4 Methods of gait analysis
Walking is often impaired by neurological and musculoskeletal pathology and is one of the health domains of the International Classification of Functioning, Disability and Health (ICF). It is a key aspect in the activities and participation component for mobility and is often adopted as the underlying framework for the assessment of mobility in clinical practice. Therefore it is very important for members of an interdisciplinary team to assess any loss of function in gait.
Gait analysis is used for two very different purposes: to aid directly in the treatment of individual patients and to improve our understanding of gait through research. Gait research use may be further subdivided into ‘fundamental’ studies of walking and clinical research. These topics are explored further in the next chapter. Clearly, no single method of analysis is suitable for such a wide range of uses and a number of different methodologies have been developed.
When considering the methods which may be used to perform gait analysis, it is helpful to regard them as being in a ‘spectrum’ or ‘continuum’, ranging from the absence of technological aids, at one extreme, to the use of complicated and expensive equipment at the other. This chapter starts with a method which requires no equipment at all and goes on to describe progressively more elaborate systems. As a general rule, the more elaborate the system, the higher the cost, but the better the quality of objective data that can be provided. However, this does not imply that some of the simpler techniques are not worth using. It has often been found, particularly in a clinical setting, that the use of high-technology gait analysis is inappropriate, because of its high cost in terms of money, space and time, and because some clinical problems can be adequately managed using simpler techniques.
It is tempting to say that the simplest form of gait analysis is that made by the unaided human eye. This, of course, neglects the remarkable abilities of the human brain to process the data received by the eye. Visual gait analysis is, in reality, the most complicated and versatile form of analysis available. Despite this, it suffers from serious limitations:
In a study on the reproducibility of visual gait analysis, Krebs et al. (1985) found it to be ‘only moderately reliable’. Saleh and Murdoch (1985) compared the performance of people skilled in visual gait analysis with the data provided by a combined kinetic/kinematic system. They found that the measurement system identified many more gait abnormalities than had been seen by the observers.
Many clinicians include the observation of a subject’s gait as part of their clinical examination. However, this is not gait analysis if it is limited to watching the subject make a single walk, up and down the room. This merely gives a superficial idea of how well they walk and perhaps identifies the most serious abnormality. A thorough visual gait analysis involves watching the subject while he or she makes a number of walks, some of which are observed from one side, some from the other side, some from the front and some from the back. As the subject walks, the observer should look for the presence or absence of a number of specific gait abnormalities, such as those described in Chapter 3 and summarised in Table 4.1. A logical order should be used for looking for the different gait abnormalities – the mixture of walking directions listed in the table is not recommended! According to Rose (1983), it is also important, when performing visual gait analysis, to compare the ranges of motion at the joints during walking with those which are observed on the examination plinth – they may be either greater or smaller.
|Gait abnormality||Observing direction|
|Lateral trunk bending||Side|
|Anterior trunk bending||Side|
|Posterior trunk bending||Side|
|Increased lumbar lordosis||Side|
|Circumduction||Front or behind|
|Hip hiking||Front or behind|
|Vaulting||Side or front|
|Abnormal hip rotation||Front or behind|
|Excessive knee extension||Side|
|Excessive knee flexion||Side|
|Inadequate dorsiflexion control||Side|
|Abnormal foot contact||Front or behind|
|Abnormal foot rotation||Front or behind|
|Insufficient push off||Side|
|Abnormal walking base||Front or behind|
The minimum length required for a gait analysis walkway is a hotly debated subject. The authors believe that 8 m (26 ft) is about the minimum for use with fit young people, but that at least 12 m (39 ft) is preferable, since it permits fast walkers to ‘get into their stride’ before any measurements are made. However, shorter walkways are satisfactory for people who walk more slowly. This particularly applies to those with a pathological gait, since the gait pattern usually stabilises within the first two or three steps. A notable exception to this, however, is the gait in parkinsonism, which ‘evolves’ over the first few strides. The width required for a walkway depends on what equipment, if any, is being used to make measurements. For visual gait analysis, as little as 3 m (10 ft) may be sufficient. If video recording is being used, the camera needs to be positioned a little further from the subject and about 4 m (13 ft) is needed. A kinematic system making simultaneous measurements from both sides of the body normally requires a width of at least 5–6 m (16–20 ft). Figure 4.1 shows the layout of a small gait laboratory used for visual gait analysis, video recording and the measurement of the general gait parameters.
