detection of the biological signal, by a sensing device which responds to a characteristic signal in the form of electrical, mechanical, electromagnetic, chemical or thermal energy

 transduction, in which the output from the sensor is converted into another form of energy, usually to a continuous electrical signal

 amplification and signal processing to extract and magnify the relevant features of the signal and reduce unwanted noise

 display and storage – the output from the instrument is presented to the operator. Storage for future use may be achieved using mechanical markers, printed copy or computer memory.

 continuous real-time detection, processing and recording of measurements

 increased range of measurements possible

 miniaturization of complicated and powerful instruments

 sophisticated artefact rejection and noise reduction algorithms

 complex on-line mathematical and statistical signal processing in upgradeable software, e.g. Fourier analysis of the EEG

 automated control of the apparatus and the timing and process of measurement, and integration of alarms

 storage in memory, permitting trend analysis, future display and further analysis

 user-friendly audio-visual display of recordings, integrating many simultaneous clinical measurements, and able to be customized by the user

 less maintenance than analogue instruments.

 dependence on electrically powered equipment

 degradation of clinical skills and alternative manual measurements through disuse

 impoverished understanding of the principles of complex measuring equipment and the requirements for correct use

 illusion of the unquestionable accuracy of measurements produced by expensive computer-controlled equipment and presented on an impressive display or typed copy.

The Importance of Repeated Measurements

Differences in repeated clinical measurements arise from three causes:

The anaesthetist must be satisfied with the accuracy and precision of any clinical measurement used in patient management. Repeated measurements which are consistent ensure that the measurement is representative, i.e. precise, but do not ensure accuracy. For example, repeated recordings of invasive arterial pressure may be extremely consistent, but erroneous if the transducer is not calibrated against the correct zero point. Defences against the uncritical acceptance of inaccurate measurements include meticulous care in calibrating instruments and recording of clinical measurements, and reflection on clinical measurements which do not fit the clinical state of the patient or other related measurements. A discrepant result should be rechecked, using a different measurement technique if possible, before it is used to change patient management. This is especially true of complex, operator-dependent techniques such as measurement of cardiac output.

Zero Stability

The ability of a measurement to maintain a zero reading on the display or record when the input signal is zero defines the zero stability. The importance of zero instability depends on the magnitude relative to the signal, e.g. a zero drift of a few millimetres of mercury is much less important for the measurement of arterial pressure than it is for intracranial pressure.

Gain Stability

The majority of biological signals are amplified before reproduction. This ‘gain’ may be fixed or controlled by the user. When set, this should remain constant over the period of recording.

Amplitude Linearity

The degree of amplification of the signal should be constant over the whole range of signal amplitudes. Manufacturers usually specify the degree of linearity of electronic components over a certain amplitude range. The amplitude linearity of a complete clinical system may be confirmed easily in an electronics laboratory by comparing the output to known, standardized test signals.

Hysteresis

Some instruments, such as thermistors and humidity sensors, may display hysteresis. This is a special case of non-linearity, in which the output differs depending on whether the input signal is increasing or decreasing.

Frequency Response

Many biological signals vary in a complicated and rapidly changing pattern. Accurate reproduction of a complex waveform requires that all of the component frequencies which make up the waveform are processed in an identical manner. This requires more than simply equal amplification irrespective of frequency, i.e. no amplitude distortion. It also implies that the relative positions of the various frequency components of the waveform are not shifted, i.e. no phase distortion. In practice, accurate reproduction up to the 10th harmonic of the fundamental frequency is sufficient for clinical purposes, e.g. 30 Hz for an arterial pressure waveform associated with heart rates up to 180 beats min^–1 (3 Hz).

