Quantification of equine sacral and iliac motion during gait: A comparison between motion capture with skin-mounted and bone-fixated sensors


email: lesley@animalphysio.com.au


Reasons for performing study: Information regarding movement at the ilium and sacrum in nonlame horses during normal gait may assist in understanding the biomechanics of the equine sacroiliac joint.

Objectives: To determine the amount and direction of motion at the ilium and sacrum using 3D orientation sensors during walk and trot in sound Thoroughbreds. To compare results from sensors fixed to the skin with results from sensors fixed to bone-implanted pins.

Methods: Three 3D wireless orientation sensors were mounted to the skin over the tuber sacrale (TS) and sacrum of 6 horses and motion at the ilium and sacrum was recorded for lateral bending (LB) flexion-extension (F-E) and axial rotation (AR) during walk and trot. This process was repeated with the orientation sensors mounted to the same pelvic landmarks via Steinmann pins.

Results: Mean walk values were greater than trot values using pin-mounted sensors for all planes of movement (P<0.05). Walk had 1.64 ± 0.22° (mean ± s.e.) more LB than trot (pin-mounted) yet 0.68 ± 0.22° less than trot when skin-mounted; 3.45 ± 0.15° more F-E (pin- and skin-mounted), and 4.99 ± 0.4° more AR (pin-mounted), but trot had 3.4 ± 0.40° more AR than walk with skin mounting. Using pinned sensors for trot resulted in less LB (2.47 ± 0.22°), F-E (1.12 ± 0.15°) and AR (10.62 ± 0.40°); and for walk less F-E (1.12 ± 0.15°) and AR (2.15 ± 0.40°) compared to skin-mounted. Poor correlation existed between mean values for skin- and pin-mounted data for walk and trot, for all planes of motion.

Conclusions: Movements were smaller at trot with bone-fixated sensors compared to walk, suggesting increased muscular control of movement at the trot. The apparent increase in skin motion at the trot and no clear correlation between skin- and bone-mounted sensors indicates inaccuracies when measuring sacral and iliac movement with skin mounting.


In contrast to the thoracolumbar region little is known about the kinematics of the equine sacroiliac joint (SIJ), despite it having an important role in locomotion as the propulsive forces generated by the hind limbs during gait are transferred via the pelvis to the vertebral column and trunk (Jeffcott et al. 1985). Studies of kinematics during gait have focused on the thoracolumbar vertebral column, using lightweight markers mounted on the skin over the spinous processes and a multi-camera computerised system (Audigie et al. 1999; Licka et al. 2001; Gómez-Álvarez et al. 2006).

A common problem with use of any skin-mounted sensor or marker for kinematic studies is movement of the soft tissue covering the bones (Forner-Cordero et al. 2008). Faber et al. (2000, 2001a,b) and Haussler et al. (2001) used bone-implanted markers to measure 3D kinematics of the equine thoracolumbar vertebral column during walk, trot and canter. Faber et al. (2001a) suggested that due to soft tissue movement, errors may be introduced when using skin-mounted markers to describe kinematics of the equine vertebral column. Inertial or orientation sensors had previously been mounted to the skin over both tuber coxae, for discrimination between lame and nonlame horses (Pfau et al. 2007); however, that study did not measure kinematics.

The aim of this study was to utilise bone-fixated orientation sensors to measure the direction (absolute orientation of segment) of motion at the sacral segment (S3) and the ilium (tuber sacrale[TS]) during the walk and trot, in horses that did not have hindlimb lameness or sacroiliac disease (SID) and to compare the motion recorded between bone and skin-fixated sensors.

Materials and methods

Ethical approval for animal use was obtained from the University of Queensland, Australia, animal ethics committee.


Six Thoroughbred horses were used in this study. There were 2 geldings and 4 mares, mean age 7.6 years (range 4–14 years); mean weight 519.6 kg (range 480–553 kg), mean height 159 cm (range 154.4–162.5 cm). The history of the horses was unavailable, as horses were acquired from a sale yard.