Some investigators permit subjects to choose their own walking speed, whereas others control the cycle time (number of steps in a set time) or the cadence (steps per minute), for example by asking them to walk in time with a metronome. The rationale for controlling the cadence is that many of the measurable parameters of gait vary with the walking speed and controlling this provides one means of reducing the variability. However, subjects are unlikely to walk naturally when trying to keep pace with a metronome, and patients with motor control problems may find it difficult or even impossible to walk at an imposed cadence. Zijlstra et al. (1995) found considerable differences in the gait of normal subjects between ‘natural’ walking and ‘constrained’ walking, in which the subject was required either to step in time with a metronome or to step on particular places on the ground. The answer to this dilemma is probably to accept the fact that subjects need to walk at different speeds and to interpret the data appropriately. This means that ‘normal’ values must be available for a range of walking speeds. An unresolved difficulty with this approach is that it may not be possible to get ‘normal’ values for very slow walking speeds, since normal individuals do not customarily walk very slowly and when asked to do so, some of the gait measurements become very variable (Brandstater et al., 1983).
Simply observing the gait and noting abnormalities is of little value by itself. It needs to be followed by gait assessment, which is the synthesis of these observations with other information about the subject, obtained from the history and physical examination, combined with the intelligence and experience of the observer (Rose, 1983). Visual gait analysis is entirely subjective and the quality of the analysis depends on the skill of the person performing it. It can be an interesting exercise to perform visual gait analysis on people walking past in the street, but without knowing their clinical details it is easy to misinterpret what is wrong with them.
When performing any type of gait analysis, one thing that must be constantly borne in mind is that you are observing effects and not causes. Putting it another way, the observed gait pattern is not the direct result of a pathological process but the net result of a pathological process and the subject’s attempts to compensate for it. The observed gait pattern is ‘what is left after the available mechanisms for compensation have been exhausted’ (Rose, 1983).
Historically the use of videotape was widespread in the 1990s, although now this is achieved by direct recording to a DVD, memory card or computer. This has provided one of the most useful enhancements to gait analysis in the clinical setting during recent years. This helps to overcome two of the limitations of visual gait analysis: the lack of a permanent record and the difficulty of observing high-speed events. In addition, it confers the following advantages:
Gait examination by video recording is not an objective method, since it does not provide quantitative data in the form of numbers. However, it does provide a permanent record, which can be extremely valuable. The presence of an earlier recording of a subject’s gait may be used to demonstrate to all concerned how much progress has been made, especially when this has occurred over a long period of time. In particular, it may convince a subject or family member that an improvement has occurred, when memory tells them that they are no better than they were several months ago!
When using video recording, the most practical system consists of camcorder(s). The majority of today’s domestic camcorders are perfectly suitable for use in gait analysis, the requirements being a zoom lens, automatic focus, the ability to operate in normal room lighting and an electronically shuttered charge-coupled device (CCD) sensor, to eliminate blurring due to movement. These allow the user to freeze-frame the picture and to advance frame by frame through successive frames, or to play at a very slow speed to allow the user to observe movements which are too fast for the unaided eye. Many gait laboratories record video data directly into a computer, which may be synchronised with data collection from motion analysis systems (described later), or in some cases used to perform kinematic analysis depending on the number of cameras used. One limitation the majority of camcorders have is the frame rate at which they collect images, commonly 25–30 frames per second, which is not fast enough to pick up some subtleties of movement. However, with the correct software, it is possible to obtain a maximum sampling rate of 50 Hz for PAL (Phase Alternating Line) and 60 Hz for NTSC (National Television Standards Committee) based systems and new developments in video technology include lower cost, higher speed cameras.
In making a thorough visual gait analysis without the use of video recording, the subject needs to make repeated walks to confirm or refute the presence of each of the gait abnormalities listed in Table 4.1. If the subject is in pain or easily fatigued, this may be an unreasonable requirement and it may be difficult to achieve a satisfactory analysis. The use of video recording permits the subject to do a much smaller number of walks, as the person performing the analysis can watch the recording as many times as necessary.