Signal-to-Noise Ratio

Biological signals are obscured to a variable degree by unwanted or extraneous signals which have similar physical characteristics and are described as noise, e.g. heart sounds become difficult to detect in the presence of continuous, noisy breath sounds. The efficiency of isolation of the signal from unwanted biological signals and electronic noise sources in the equipment is defined by the signal-to-noise ratio. The variability of the amplitudes of signal and noise is enormous and the signal-to-noise ratio is described using a logarithmic scale of decibels. Microvolt EEG measurements are particularly susceptible to noise from many sources. Biological noise includes contaminating ECG and EMG potentials, particularly from the scalp muscles, and interference from electrochemical activity at the skin–electrode interface. Electrostatic and electromagnetic linkage between the recording wires and nearby sources of mains electricity generates noise which is predominantly 50 Hz frequency and harmonics. Radiofrequency noise from diathermy or transmitters may also be picked up at this stage. Physical disturbance of the recording wires causes tiny changes in capacitive potentials and may add low-frequency noise, called microphony. Thermal noise is added during amplification, particularly at the input stage when the signal is in the microvolt range. Good amplifier design, electronic filtering of unwanted frequencies and modern techniques of digital signal processing may extract small signals from considerable background noise, but this inevitably introduces some distortion of the signal. Prevention of contamination of the signal by minimizing sources of noise before the signal is amplified is always preferable. The operator is responsible for correctly using measuring instruments to optimize the signal and for applying knowledge of the physical principles of the measurement to minimize contamination by noise.

Mechanical Measuring Instruments

Measuring instruments based on mechanical principles lack the flexibility and automated control of computerized devices, but use ingenious methods for processing and displaying analogue measurements. For example, mechanical spirometers use precision-engineered gears to translate the movement of a piston or vane into the rotation of a calibrated dial.

Analogue Computers

Analogue computers use hardware comprising electronic circuits and operational amplifiers. Signals are processed in the form of continuously variable electric potentials. Analogue hardware components continuously perform a wide variety of mathematical functions on a rapidly changing input waveform. Integration and differentiation are formidable mathematical tasks for a digital computer, which can be solved simply and cheaply using analogue circuits comprising capacitors and resistors. Integration of the flow signal from a pneumotachograph produces a volume waveform.

Microcomputers and Digital Signal Processing

Digital signal processing offers a powerful alternative to mechanical processing and analogue computation. A fundamental step in this process is the conversion of a continuous analogue electrical signal into a discrete digital form. This analogue-to-digital conversion is achieved by measuring or ‘sampling’ the continuous input signal at regular intervals, to produce a series of discrete measurements over time which are in a suitable format for digital computation. The overwhelming advantage of digital processing is that the manipulation of the digitized signal is performed by a flexible and unlimited series of software calculations which range from mathematical functions to the analysis of statistical properties and trends.

Analogue-to-Digital Conversion

The core processing units of digital computers assume one of two stable states, i.e. a binary, rather than decimal, code. This imposes a limit on the resolving power of the digital processor, although with increasing processing power this limit has become negligible. Earlier 8-bit computing comprised a binary number of eight digits representing 2⁸ integer decimal numbers, from binary 00000000 = decimal 0 to binary 11111111 = decimal 255. In short, an 8-bit converter can resolve an analogue signal with an accuracy of one part in 255, i.e. with an amplitude resolution of 0.4% of full scale. A 12-bit converter is more accurate, with a resolving power of one part in 4095 or 0.02% of full scale. More modern processors are capable of 32-bit computing corresponding to a range of 2³² integer decimal numbers (a resolving power of 0.00000002%), with 64-bit processors now commonly found in domestic computer equipment. Whilst highly accurate, the cost of this improvement in resolution is more expensive hardware to digitize, process and store considerably more digital information.

Amplitude resolution is not the only determinant of the accuracy of analogue-to-digital conversion. Resolution over time, determined by the sampling frequency, is also important. A relatively low sampling frequency may provide a representative sample of values for a slowly changing waveform but it may inadequately represent high-frequency components and introduce an aliasing error, in which different signals become indistinguishable. The Nyquist theorem suggests that the minimum sampling frequency to maintain the integrity of the waveform is at least twice the highest frequency component with significant amplitude in the input signal waveform, e.g. a sampling frequency of 100 Hz would adequately capture the fastest rate of change in a physiological pressure signal.

The immensely powerful and complicated hardware and software programming instructions responsible for performing the tasks of digitizing, processing, storing and displaying the input signal are hidden from view in the commercial ‘black box’.