The 6 Thoroughbred horses were each acclimatised to locomotion on a high-speed equine treadmill (Mustang 2000)1 at inclination 5%, on 2 occasions over 2 days prior to commencement of data collection. Prior to entering the treadmill, horses were placed in stocks and the treadmill safety harness was applied. Walking (1.5 m/s), trotting (3.5 m/s) and cantering (5.5–8.0 m/s) were successfully performed on the treadmill by all horses prior to data collection. The horses were assessed by a veterinarian (C.M.) and a physiotherapist (L.G.) for the presence of any hindlimb lameness. None of the horses were lame. Horse 4 displayed minimal asymmetry of the TS and minimal muscle atrophy of the left pelvis musculature (middle gluteal), but there was no alteration of gait, or any other signs of sacroiliac dysfunction. Horse 2 was given 2 g of phenylbutazone (PBZ) paste following initial treadmill acclimatisation due to the presence of a subcutaneous swelling in the right forelimb, not causing lameness. The swelling subsequently reduced over 48 h.


Segment angles of the sacral segment (S3) and the ilium (TS) were recorded using 3 wireless orientation sensors (Inertia Cube 3, IC3)2. The wireless range of the sensor was approximately 10 m and the weight of each sensor 20.0 g. The IC3 sensors measured absolute orientation of any object relative to gravity and magnetic north. The collection frequency for the sensors was 100 Hz. Previous work has shown that the sensors have a static accuracy of better than 0.05° when appropriately configured (Foxlin et al. 1998).

The IC3 sensors contain an accelerometer, magnetometer and gyroscope in each orthogonal plane. The orthogonal planes referred to are those denoted by the standard right-handed orthogonal Cartesian coordinate system. Flexion-extension (F-E) is described as rotation around the x-axis; lateral bending (LB) is described as rotation around the z-axis; axial rotation (AR) is described as rotation around the y-axis. Orientation in this study was reported as Euler angles. All data were collected and analysed using Labview 7.13.


Skin-mounted markers: Xylazine (150 mg) was administered i.v. to each horse prior to the horse being clipped over the region of both TS, sacral spinous processes (SPs) and caudal lumbar SPs. Fixomull stretch tape was applied over the bony prominences of each TS and the S3 SP of the square standing horse, and an ink marker denoted the mid-point of each bony prominence. IC3 sensors were placed over the ink mark on the bony prominences, with the middle of the sensor lining up with the ink mark, fastened with double sided tape and further fastened down with Fixomull stretch tape.

Sensor 1 was attached over the left TS, sensor 2 was attached over the right TS and sensor 3 was attached over the S3 SP.

There were no observable effects of sedation when the horses entered the treadmill approximately 45 min following sedation. When steady state locomotion was reached, at a speed of 1.5 m/s, 3 sets of walking motion (W1–3) data were recorded. Steady state was defined 3 the horse having a consistent cadence of gait at the given treadmill speed. Three sets of trotting data (T4–6) were then collected at steady state locomotion at a speed of 3.5 m/s. Speed was reduced to 1.5 m/s and a final set of walking data (W10) was collected. The W10 set was to be compared to W1–3, to see if there were differences between ranges of motion for walk following trot. This was to assess if there was any dislodgement of the skin-mounted markers due to change in adhesion following sweating and/or change in skin temperature during trot.

Bone-implanted (pin-mounted) markers: Within 24 h of collection of data using skin-mounted sensors, horses were sedated with xylazine 200 mg and butorphanol 20 mg i.v. in preparation for insertion of Steinmann pins. Prior to pin insertion, gentamicin (6.6 mg/kg bwt) and 2 g PBZ i.v. were administered, the areas of pin insertion clipped and prepared aseptically and a stab incision made through the skin and underlying tissue onto the underlying bone, using a No. 11 scalpel blade. A 4–8 cm long, 3.0 mm thick Steinmann pin was then placed into the spinous processes (last lumbar and S3) and both TS without predrilling and cut such that each pin protruded approximately 1 cm above the skin. Custom built lightweight brackets, weighing 9 g and measuring 34 × 25 × 20 mm (Fig 1) with IC3 sensors screwed to the same, were fixed via 2 tightening nuts to the protruding end of each of the Steinmann pins on the left and right TS and the S3 SP in the same configuration as for the skin mounted situation. There was a fourth sensor pinned into the last lumbar vertebral SP. Orientation of the left and right TS and the S3 SP were simultaneously recorded by the sensors in 3 rotations (F-E, LB and AR) at the walk and trot on the treadmill. Data from sensor 4 were discarded as, during canter, the pin in the SP worked loose. For this reason, canter was not included in the data collection.

Figure 1.

The custom built lightweight bracket, with Steinmann pin clamped.