Video recording facilitates the process of teaching visual gait analysis, in which the student often needs to see small abnormal movements which happen very quickly. It is much easier to see such movements if the gait can be examined in slow motion, with the instructor pointing out details on the television or computer monitor. The use of video recording also makes it possible to observe a variety of abnormal gaits which have been archived. A number of teaching animations are now available (Appendix 3).
Showing the subject a video recording of their own gait is not exactly ‘biofeedback’, since there is a time delay involved, but nonetheless it can be very helpful. When a therapist is working with a subject to correct a gait abnormality, the subject may gain a clearer idea of exactly what the therapist is concerned about if they can observe their own gait from the ‘outside’.
Although visual gait analysis using video recording is subjective, it is easy, at the same time, to derive some objective data. The general gait parameters (cycle time or cadence, stride length and speed) can be measured by a method which will be described in the next section. It is also possible to measure joint angles by using some form of on-screen digitiser such as Siliconcoach and Dartfish. Such measurements tend to be susceptible to some errors as the joints may not be viewed from precisely the correct angle, although reasonable joint movements of the lower limb in the sagittal plane and some joint movements in the coronal plane can be obtained.
Individual investigators will find their own ways of performing gait analysis using video recording. A common routine used involves the subject being asked to wear shorts or a swim suit, so that the majority of the leg is visible. It is important that the subject should walk as ‘normally’ as possible, so they are asked to wear their own indoor or outdoor shoes, with socks if preferred. Unless it would unduly tire the subject, it is a good idea to make one or two ‘practice’ walks, before starting video recording. Two camera positions are often used, one viewing from the side (sagittal plane) and the front (coronal plane) (Fig. 4.1). These are first adjusted to show the whole body from head to feet and the subject is recorded as they walk the length of the walkway in one direction. At the end of the walkway, the subject turns around, with a rest if necessary, and is recorded as they walk back again. The whole process may then be repeated with the cameras adjusted to show a close-up of the body from the waist down, or of the body segment of interest.
It is often helpful to mark the subject’s skin, for example using an eyebrow pencil or whiteboard marker, to enhance the visibility of anatomical landmarks on the recording. Hillman et al. (1998) fitted subjects with surface-mounted ‘rotation indicators’, to improve the accuracy with which transverse plane rotations were estimated from video recording.
Subjects should not be able to see themselves on a monitor while they are walking, as this provides a distraction, particularly for children. The cameras should also be placed in a manner so that they are not a focus point. Whether they are shown the video recording afterwards is at the discretion of the investigator, although it is important to review the recording before the subject leaves, in case it needs to be repeated for some reason.
The analysis is performed by replaying the video recording, looking for specific gait abnormalities in the different views and interpreting what is seen in the light of the subject’s history and physical examination. It is particularly helpful if two or more people work together to perform the analysis. Rose (1983) suggested that gait analysis should be based on the team approach, with discussion and hypothesis testing. As will be described in Chapter 5, hypothesis testing may involve an attempted modification of the gait, for instance by fitting an orthosis or by paralysing a muscle using local anaesthesia.
Temporal and spatial parameters of gait, sometimes referred to as the general gait parameters, include cycle time (or stride time), stride length and speed. These provide the simplest form of objective gait evaluation (Robinson and Smidt, 1981) and may be made using only a stopwatch and a tape measure. Other temporal and spatial parameters of gait include, step time, double support time, single support time, step length, base width and foot angle. However these measurements require the use of specialist equipment which will be described in the next section.
Cycle time, stride length and speed tend to change together in most locomotor disabilities, so that a subject with a long cycle time will usually also have a short stride length and a low speed (speed being stride length divided by cycle time). The general gait parameters give a guide to the walking ability of a subject, but little specific information. They should always be interpreted in terms of the expected values for the subject’s age and sex, such as those covered in Chapter 2. Figure 4.2 shows one way in which these data may be presented; the diamonds represent the 95% confidence limits for a normal subject of the same age and sex as the subject under investigation. Although cycle time is gradually replacing cadence in the gait analysis community, it is more convenient to use cadence on plots of this type, since abnormally slow gait will give values on the left-hand side of the graph for all three of the general gait parameters.
Fig. 4.2 • Display of the general gait parameters, with normal ranges appropriate for a patient’s age and sex. Pre- and postoperative values for a 70-year-old female patient undergoing knee replacement surgery.