Analogue Displays

A continuously variable signal, such as pressure or temperature, is represented by an analogue display in terms of the amplitude of a physical quantity on a calibrated scale, dial, electrical meter or printed record. The glass thermometer incorporates a wedge-shaped lens which magnifies the appearance of the mercury column against the calibrated background scale. The height of a water column manometer is a linear, visual scale of pressure. Simple mechanical displays are accurate and easily understood, but are inconvenient to read and most suitable for intermittent discrete measurements.

Mechanical spirometers and flowmeters record flow on a dial driven by gears. Electrical moving coil meters use a coil of wire suspended in a magnetic field which rotates in proportion to the applied current and moves a pointer on a calibrated dial. Alternatively, the amplified and filtered electrical signal could drive a chart recorder which produces a continuous printed record of the amplitude of measurements against time. Limitations common to these mechanical devices include fragile moving parts, and inertia which impairs the frequency response to rapidly changing signals.

The cathode ray oscilloscope is an effective screen-based display for continuous analogue electrical signals. A heated cathode generates a stream of electrons which are focused and accelerated onto a phosphorescent coating which lines the flat surface of the tube to generate a bright spot. The position of the electron beam in both x- and y-axes is controlled by electrostatic plates. The continuously varying input signal is applied to the y-plates so that deflection in the vertical y-axis is proportional to the amplitude of the signal. The absence of mechanical parts results in a high-frequency response. An electronic time-base circuit delivers a saw-tooth voltage to the x-plates which drives the electron beam across the x-axis at a constant rate and returns the beam to the left-hand side at the start of each sweep. This produces a dynamic image of signal amplitude against time. Alternatively, a second input signal may be applied to the x-plates to produce an x-y graphical plot, e.g. pressure–volume loop. Cathode ray oscilloscopes are widely used in electronic engineering and signal processing, but have been replaced in clinical practice by microprocessor-controlled displays.

Microprocessor-Controlled Displays

Digital signal processing has revolutionized clinical measurement. Modern monitors comprise a single system integrating various measurements of physiology, and display information as discrete measurements as well as continuous analogue waveforms (Fig. 16.1). This paradox, the conversion of analogue information into digital and then back to analogue, illustrates the real power of digital signal processing to manipulate and present information in a relevant and user-friendly manner. Data patterns (waveforms, trends, graphs) can be recreated or processed in other ways from the original digital signal to assist the anaesthetist, with the original digital signal being stored in a computer record without degradation of the quality of the signal.

Most of the current monitoring systems follow good ergonomic principles, with different variables separated consistently by position on the screen and by colour. This allows the most important information to be placed centrally in large symbols or fonts and in bold colours, with less important data either relegated to small print, or placed in submenus. However, the flexibility of most monitors implies that it is still possible for individuals to change colours and priorities, often making the monitor much less effective. Whenever possible, departments should ensure that all monitors have identical default settings to reduce confusion (these are usually password protected). Unfortunately, the lack of international standards means that confusion may still occur if monitors from multiple sources are used in the same unit.

Despite many attempts to simplify patient data into geometric shapes or bar graphs, data continue to be displayed most often as simple numbers, supported by waveforms, e.g. invasive pressure, and a graphical display of trends over time. Trends are particularly useful when clinical problems may produce gradual change. For example, in neurosurgery, a gradual decrease in end-tidal carbon dioxide concentration is often associated with multiple air emboli.

Input Impedance and Common Mode Rejection

Amplifiers for biological signals require high common mode rejection and high input impedance. The input and electrode impedances act as a potential divider: high electrode impedance and low amplifier input impedance attenuate the electrical signal across the amplifier. The input impedance of modern amplifiers exceeds 5 MΩ to avoid problems, and careful attention must be paid to minimizing electrode impedance, particularly for EEG recordings.

Differential amplification is a powerful method of reducing unwanted noise. The potential difference between two input signals is amplified, but electrical signals common to both are attenuated. This feature is termed ‘common mode rejection’ and very effectively reduces mains interference in all biological signals and electrocardiographic contamination of much smaller electroencephalographic signals. The common mode rejection ratio (CMRR) for a typical differential amplifier exceeds 10 000:1. In other words, a signal applied equally to both input terminals would need to be 10 000 times larger than a signal applied between them for the same change in output.

