Automatic Posture Recognition System

Insole Gait Analysis System

Gait abnormalities are common in clinical practice and there is a global imperative to improve technologies that facilitate their detection, evaluation, monitoring and management. Real time evaluation using digital technology supports the development of digital healthcare.  Currently gait assessment relies on visual observation of structured clinical tests such as the “Timed Get up and Go Test”. Gold standard methods such as “Qualisys Motion Capture System” require sophisticated equipment in gait laboratories. These are not widely available due to expense, analysis time and requirement of trained technicians. Developing low cost, portable, easy to use digital technology is important to enable sophisticated assessment of gait at home or in clinics. To facilitate the evaluation and interpretation of locomotive information, a tool to visualize gait in real-time is proposed. The proposed tool consists of five approaches (1: Realtime dial visualization, 2: Visualization of individual leg time variation, 3: Visualization of both legs asymmetry, 4: Boxplot visualization, and 5: Evaluation considering all features).  Results show that wearable Inertial Measurement Unit (IMU) can be used for extraction of objective gait features. This system opens possibilities for home-based assessment of gait without the requirement and expense of an elaborate laboratory setup and supports the development of digital healthcare.

1)  Real time dial visualization

Figure 1 demonstrates spatiotemporal measurements in a dial fashion taken from one subject. Both legs should theoretically give identical results and therefore perfect asymmetry should give dial indicator readings of zero. The first dial is an asymmetry display for stride length and time comparing both legs. The second dial displays the real time measurement of step length and time. It is noted that there is a difference in the level of asymmetry. The third dial shows the swing phase distance and time. Similarly, there is little asymmetry between two legs. The scales on all three dials represent the CIs and the pointer represents the instantaneous real time difference between two legs. For example, in this case although the dials for stride and swing show near perfect symmetry, measurements relating to step are not. Step measurement entails information on distance and time. The distance dial shows that right leg is travelling longer (0.51m) than left leg (0.39m). The patient therefore needs to shorten the distance travelled by the right leg and/or make the left step longer.  Time dial demonstrates the right leg travels the longer distance in a shorter time (0.26s) compared to the left leg (0.34s). Digital number below the dial is showing the absolute measure for all three markers.

2)  Visualization of individual leg time variation

30 strides are performed and time of stride, stance and swing phases is presented in Figure 2 where each bar shows the stride time. Cyan and yellow colors represent the time of stance and swing phases respectively. There is a small variation of stance and swing phase timing. This visualization clearly represents the variability of stance and swing phases in each stride of the legs. The ratio of stance and swing is found closest to the 60:40% split for average stride, stance and swing information (Figure 2).

Figure 2. Time of stride, stance and swing phases from right and left legs

3)  Visualization of both legs asymmetry

In this visualization the stride and step asymmetry information for both time and distance from both legs are presented in Figure 3. We observe that while there is good symmetry in the stride there is strong variation in the step phases.

Figure 3. Gait asymmetry of stride and step phases from right and left legs

4) Boxplot-based visualization

Figure 4 shows a boxplot of the distribution of values for individual factors where the mean values obtained for stride and step for both legs. The quartile ranges are identified in the boxplot and show low variation for the stride. First box plot shows higher variation in the step length on the left leg than on the right. This demonstrates that although the stride length is similar on the right and left there can be a higher variation in the step length. Boxplots for time indicate that variation is low for both legs. If the first and last stride of each walking are excluded on the corridor, the asymmetry is not that high. Those phases consist of more variation due to initial acceleration and ending momentum. It is noted that the observations identified by the boxplots are not especially extreme.

Figure 4. Boxplot of stride, stance and swing information

5)Evaluation considering all features

Gait features are used to develop a shape of the individual gait which represents the gait of an individual person. In a normal situation of near symmetry in gait the shape should be at a 45o angle as the extracted features on the right is roughly the same as that on the left. Procrustes analysis is used to obtain a best fit amongst all shapes of gait features extracted from all 10 young subjects by performing translation and rotation. The mean of all best fit shapes is considered as NMGS.

Figure 5: Lowest and highest shape differences from (a) young and (b) older subjects

In Figure 5(a), the highest variation is in Young 8 and the lowest is in Young 1 and there is a clear difference in the gait shapes between the two subjects. Similarly, the highest variation is in Older 13 and the lowest is in Older 8 shown in Figure 5(b). This different shape for each individual arises from the differences in the values of the extracted features. For example, in Young 8 the highest variation is due to travel distance and time features. Similar findings are also found for older subjects. These shape difference have clinical relevance for example in Older 13 had a stroke with a numb right leg. He used crutches when moving and was unable to move the right leg. Thus most of the movement during walking was covered by the left leg and crutches are used to keep body balance. While the above captures the abnormality it is important that the abnormality is quantified. This is important to monitor progression of an illness as well as to monitor improvement with treatment. For this purpose, EDMA is used to locate the gait feature contributing to the abnormality. A mean form is estimated from 10 young subjects. This is used to estimate the inter-feature distances that represents the distance between each feature from one to another (29). The form difference matrix is then estimated between the mean form and all gaits forms. Figure 6(a), Older 13 has the highest value of 19.82 indicative that this patient has the highest variance.

Figure 6: Degree of abnormality from (a) young and (b) older adults

I am interested to investigate the following

  • Early detection of diabetic ulceration and evaluation of treatment outcome
  • Early detection of Parkinson Disease and evaluation of treatment outcome
  • Detection of different stages of Alzheimer disease and effective treatment plan
  • Quantitative analysis of gait pattern before and after knee surgery
  • Remote monitoring of gait for older adults
  • Assessment of mental wellbeing based on movement analysis Remote monitoring of exercise and prevention of injury