Horses were introduced to the treadmill approximately 60 min following initial sedation and in all cases there were no observed effects of sedation. Three sets of walking (W1–3), trotting data (T4–6) and a final set of walking data (W10) were recorded, as for the skin-mounted sensors. The W10 set of data was to be compared to W1–3, to see if there were differences between ranges of motion for walk following trot. This was to assess if there was any dislodgement of the Steinmann pins during trot.

Data processing

During data collection using both skin- and pin-mounted sensors, data were sampled at 100–150 samples/s. The spatial orientation of each segment (left and right TS and S3) during gait was represented by Euler angles, as a composition of rotations from a reference frame recorded from the sensors on each bony prominence. The recording of the absolute motion (F-E, LB and AR) of each segment was compared for walk and trot. Then differences in motion between bone-implanted (pin-mounted) and skin-mounted sensors were noted. Statistical analysis was carried out using the GLM procedure in SAS version 8.2. Contrasts were used to determine consistency of movements during gait and if there was consistency between final walk (after trot) and initial walking data. Analysis of variance (ANOVA) was used to compare the effects of gait, mounting and sensor. Values are presented as mean ± s.e. Correlations were performed between skin- and pin-mounted sensors for both walk and trot.


There were significant effects of gait, mounting and sensor. There was no significant variation between the sets of walking data and the sets of trotting data, for both skin- and pin-mounted situations, however, there were significant consistent differences between walk and trot with pin mounting (P<0.01) and significant consistent differences between pinned and skin mountings at the trot (P<0.01). Overall, there was no significant difference between the final walk (W10) and initial set of walk data (W1–3) for the pin-mounted sensors. There were 2 instances of significant differences found between W10 and W1–3 for skin-mounted sensors - this was in the directions of LB for the sensor mounted over the sacrum (P<0.05) and for F-E for the sensor mounted over the right TS (P<0.01).

Mean values for walk (W1–3)

The mean values recorded from skin- and pin-mounted sensors during walk are listed in Table 1, in Euler angles. Figure 2 is an example in graph form that shows the mean values for AR during walk and trot, with sensors in skin- and pin-mounted situations, highlighting the differences between the 2 mounting techniques during the same gait.

Table 1. Mean values (Euler angles) recorded at each sensor, for walk and trot, skin- and pin-mountings
Gait and planeSensorEuler angles (o)Pooled s.e.Significant between mountings
Skin mountPin mount
  1. Sensor 1 (left TS); sensor 2 (right TS); sensor 3 (S3). LB = lateral bend; F-E = flexion-extension. AR = axial rotation.

Figure 2.

Graph showing mean values for AR during walk and trot, with sensors in skin and pin-mounted situations. Asterisk denotes significant difference between the 2 mounting techniques during the same gait (P<0.05). Error bars denote s.e.

There was no significant difference between skin and pin-mounted data for LB at the walk. There was a significantly greater range of F-E for skin-mounted data compared to pin-mounted data for reading from each of the 3 sensors (P<0.02) (Fig 2b). There was no significant difference in the amount of AR measured at skin-mounted sensors vs. pin-mounted sensors over the TS. However, there was a significantly greater range of AR of the sacral segment (sensor 3) for skin-mounted data compared to pin-mounted data (P<0.01).

Mean values for trot (T4–6)

There was a significant difference between skin- and pin-mounted data for all LB and F-E (P<0.01) at the trot, with range of LB for skin-mounted data being greater than pin-mounted data.

There was also a significant difference between skin- and pin-mounted data for all AR (P<0.01) at the trot, with range of AR for skin-mounted data being greater than pin-mounted data. The difference between pin- and skin-mounted sensors was greatest during AR compared to the other orthogonal planes. The mean values for AR for skin-mounted markers during trot were the largest overall.

Effect of gait

Lateral bending was 1.65° ± 0.22 less for trot than during walk with pinned sensors (P<0.01), and trot was 0.68° ± 0.22 greater than walk with skin mounting (P<0.01).

Flexion-extension was 3.46° ± 0.15 less for trot than walk, with pin-mounted sensor (P<0.01) and 3.45° ± 0.15 less for trot than walk with skin-mounted sensor (P<0.01).

Axial rotation was 4.99° ± 0.40 less during trot than during walk with pin-mounted sensors (P<0.01) and 3.48° ± 0.40 greater for trot than walk with skin-mounted sensors (P<0.01)

Effect of mounting

Trot had 2.48° ± 0.22 greater LB with skin mounting vs. pin mounting (P<0.01), but there was no significant difference between mountings for LB during walk.