Frequency Response

The bandwidth of the amplifier must cover the range of frequencies which are important in the signal. In practice, amplifiers require a flat frequency response for ECG from 0.14 to 50 Hz, for EEG from 0.5 to 100 Hz and for EMG from 20 Hz to at least 2 kHz.

Low-frequency interference, largely caused by slow fluctuating potentials generated in the electrodes, produces baseline instability and drift. This is removed by incorporating a network of resistors and capacitors which function as a simple high-pass filter allowing biological signals to pass, but attenuating low-frequency noise. This introduces a compromise in amplifier design between signal trace fidelity and stability of recording. For example, amplifiers designed for diagnostic electrocardiography have long time constants with optimal reproduction of the waveform at the expense of baseline instability, especially to movement. In comparison, continuity of recording is more important when the electrocardiogram is used for monitoring during anaesthesia; high-pass filtering produces a short time constant and good baseline stability at the expense of waveform reproduction. Low-frequency elements of the ECG, such as the T wave, may become differentiated by phase shift in the high-pass filter and appear distorted or biphasic.

Other filters can attenuate particular frequencies. Highly selective band reject filters attenuate 50 Hz interference from the signal. Low-pass filters are used to eliminate higher-frequency artefacts from an EEG signal. The purpose of filtering is to reduce unwanted noise relative to the signal. When the frequency range of signal and noise overlap, some degree of signal degradation is inevitable.

Noise and Interference

Electrical noise arising from the patient, the patient-electrode interface or the surroundings may seriously interfere with accurate recording of biological potentials.

Indirect Methods

An adequate arterial pressure is essential for tissue perfusion; even when perfusion is adequate, hypotension may lead to renal failure. Indirect methods of measuring arterial pressure do not depend on contact between arterial blood and the system for signal recognition and transduction. Arterial pressure may be most rapidly estimated by palpating a pulse, although this method is too unreliable as a single technique. The majority depend on signals generated by the occlusion of a major artery using a cuff, known as the Riva-Rocci method. Systolic pressure can be estimated by the return of a palpable distal pulse; auscultation of the Korotkoff sounds can determine systolic and diastolic pressures. These methods, however, are too time-consuming during anaesthesia and often impossible because of poor access to the patient’s arm.

Oscillometric Measurement of Arterial Pressure

The oscillometric measurement of arterial pressure estimates arterial pressure by analysis of the pressure oscillations which are produced in an occluding cuff by pulsatile blood flow in the underlying artery during deflation of the cuff (Fig. 16.3). The original automatic oscillometers used two cuffs. The upper cuff was inflated to occlude the arterial flow and then gradually deflated. As the blood flow began to pass under the upper cuff, the small changes in volume were detected by the lower cuff with an electromechanical pressure transducer. Modern machines use a single cuff with two tubes for inflation/measurement. During slow deflation, each pulse wave produces a pressure transient in the cuff which may be distinguished from the slowly decreasing ambient pressure in the cuff. Above systolic pressure, the transients are small, but suddenly increase in magnitude when the cuff pressure reaches the systolic point. As the cuff pressure decreases further, the amplitude reaches a peak and then starts to diminish. The mean arterial pressure correlates closely with the lowest cuff pressure at which the maximum amplitude is maintained. As the cuff pressure reaches diastolic pressure, the transients abruptly diminish in amplitude. To avoid high cuff pressures and long deflation times, monitors inflate the cuff to just above a normal systolic pressure and then slowly decrease the pressure until a pulse is detected. Consequently, estimates of diastolic pressure can be unreliable. If a pulse is not detected, the cuff is then inflated to a higher pressure. This process may be repeated several times before a measurement is made.

Commercial instruments incorporate mechanisms for improving the reliability of the measurement. For example, at each successive plateau pressure during the controlled deflation, successive pressure fluctuations are compared and accepted only if they are similar. All automatic oscillometric instruments require a regular cardiac cycle with no great differences between successive pulses. Accurate and consistent readings may be impossible in patients with an irregular rhythm, particularly atrial fibrillation.