Cycle time or cadence may be measured with the aid of a stopwatch, by counting the number of individual steps taken during a known period of time. It is seldom practical to count for a full minute, so a period of 10 or 15 seconds is usually chosen. The loss of accuracy incurred by counting for such short periods of time is unlikely to be of any practical significance. The subject should be told to walk naturally and they should be allowed to get up to their full walking speed before the observer starts to count the steps. The cycle time is calculated using the formula:
Stride length can be determined in two ways: by direct measurement or indirectly from the speed and cycle time. The simplest direct method of measurement is to count the strides taken while the subject covers a known distance. More useful methods include putting ink pads on the soles of the subject’s shoes and walking on paper (Rafferty and Bell, 1995), using marker pens attached to shoes (Gerny, 1983) and a more messy option is to have the subject step with both feet in a shallow tray of talcum powder and then walk across a polished floor or along a strip of brown wrapping paper or coloured ‘construction paper’, leaving a trail of footprints. These may be measured, as shown in Figure 2.3, to derive left and right step lengths, stride length, walking base, toe out angle and some idea of the foot contact pattern. This investigation is able to provide a great deal of useful and surprisingly accurate information, for the sake of a few minutes of mopping up the floor afterwards! As an alternative to using talcum powder, felt adhesive pads, soaked in different coloured dyes, may be fixed to the feet (Rose, 1983). The subject walks along a strip of paper and leaves a pattern of dots, which give an accurate indication of the locations of both feet.
The multiplication by ‘2’ converts steps to strides and by ‘60’ converts minutes to seconds. For accurate results, the cycle time and speed should be measured during the same walk. However, the simultaneous counting, measuring and timing may prove too difficult and the errors introduced by using data from different walks are not likely to be important, unless the subject’s gait varies markedly from one walk to another.
The speed may be measured by timing the subject while he or she walks a known distance, for example between two marks on the floor or between two pillars in a corridor. The distance walked is a matter of convenience but somewhere in the region of 6–10 m (20–33 ft) is probably adequate. Again, the subject should be told to walk at their natural speed and they should be allowed to get into their stride before the measurement starts. The speed is calculated as follows:
Determination of the general gait parameters from a video recording of the subject walking is easy, providing the recording shows the subject passing two landmarks whose positions are known. One simple method is to have the subject walk across two lines on the floor, a known distance apart, made with adhesive tape. Space should be allowed for acceleration before the first line and for slowing down after the second one. When the recording is replayed, the time taken to cover the distance is measured and the steps taken are counted. It is easiest to take the first initial contact beyond the start line as the point to begin both timing and counting and the first initial contact beyond the finish line as the point to end both timing and counting. The first step beyond the start line must be counted as ‘zero’, not ‘one’. This method of measurement is not strictly accurate, since the position of the foot at initial contact is an unknown distance beyond the start and finish line, but the errors introduced are unlikely to be significant. As mentioned above, this method can also be employed without the use of video recording.
A number of systems have been described which perform the automatic measurement of the timing of the gait cycle, sometimes called the temporal gait parameters. Such systems may be divided into two main classes: footswitches and instrumented walkways. Figure 4.3 shows typical data, which could be obtained from either type of system.
Footswitches are used to record the timing of gait. If one switch is fixed beneath the heel and one beneath the forefoot, it is possible to measure the timing of initial contact, foot flat, heel rise and toe off, and the duration of the stance phase (see Figs 2.2 and 4.3). Data from two or more strides make it possible to calculate cycle time and swing phase duration. If switches are mounted on both feet, the single and double support times can also be measured. The footswitches are usually connected through a trailing wire to a computer, although alternatively either a radio transmitter or a portable recording device may be used to collect the data and transfer them to the measuring equipment.
A footswitch is exposed to very high forces, which may cause problems with reliability. This has led to many different designs being tried over the years. A fairly reliable footswitch may be made from two layers of metal mesh, separated by a thin sheet of plastic foam with a hole in it. When pressure is applied, the sheets of mesh contact each other through the hole and complete an electrical circuit. Foot-switches are most conveniently used with shoes, although suitably designed ones may be taped directly beneath the foot. Small switches may also be mounted in an insole and worn inside the shoe. In addition to the basic heel and forefoot switches, further switches may be used in other areas of the foot, to give greater detail on the temporal patterns of loading and unloading. In addition to the home-made varieties, a number of companies also manufacture footswitches.
