Computer-based System for Weaning of Conscious Patients from Mechanical Ventilation

Computer-based System for Weaning of Conscious Patients from Mechanical VentilationIvan G. Buliev Dept. of Medical Physics, University of Patras, Patras, 26500 GreeceKriton S. FilosDept. of Anaesthesiology andCritical Care Medicine,University of Patras,Patras, 26500 GreeceNicolas PallikarakisDept. of Medical Physics,University of Patras,Patras, 26500 GreeceAbstract-In this paper, we describe a system intended tominimize the duration of the weaning process in patientsundergoing mechanical ventilation. Different medical devices are remotely accessed and parameters like respiratory rate,tidal volume, airway pressures, heart rate, arterial bloodpressures, etc., are collected, analyzed and the current patient status and efforts are estimated. Based on that, the system gives the opportunity for a continuous and automatic refinement of the ventilation settings, according to a defined algorithm.At a preliminary stage, we only monitor different weaningcases, and record patient respiratory- and cardio-circulatory parameters, as well as ventilation settings, chosen by the physicians.Further work will include deriving optimal weaning algorithms, on the base of the collected case records and physicians' experience, and using the system in practice, initially as an advisor.Keywords:Weaning, mechanical ventilation, intelligent control.I.I NTRODUCTIONAlthough the majority of patients receiving mechanical ventilation can be successfully extubated, approximately 20% of ventilated patients [1] need weaning - gradual reduction of mechanical support while they resume spontaneous breathing. A significant fraction of a patient's time in the intensive care unit (ICU) is typically taken up with weaning. Some surveys [2] show that weaning time accounted for over 40% of total ventilator time.For the majority of mechanically ventilated patients this process can be accomplished quickly and easily. There are, however, a significant percentage of patients in whom weaning fails. These patients present a great challenge for clinicians since the pathophysiology underlying a weaning failure is complex, multifactorial, and to a large extent, not well established. Part of the problem probably results from the fact that even excellent physicians often do not accurately judge when a patient is ready to wean. However, it is also true that the clinical approach to weaning, if poorly organized, adds additional time to the duration of mechanical ventilation. Typically, adjustment of the ventilatory parameters is performed only a few times per day, while no adjustments during the night are achieved. Therefore, weaning is unduly prolonged when these changes are not made timely. Furthermore, in case that a patient, for different reasons, needs momentary increase of mechanical support, usually a time delay of up to half an hour can occur until the adjustment is made. This has a negative effect on the duration of the weaning process, even making it impossible in some cases.Currently, weaning tends to be dictated by the experience and intuition of the attending physician who tries to maintain the patient in a state of 'comfort'. However, there is evidence that weaning may proceed more efficiently if directed according to a specified protocol. This allows the weaning process to be applied in form of an algorithm, which could be automated on a computer, thus permitting much more frequent assessment and adjustment of ventilation settings. Obviously, any similar algorithm must utilize feedback control.Several attempts have been made in this direction. Knowledge-based systems [3], [4], fuzzy logic algorithms [5] [6], and adaptive lung ventilation (ALV) controllers [7] are reported to significantly improve the quality of the weaning process. All these systems use on average up to three monitored respiratory parameters, usually limited because of the way of describing the controller function. We intend to monitor, investigate, and take into account larger number of respiratory and cardio-circulatory parameters, namely: respiratory rate (RR), tidal volume (V T), the ratio RR/V T, pulse oxygen saturation (SpO2), end-tidal CO2 partial pressure (P ET CO2), heart rate (HR), arterial blood pressures (BP), mean airway pressure (P AW), peak inspiratory pressure (PIP), and the end-expiratory pressure (PEEP). We believe that any of these parameters could help in determining the more accurate individual weaning support level of the patient.Therefore, the system presented here includes a mechanical ventilator allowing external control and a limited number of devices providing the above mentioned parameters at regular basis.II.S YSTEM D ESCRIPTIONThe components of the system were largely determined by the available