There was 1.12° ± 0.15 greater F-E during walking with skin mount vs. walking with pin-mounted sensors (P<0.01), and 1.13° ± 0.15 greater F-E during trotting with skin mount vs. pin-mounted sensors (P<0.01). AR was 2.15° ± 0.40 greater during skin-mounted walk compared to pin-mounted walk (P<0.01) and 10.62° ± 0.40 greater during skin-mounted trot than pin-mounted trot (P<0.01).

Effect of sensor

Sensor 3 (S3) showed 0.56° ± 0.17 less movement than Sensor 1 (left TS) during LB (P<0.01). Sensor 1 (left TS) showed 0.31° ± 0.13 greater F-E than Sensor 2 (right TS) (P<0.05), and 0.46° ± 0.13 greater F-E than Sensor 3 (S3) (P<0.01). There were no significant differences in patterns for AR, between sensors.

Correlation of skin-mounted results from pin-mounted results

The average of LB, F-E and AR from skin-mounted sensors and the pin-mounted sensors were compared for each horse, for walk and trot. There was poor correlation between skin and pin-mounted data for walk and trot, for all planes of motion (Figs 3a–c). Correlation coefficients ranged from −0.011−0.328.

Figure 3a.

Scatter plot for LB for walk and trot. Walk = open circle; trot closed circle.

Figure 3b.

Scatter plot for F-E for walk and trot. Walk = open circle; trot closed circle.

Figure 3c.

Scatter plot for AR for walk and trot. Walk = open circle; trot closed circle.


The results of this study have described the absolute motion of the sacrum and ilium in 3 orthogonal planes at the walk and trot. Motion was significantly reduced in the trot for bone-fixated sensors although this became less apparent when using the more variable and poorly correlated skin-fixated sensor values. There was good repeatability between the 3 sets of consecutive walk and trot data, for both pinned and skin-mounted sensors, averaged over the 6 horses. The stability of the pin mounted sensors during data collection was confirmed by the lack of difference between the final group of walk data that followed the trot and the 3 initial walking sets of data. Slight difference existed between the final groups of walk data when skin-mounted sensors were used. The differences between final and initial walk values may have been due to movement of skin over the bony landmarks, leading to dislodgement of adhesive. The higher values obtained during the trot for skin-mounted vs. pin-mounted sensors supports the notion of increased skin movement with skin mounted sensors.

During walk, recordings of LB were not significantly different between skin-mounted and pin-mounted sensors. There may be appreciably little skin movement artefact in the LB plane during walk. For all sensors, there was greater motion recorded in the F-E plane when they were skin-mounted during walk, compared to pin mounting, suggesting that some movement of the skin occurs in the sagittal plane during walk. There was significantly greater AR movement recorded at the sensor mounted to the skin over S3 compared to the sensor pinned to the same landmark, but no difference in AR recorded at the sensors mounted to the skin over the TS compared to the pinned sensors there. Together, these data may indicate movement of the TS and the sacral SP relative to the skin at walk in a sagittal plane and movement of the sacral SP relative to the skin in an AR direction.

During trot there was a significantly greater LB, F-E and AR recorded with the skin-mounted sensors compared to the pin-mounted sensors, with the difference between the 2 mountings being the greatest for AR. This suggests that there is movement of the TS and sacral SP relative to the skin in all directions at the trot, especially in the AR plane. Axial rotation was the largest overall plane of movement recorded at all sensors for walk and during trot with skin-mounted sensors. During pin-mounted trot the component of AR was significantly smaller than during skin-mounted trot, yet still significantly larger than LB and F-E recorded during pin-mounted trot. This again points to the movement of pelvic bony landmarks relative to skin occurring mostly in an axial plane, at both the walk and trot. As the mass of the orientation sensor is 20 g, in a skin-mounting there may also be expected to be a greater effect of inertia on the sensor during trot compared to walk.

The mean values recorded at the sacral segment with pin-mounted sensors were larger than those of Haussler et al. (2001) who recorded segmental motion at the L6 and sacral SPs at the walk and trot, but used a transducer mounted to a Steinmann pin rather than a sensor. Despite the discrepancy in magnitude, the patterns of movement recorded by Haussler et al. (2001) were similar to in this study. This was reflected in F-E being the smallest plane of motion at the sacral segment during trot, and AR being the largest plane of movement measured from the sacral segment; patterns similar to the current study.