An instrumented walkway is used to measure the timing of foot contact, the position of the foot on the ground, or both. Many different designs have been developed, usually individually built for a single laboratory. The descriptions which follow refer to typical designs, rather than to any particular system.
A conductive walkway is a gait analysis walkway which is covered with an electrically conductive substance, such as sheet metal, metal mesh or conductive rubber. Suitably positioned electrical contacts on the subject’s shoes complete an electrical circuit. The conductive walkway is thus a slightly different method of implementing footswitches and provides essentially the same information. Again, the subject usually trails an electrical cable, which connects the foot contacts to a computer. The speed needs to be determined independently, typically by having the body of the subject interrupt the beams of two photoelectric cells, one at each end of the walkway, again connected to the computer. Timing information from the foot contacts is used to calculate the cycle time, and the combination of cycle time and speed may be used to calculate the stride length.
An alternative arrangement is to have the walkway itself contain a large number of switch contacts, which detect the position of the foot, as well as the timing of heel contact and toe off. This has the advantage that no trailing wires are required and the walkway can be used to measure both step lengths and the stride length. A number of commercial systems are available to make this type of measurement, often also providing some information on the magnitude of the forces between the foot and the ground. One such system, which is now in common use, is the ‘GAITRite’ (Bilney et al., 2003; Menz et al., 2004) (Fig. 4.4).
Kinematics is the measurement of movement or, more specifically, the geometrical description of motion, in terms of displacements, velocities and accelerations. Kinematic systems are used in gait analysis to record the position and orientation of the body segments, the angles of the joints, and the corresponding linear and angular velocities and accelerations.
Following the pioneering work of Marey and Muybridge in the 1870s, photography remained the method of choice for the measurement of human movement for about 100 years, until it was replaced by electronic systems. Two basic photographic techniques were used: cine photography and multiple-exposure photography. Cine photography is achieved by the use of a series of separate photographs, taken in quick succession. Multiple exposure photography has existed in many different forms over the years. It is based on the use of a single photograph, or a strip of film on which a series of images are superimposed, sometimes with a horizontal displacement between each image and the next. The 1960s and 1970s saw the development of gait analysis systems based on optoelectronic techniques, including television, and these have now superseded photographic methods. The general principles of kinematic measurement are common to all systems and will be discussed before considering particular systems in detail.
Kinematic measurement may be made in either two dimensions or three. Three-dimensional measurements usually require the use of two or more cameras, although methods have been devised in which a single camera can be used to make limited three-dimensional measurements.
The simplest kinematic measurements are made using a single camera, in an uncalibrated system. Such measurements are fairly inaccurate but they may be useful for some purposes. Without calibration, it is impossible to measure distances accurately and such a system is usually used only to measure joint angles in the sagittal plane. The camera is positioned at right angles to the plane of motion and as far away as possible, to minimise the distortions introduced by perspective. To give a reasonable size image, with a long camera-to-subject distance, a ‘telephoto’ (long focal length) lens is used. The angles measured from the image are projections of three-dimensional angles onto a two-dimensional plane and any part of the angulation which occurs out of that plane is ignored. Commercial systems of this type are available for measuring joint angles from television images. Such measurements may be subject to yet another form of error, since the horizontal and vertical scales of a television image may be different.
A single-camera system can be used to make approximate measurements of distance, if some form of calibration object is used, such as a grid of known dimensions behind the subject. Measurement accuracy will be lost by any movement towards or away from the camera, but this effect can again be minimised if the camera is a long distance away from the subject, using a telephoto lens. Angulations of the limb segments, either towards or away from the camera, will also interfere with length measurements.
To achieve reasonable accuracy in kinematic measurement, it is necessary to use a calibrated three-dimensional system, which involves making measurements from more than one viewpoint. A detailed review of the technical aspects of the three-dimensional measurement of human movement was given in four companion papers by Cappozzo et al. (2005), Chiari et al. (2005), Leardini et al. (2005) and Della Croce et al. (2005). Although there are considerable differences in convenience and accuracy between cine film, video recording, television/computer and optoelectronic systems, the data processing for the different types of three-dimensional measurement systems is similar.