instrumentation. At the ICU of the University Hospital of Patras, Greece, we had access to mechanical ventilators - Siemens Servo Ventilator 300A, patient monitors - Spacelabs 91369 and Siemens SC7000, and a non-invasive cardiac output meter - Novametrix NICO2.We based our system on the available mechanical ventilator and added two more devices to complete the acquisition of a required set of breathing and cardiovascular status parameters. We also prepared a ventilator control module, and included a mobile phone for sending alerts in case of incidental malfunctioning of the system components. Fig. 1 presents the structure of the system.A.Servo Ventilator SV300A (Siemens) This is a mechanical lung ventilator intended for adult, pediatric and neonatal patients. It runs supported and controlled ventilation modes, allows customizedrespiratory patterns, and offers context sensitive controls.The SV300A measures a large set of breathingparameters. Ventilation settings, alarm settings, measuredbreath data, and some predefined data trends can berequested and received through an RS232 channel. TheSV300A supports a two-way hand-shaking command-based communication protocol for data transfer. Most of the required for the automatic weaning respiratory parameters are read directly from the ventilator.The ventilator allows an external setting of the ventilation parameters, through a special connector. This makes it appropriate for using it in closed-loop systems for intelligent air support for the patients, aiming at the fastest possible definite weaning.B. Spacelabs 91369 patient monitorThis is a lightweight, portable monitor designed to be used as a compact bedside monitor or as a battery-operated transport monitor. Equipped with its Ultraview SL-A module, it performs electrocardiography recordings (ECG), non-invasive blood pressure (NiBP), SpO 2 and body temperature measurements. It measures the heart rate from either the ECG or SpO 2 channel.The only possibility to obtain data from this device was to monitor its serial printer output and to interpret the texts sent through it. The device does it every 15 sec. C. NICO2 meter (Novametrix)The measurement of cardiac output is accomplished byinterpreting data collected by proprietary sensors thatmeasure flow, airway pressure, and CO 2 concentration,Fig. 1. Schematic representation of the system components and connections.Fig. 2. Block scheme of the ventilator control module.and then combining these signals to calculate CO2 elimination. Using these variables, a Fick partial rebreathing technique is applied to calculate cardiac output.The device is equipped with one-way RS232 interface, through which, on every patient's breath, it send breathing cycle parameters.D.Mobile phoneThe system provides extensive checks for error conditions. All activities are time stamped and recorded into log files. Although the physicians are always nearby and can react, in case of any connection or component failure, sometimes, for different reasons, reaction delays can be accumulated or even system alerts may be missed. Most severe error conditions are reserved by sending warning SMS messages to the technical staff or/and the head of the ICU. Serial connection to a mobile phone has been programmed and tested with two types of phones - Sony J70 and Nokia 3510i, for which communication cables were available.E.Ventilator control boardThe external control of the SV300A is accomplished through 13 analogue inputs (0..+5V), routed to a special connector on the ventilator side panel. Any desired ventilation setting is selected by applying a proportional voltage signal at the corresponding analogue input. A number of digital inputs and outputs are also involved. Since the personal computers (PC), laptops, notebooks are generally not provided with such I/O lines, an intermediate board was needed.The block structure of the control board, which we designed and constructed is shown in Fig. 2. It is based on a Dallas' DS89C450 microcontroller. The power supply comes from the ventilator. The board provides three peripheral connectors - two serial RS232 channels and a DSUB multi-pin connector with 16 analogue outputs, 13 digital input, and 3 digital outputs. An alarm beeper, a real-time clock circuit, E2PROM data memory, are installed on the board. Eight on-board potentiometers can be read and used for setting inputs. A two-line text LCD screen is also installed and displays different information during the controller functioning.A watchdog timer takes care for any functionality failure of the device. In case of board malfunctioning, the control of the ventilator is released to the ventilator knobs. One of the RS232 channels is intended to be connected to a data concentrator from where the controller can collect all patient data and perform