Clinical studies comparing automatic oscillometric instruments with direct arterial pressure have demonstrated good correlation for systolic pressure with a tendency to overestimate at low pressures and underestimate at high pressures. Mean and diastolic pressures were less reliable. The 95% confidence interval for all three indices exceeded 15 mmHg. The disadvantages of automated oscillometry are shown in Table 16.3.

Direct Measurement

To measure arterial pressure directly, a cannula (usually 20–22G parallel-sided Teflon) must first be inserted into an artery (usually the radial because occlusion of the artery may be compensated for by flow through the ulnar artery). As fluids are incompressible, the pressure in the artery is transmitted directly to a transducer, which converts pressure into an electrical signal which is displayed by the monitor. The cost and complexity of pressure transducers are compensated by convenience, accuracy, continuity of measurement and an electrical output which may be processed, stored and displayed according to the requirements. The transducer should be at the level of the left ventricle and the transducer opened to the atmosphere to provide a zero reading before use. Monitors usually display systolic and diastolic pressures as well as the mean pressure, calculated automatically by integrating the area under the pressure waveform. The waveform provides useful additional information: a rapidly appreciated estimate of pressure, a qualitative assessment of the adequacy of the frequency response and damping, and an assessment of relative hypovolaemia during positive pressure as identified by the variability or ‘swing’ in the waveform.

The advantage of direct measurements is a real-time measure of arterial pressure, which is essential when administering drugs such as vasopressors to critically ill patients (Table 16.4). Such measurement systems also provide a means for obtaining samples for arterial blood gas analysis and other blood tests. The use of arterial cannulae has therefore become standard practice for severely ill patients, both in the operating theatre and in the intensive care unit.

However, errors are common as a result of malpositioning of the transducers and failure to zero the transducer before use. For example, if the operating table is moved upwards while the transducer remains static, the difference in height artificially increases the pressure reading. Further, while modern disposable sets are usually reliable and accurate, they may occasionally malfunction. Unusual readings should therefore be checked against a reading from a non-invasive monitor. Complications relating to arterial cannulae are shown in Table 16.5.

Accuracy of Arterial Pressure Measurements

The accuracy of pressure measurements, particularly using indirect methods, needs to be considered. Invasive direct measurement of arterial pressure is the usual standard for comparison. However, the catheter-transducer system must be carefully set up and tested for optimal performance and this is hard to achieve in clinical practice. Arterial pressure varies throughout the arterial tree and the measured pressure depends on the site of measurement. As the pulse wave travels from the ventricle to peripheral arteries, changes in vessel diameter and elasticity affect the pressure waveform, which becomes narrower with increased amplitude. Differences in arterial pressure between limbs are common, particularly in patients with arterial disease.

Indirect methods using an occluding cuff make intermittent measurements, with the systolic and diastolic readings reflecting the conditions in the artery at two instants at which the end-points are detected. By contrast, direct pressure measurements are the average of a number of cycles, more precisely reflecting mean pressures. Indirect measurements may be compromised by taking a small number of infrequent samples from a variable signal.

 Long catheters inserted via the antecubital fossa are relatively easy and safe to insert but are of small diameter. Catheters inserted via the basilic or cephalic vein are sometimes difficult to advance past the shoulder. It is also difficult to determine if the tip of the catheter is within a central vein without X-ray imaging. Thrombosis of the veins is common if the catheter is left in situ for more than 24 h.

 Femoral venous catheters are inserted just below the inguinal ligament. They are also relatively easy to insert and may be of large gauge to allow rapid transfusion of fluids. This route is often chosen in children. However, the site of insertion is often within a skin fold, making skin sepsis more likely.

 Internal jugular catheters are used most commonly because the vein is superficial, of larger diameter, and easily managed. This is the route which is often most appropriate for use in an emergency. However, the insertion point is adjacent to several vital structures, including the carotid artery, lung, brachial plexus and cervical spine, with the result that direct needle trauma to these structures can occur. Current guidelines recommend the use of an ultrasound probe for insertion of a catheter via the internal jugular route to improve the accuracy of insertion and to minimize complications.