Feature Article

By: D. Hodgins, European Technology for Business (ETB) Ltd, Codicote, UK
G. Evans,London Knee Clinic, London, UK
R. Orpwood, Bath Institute of Medical Engineering, Bath, UK

The measurement of a person’s gait is an important part of diagnosing a variety of medical conditions, yet it is rarely performed objectively.¹ This is because there is no equipment that can be readily used in an outpatient clinic or home environment. To diagnose a knee injury, the orthopaedic surgeon will ask the patient to walk and to state when pain occurs. The knee flexion angle when stationary may then be measured using a goniometer, or more generally, estimated by eye. A course of treatment will follow that may involve surgery and/or a specific exercise programme. During rehabilitation the same procedures will be used at defined intervals. At the end, the orthopaedic surgeon and patient have their subjective views of the success of the treatment. There is no quantifiable information to support those views. The same scenario exists for anyone undergoing medical treatment or physiotherapy that affects their way of walking.

 

The problem of being unable to quantify a person’s gait outside of the “gait laboratory” environment has been known for some time. However, it is only with the evolution of new microsensors that solutions have been forthcoming. In the past few years sensor based products have been produced that are small and light enough to be attached to the lower limbs without affecting the gait pattern. This article describes one system that is being developed and used in pilot trials on orthopaedic patients and elderly people in the United Kingdom. This work is part funded by the TSB Assisted Living Programme (www.innovateuk.org/ourstrategy/innovationplatforms/assistedliving.ashx).

 

The inertial measuring units (IMUs) record six-degrees-of-freedom (Figure 1); these were developed under a European Union Framework 6 Healthy Aims project by ETB. Each IMU weighs 54 g, measures 73 × 36 × 19 mm and contains a tri-axial 5 g accelerometer and three single axis, 1200 degree/s gyroscopes. There are also anti-aliasing filters with a cut-off frequency of approximately 50 Hz; the filter outputs are sampled with a 12 bit analogue-to-digital converter at a frequency of 102.4 Hz. Before each trial, each IMU is factory set to within 1 ppm (equivalent to 3.6 ms/h) of a reference time that is traceable to national standards, with the aim of achieving less than 10 ms per hour relative drift between units after synchronisation. The IMU sample interval is 9.77 ms, which is greater than the relative drift between two units in one hour. Each IMU is time synchronised at the start of each trial by a simultaneous pulse sent to the respective units.

 

The sensors produce data on angular rotation from the gyroscopes and angular and linear acceleration from the accelerometers. To produce an output that is meaningful to the clinician, physiotherapist and patient, the sensor data must be processed. For the gait analysis application the following parameters are calculated:

The stride duration and the temporal phasing are calculated for each stride and the knee flexion angles are calculated every 10 ms for the entire duration of the trial. The angles, temporal phasing and stride duration are plotted for the entire trial. The user can interrogate any group of strides of interest; for example, the steady section in level walking, stair climbing and stair descent. The typical stride in the chosen region is plotted, together with the standard deviation. In addition, a summary statistic table of the region and a histogram of the knee flexion angles is supplied.

 

In a clinical environment, the procedure should require minimal set up time; the sensors attached to the legs should not affect the person’s movement and must not move during the test. The sensors are strapped to upper and lower legs, one by one, and switched on in the process. The person is then asked to stand still for 5 s to calibrate the sensors and then the trial begins. Each application can define the person’s protocol and this can last for 5–15 min depending on how many different environments the person is to be tested under; for example, it can be level walking or it can include stair climbing and descent. In an extensive test, slopes and different surfaces can be included. Stopping for approximately 10 s between activities enables each section to be clearly identified.

 

An example is provided in which a person is walking on level ground and the clinician analyses a region. The statistics for the region are automatically calculated and produced as a table. The average stride duration is 1.23 s, which is slow for a normal person, with a 5% variation. The relative phasing between the left and right leg is 49%, only 1% lower than the 50% expected for a symmetric gait. Figure 2 shows the typical stride from the region chosen, plus the standard deviation. From this, it is clear that the right leg does not bend as much as the left and it never fully straightens. There is also no extension-flexion when on load on the right leg, whereas with the left leg on load there is no extension on lift off. The conclusion is that the flexion of the right knee is not comparable to the left knee. This may put undue loading onto the knee joint and also an asymmetric loading on the back.

 

So far, the system has been used extensively at the London Knee Clinic, UK, to monitor patients going through the rehabilitation process after surgery. The results demonstrate that the symmetry of gait improves throughout the rehabilitation process,
particularly in terms of the symmetry of the knee flexion angle, comparing the left and right knee. The Clinic is now extending its trials to quantify the condition before treatment, so that both patient and clinician can measure the improvement that the treatment has made.

 

Further studies are planned including monitoring the joint pattern in individuals having total knee replacements compared with those who have partial knee replacements. Establishing the evolutionary changes that occur in recovery after anterior cruciate ligament reconstruction is also under consideration.

 

Enough experience has now been achieved using the device during normal walking and during stair climbing to establish that the device can be used during running indoors on a treadmill and on the sports field. This is an exciting extension of the monitoring capability of gait analysis and will enable athletes who are returning to their sport following a knee injury to be monitored.

A further application being explored is the use of the system for identifying older people who are at risk of a fall. The main aim of this work, which is being conducted by the Bath Institute of Medical Engineering (BIME), is to understand the constraints that surround the use of the system in a clinical setting such as in a doctor’s surgery; these constraints include limited measurement time and limited physical space. The overall approach for this application is to use the system primarily to allow calculation of the variability of stride time, which has been shown in the literature to be a good indicator of falls risk. This could lead to earlier referral of at-risk patients to a specialist falls clinic prior to a fall, thus reducing the significant personal and financial cost associated with falls. A pilot study of healthy volunteers over 60 years old has indicated an appropriate walking distance for the next trial, which will compare the gait of healthy volunteers with “at-risk” patients in a clinical environment.

Other applications include medical conditions such as back pain, Parkinson’s disease, osteoporosis and diabetes, which all affect the gait profile. Regular monitoring could identify changes and trigger new treatment or exercises where appropriate.

 

The system can be used on people of all ages and with abnormal gaits caused by illness, disability or age. The initial work done by the London Knee Clinic and BIME will be used to develop custom software for different applications.