Most commercial kinematic systems use a three-dimensional calibration object, which is viewed by all the cameras, either simultaneously or in sequence. Computer software is used to calculate the relationship between the known three-dimensional positions of ‘markers’ on the calibration object and the two-dimensional positions of those markers in the fields of view of the different cameras. An alternative method of calibration is used by the Codamotion system, whose optoelectronic sensors are in fixed relation to each other, permitting the system to be calibrated in the factory.
When a subject walks in front of the cameras, the calibration process is reversed and three-dimensional positions are calculated for the markers fixed to the subject’s limbs, so long as they are visible to at least two cameras. Data are collected at a series of time intervals known as ‘frames’. Most systems have an interval between frames of either 20 ms, 16.7 ms or 5 ms, corresponding to data collection frequencies of 50 Hz, 60 Hz or 200 Hz, with some systems now offering frame rates of up to 500 Hz and beyond. When a marker can be seen by only one camera, its three-dimensional position cannot be calculated, although it may be estimated by ‘interpolation’, using data from earlier and later frames.
All measurement systems, including the kinematic systems to be described, suffer from measurement errors. Measurement accuracy depends to a large extent on the field of view of the cameras, although it also differs somewhat between the different systems. The earlier systems had measurement errors of 2–3 mm in all three dimensions, throughout a volume large enough to cover a complete gait cycle (Whittle, 1982). Recently, design and (especially) calibration improvements have reduced typical errors to less than 1 mm. Some commercial systems claim to provide much higher accuracy than this, but the authors treat such claims with scepticism, particularly when applied to the measurement of moving markers under realistic gait laboratory conditions.
Technical descriptions of kinematic systems use, and sometimes misuse, the terms ‘resolution’, ‘precision’ and ‘accuracy’. In practical terms, resolution means the ability of the system to measure small changes in marker position. Precision is a measure of system ‘noise’, being based on the amount of variability there is between one frame of data and the next. For the majority of users, the most important parameter is accuracy, which describes the relationship between where the markers really are and where the system says they are!
Most commercial systems are sufficiently accurate to measure the positions of the limbs and the angles of the joints. However, the calculation of linear or angular velocity requires the mathematical differentiation of the position data, which magnifies any measurement errors. A second differentiation is required to determine acceleration, and a small amount of measurement ‘noise’ in the original data leads to wildly erratic and often unusable results for acceleration. The usual way of avoiding this problem is to smooth the position data, using a low-pass filter, before differentiation. This achieves the desired object but means that any genuinely high accelerations, such as that at the heelstrike transient, may be lost.
Thus, kinematic systems are good at measuring position but poorer at determining acceleration, because of the problems of differentiating even slightly noisy data. Conversely, accelerometers are good at measuring acceleration but poor at estimating position, because of the problems of integrating data with baseline drift. Really accurate data could be obtained by combining the two methods, using each to correct the other and calculating the velocity from both. Some research studies have been conducted using this combined approach.
As well as the errors inherent in measuring the positions of the markers, further errors are introduced because considerable movement may take place between a skin marker and the underlying bone. A few studies have been performed (e.g. Holden et al., 1997; Reinschmidt et al., 1997) in which steel pins were inserted into the bones of ‘volunteers’ (usually the investigators themselves) and the positions of skin markers compared with the positions of markers on the pins. The amount of skin movement revealed by such studies is generally somewhat worrying! The amount of error this causes in the final result depends on which parameter is being measured. For example, marker movement has little effect on the sagittal plane knee angle, because it causes only a small relative change in the length of fairly long segments, but it may cause considerable errors in transverse plane measurements or on measurements involving shorter segments, such as in the foot. In some cases, the magnitude of the error is greater than the measurement itself! Skin movement may also introduce errors in the calculation of joint moments and powers. A possibility for the future is to correct for marker movement, by estimating the movement relative to the underlying bone. A further error is introduced when the positions of the joints are estimated from anthropometric measures (e.g. leg length) and the positions of skin markers, particularly where it is possible to place the markers in the wrong position. Even for subjects with normal anatomy, these errors can be substantial; for patients with bony deformity, the errors may be even greater. For these reasons the development of marker sets and anatomical models has been at the forefront of the evolution of gait analysis alongside the ability of new motion analysis systems capable of coping with larger numbers of markers.