weaning control without the help of a personal computer. The microcontroller has sufficient amount of program memory for implementing an autonomic weaning algorithm in the future.The second RS232 channel is used either in service mode for upgrading the program of the microcontroller or normally to connect the PC and receive commands for setting up the ventilation parameters. The basic commands it recognizes are for setting up to a desired value of an analogue or digital output, or for reading a digital input. There are also commands for reading the onboard potentiometers, setting the LCD display mode, and controlling the alarm beeper.F.Personal Computer (Notebook)This is the core of the weaning system. In our case it is an HP Centrino notebook. Notebooks, including the onewe used, do not offer anymore RS232 serial channels. A Fig. 3. A snapshot from the IWC program.USB hub and USB-to-RS232 adapters were used to provide the needed serial channels. The notebook then connects to all medical devices described above, gathers the information they measure and sets remotely the ventilation parameters. All this is done by running dedicated control software, called IWC.The IWC program, written in Borland Builder C/C++ 5.0, offers a friendly graphical user interface (GUI), through which the users are able to configure the system, to monitor all measured parameters, to initiate and stop case recordings, to change to some extent the ventilator settings, and to create and refine different rule-based weaning algorithms. The whole system event history can be optionally recorded, as well as the communication to any of the connected devices. Individual weaning case records can be created and stored for further reference. Exporting to MATLAB is implemented, which will allow fast developing and testing of weaning control algorithms. Remote network access is also being implemented. A snapshot of the main program window is presented in Fig. 3.III.I NITIAL S YSTEM T RIALAt a preliminary stage, we only monitor different weaning cases, and record with a sampling period of 5 seconds patient status parameters and ventilation settings chosen by the physicians. Fig. 4 is an example of a part of weaning case record (non-filtered raw data). It presents selected parameters in a graphical way. Analyzing the recorded data, periods of patient 'discomfort' (lowest subplot) are determined. The criterion was either RR>35 or SpO2<90.IV.D ISCUSSIONThe system proves to work stable and successfully collects weaning data. Up to 58 parameters are monitored, recorded, and can be used to estimate the patient efforts, breathing work, and endurance.The collected data are quite noisy and this is attributed to various reasons, mostly, however, to the non-stationary breathing pattern of the patient. Smart filtering is required before continuing towards weaning control.Collecting of a large database of weaning records could be helpful in several aspects. Analyzing retrospectively the records, efficiency of the present weaning practice can be estimated. Classic control algorithms can be tested and optimized for multivariable input. Other control approaches like fuzzy logic and neural networks can be programmed and tested. Non-obvious rules can be discovered and added to the description of the weaning algorithm.The system is the base for a future research work on creating and testing efficient automatic weaning algorithms, based on closed-loop control.R EFERENCES[1] A. Esteban, I. Alia, J. Ibanez, S. Benito and M.J. Tobin, “Modes ofmechanical ventilation. A national survey of Spanish hospitals,”Chest, vol.106(4), pp.1188-1193, October 1994[2] A. Esteban, and I. Alia, “Clinical management of weaning frommechanical ventilation,” Int. Care Med, vol.24(10), pp.999-1008,October 1998[3]J.H. Strickland Jr, and J.H. Hasson, “A computer-controlledventilator weaning system. A clinical trial,” Chest, vol.103(4), pp.1220-1226, April 1993[4]M. Dojat, A. Harf, D. Touchard, F. Lemaire, and L. Brochard,“Clinical evaluation of a computer controlled pressure support mode,” Am. J. Crit. Care Med, vol.161(4 Pt 1), pp.1161-1166,April 2000[5]T. Nemoto, G.E. Hatzakis, C.W. Thorpe, R. Olivenstein, S. Dial,J.H. Bates, “Automatic control of pressure support mechanicalventilation using fuzzy logic,” Am. J. Respir. Crit. Care Med,vol.160(2), pp.550-556, August 1999[6]J. Schaublin, M. Derighetti, P. Feigenwinter, S. Petersen-Felix,A.M. Zbinden, “Fuzzy logic control of mechanical ventilationduring anaesthesia,” Br. J. Anaesth., vol.77(5), pp636-641, November 1996[7] D.M. Linton, P.D. Potgieter, S. Davis, A.T. Fourie, J.X. Brunner,T.P. Laubscher, “Automatic weaning from mechanical ventilationusing an adaptive lung controller,” Chest, vol.106(6), pp.1843-1850, December 1994Fig. 4. Example of weaning case raw data.。