Sen& Barro Ameneiro, Dept. Electrhica y Computacibn, Facultad de Fisica, Universidad de Santiago, E-15706 Santia g o de Compostela, Spain. Tel: +34-981-563100, ext. 13560. Fax: +34-981-599412. E-mail: senen@ dec.usc.es.
References 1. Special issue on Intelligent Systems for Patient Monitoring and Management. IEEE Eng Med Biol Mag 12(4), 1993.
2. Ravden SJ, Johnson GI: Evaluating Usability of Human-Computer Interfaces: A Practical Method. Ellis Hoorwold Limited, Chicester, West Sussex, UK 1989. 3. Treu S: User Interface Design: A Structured Approach. Plenum Press, New York, 1994. 4. Treu S: User Inteface Evaluation: A StructuredApproach. Plenum Press, New York, 1994. 5 . Vila J, Presedo J, Delgado M, Barro S, Ruiz R, Palacios F: Sutil: Intelligent ischemia monitoring system. I n t J Medial Informatics 47~193-214,1997. 6. Marquette: Solar 7000/8000 patient monitor. Operator’s manual. Marquette Electronics Inc., 1995. 7. The MIT-BIH arrhythmia database CDROM (second edition). Harvard-MIT Division of Health Sciences and Technology, 1992.
8. Fraga S, FCIk P, Lama M, Shnchez E, Barro S: A proposal for a real time signal perception specialist. Proc EIS’98 3:261-267, 1998. 9. Erman L, Hayes-Roth F, Lesser V, Reddy D: The Hearsay-I1 speech understanding system: Integrating knowledge to resolve uncertainty. ACM Computing Surveys 12(2):213-253,1980. 10.Barro S, Bugarin A, FClix P, Ruiz R, Marin R, Palacios F: Fuzzy logic applications in cardiology: Study of some cases. Proc IPMU’94 2: 885-891, 1994. 11.Presedo J, Vila J, Barro S, Palacios F, Rniz R, et al.: Fuzzy modelling of the expert’s knowledge in ECG-based ischaemia detection. F u u y Sets and Systems 77:63-75,1996. 12. Felix P, Fraga S, Marin R, Barro S: Trenddetection based on a fuzzy temporal profile model. Artificial Intelligence in Engineering. In press. 13. Stevens WR: Unix Network Programming. Prentice Hall, NJ, 1990. 14. Horstmann CS, Cornell 6:Core Java. Volume I - Fundamentals. Prentice Hall Java Series, 1997. 15. Horstmann CS, Cornell G: Core Java. Volume II - Advancedfeatures. Prentice Hall Java Series, 1998. 16. Nelwan S, Meij S, Fuchs K, van Dam T: Ubiquitous access to real-time patient monitoring data. Computers in Cardiology 24:211-274, 1997. 17. Zhang Y, Bai J, Zhou X, Dai B, Cui Z, et al.: First trial of home ECG and blood pressure telemonitoring system in Macau. Telemed J 3(1):67-72, 1997. 18. Curry GR, Harrop N: The Lancashire telemedicine ambulance. J Telemed and Telecare 4:231-238, 1998. 19. Silva Cnnha JP, Baptista M, Ribeiro A, Sousa Pereira A: Telecardio: especifica@o tkcnica e funcional do demonstrador de telecardiologia. Internal report, INESC, INESCTEL and DID/CET, Portugal, 1998.
Brain Teasers 1. In an arm-wrestling contest: Jim beat Frank and John; Frank beat Joe, Tom, and John; Joe beat Jim and Tom; Tom beat Jim and John; and John beat Joe. Rank the players in order of ability.
2. Four snails start at the corners of a unit square and move directly toward one another in cyclic order, at a unit rate. How far will they travel before they meet?
88
3. Each of the letters A, B, and C represents a particular digit. What is the minimum value of the whole number ABC divided by (A + B +C)? (answer not unity)
See page 99 for the answers.
IEEE ENGINEERING IN MEDICINE AND BIOLOGY
July/August 1999