Error associated with using skin-mounted markers in equine kinematics has been recognised previously by van den Bogert et al. (1990) and also in other quadruped kinematic studies (Taylor et al. 2005). Forner-Cordero et al. (2008) stated that a common problem shared by accelerometers, inertial sensors and any measurement system based on a skin-mounted sensor is the signal distortion caused by vibration of the skin and soft tissue. They also recognised that there is preload compression of the soft tissue underneath the sensor due to the adhesive tape used to fix the sensor to the skin and this can increase the stiffness of the sensor attachment. In this current study, the Fixomull stretch tape passed over the sensor at each site of attachment and was firmly fastened to the skin, so these factors need to be considered when interpreting results. A conclusion from a study that compared sensors mounted to the skin and directly to the bone was that sensors must ideally be 3 g or less in mass, to reduce the error associated with skin-mounted sensors (Trujillo and Busby 1990). The sensors used in the current study had a mass of 20 g each, for which reason they represent a potential source of error. Forner-Cordero et al. (2008) suggest the only way to overcome error due to soft tissue is to attach the sensor to the bone, as we have done in this study, by way of Steinmann pins.

Taylor et al. (2005) used optimal common shape technique to account for the effect of local deformations due to skin elasticity on marker sets mounted to the legs of sheep during gait. Error in prediction of position of skin markers from markers mounted to screws inserted into the femur, tibia and metatarsus of the sheep was strongly associated with the amount of soft tissue covering the region measured. These authors concluded that it is difficult to gain an accurate representation of kinematics of underlying bone from skin markers due to muscle firing and also inertial effects of heel impact. However, whether these effects apply to sensors or markers mounted to the axial skeleton, as in the current study and previous equine kinematic data, is unknown.

In this study, there was a poor correlation between the averaged skin and pin-mounted data, for all planes of motion, for both walk and trot. Although Faber et al. (2001c) were able to evaluate kinematics of the SIJ using lightweight skin-mounted markers with video analysis in a comparative way, from the results of this current study we cannot predict motion that occurs within a given bony segment of the SIJ region from skin-mounted sensors during gait. The markers used in Faber et al. (2001a,b,c) were lightweight, whereas the orientation sensors had a mass of 20 g. Until advances in technology produce orientation sensors with a mass similar to the lightweight reflective marker, the capabilities of the orientation sensor to record motion during locomotion whilst mounted to the skin, are limited.

There may be a source of error in the use of bone-implanted sensors due to the distances from the S3 spinous process to the sacrum and the distance from the TS to the articular surface on the wing of the ilium. In addition, the lightweight aluminium brackets were a further 10 mm away along the Steinmann pin, from the surface of the bony landmark. Therefore, the measurements taken at the TS and S3, even with bone-implanted sensors, could only be considered to be the orientation of that bony landmark at a given point in time and not a reflection of the movement occurring at the SIJ.

Another source of error may lie with the assumption that motion segments of the vertebral column behave like rigid bodies (Faber et al. 2001b). Haussler et al. (2009) concluded from an in vitro study that the bones of the pelvis are not rigid structures and bony pelvic deformation is a normal occurrence in any SIJ movement. In vivo, the degree of bony deformation may depend on the age of the horse. Haussler et al. (1997) had noted previously that equine pelvic physeal closure occurred at age 5.2−5.8 years of age, with iliac crest and ischial arch epiphyseal formation and fusion to the underlying bone occurring at 7.2 years and 5.4 years, respectively. However, only one of the 6 horses used in the current study was aged <5 years, which may have minimised the overall effect of intrinsic pelvic movement associated with incomplete pelvic physeal closure.

In this study there appeared to be an effect of sensor for the orthogonal planes of LB and F-E. The sensor attached to the sacral segment measured less motion in the orthogonal planes of LB for both mountings and both gaits, than the sensor on the left TS, but not the sensor on the right TS. The sensor on the left TS had greater amount of F-E motion recorded during both mountings and both gaits than the other sensors. The significance of this finding is unclear, but could possibly relate to the direction of exercise the horses had previously been exposed to when in race training, as the sensor on the left TS had greater motion occurring in a number of situations.

Therefore, in this study, some of the problems encountered initially with the earlier in vivo studies in thoracolumbar kinematics during gait were apparent when movements at the TS and sacrum were measured. Anatomically, the location of the SIJ is more inaccessible than other regions of the vertebral column, with its joint centre of rotation quite a distance to surface landmarks and the landmarks covered by mobile skin and soft tissue. Skin-mounted sensors, being noninvasive, are perhaps more acceptable and readily utilised for measuring motion during gait in vivo than bone-fixated sensors, but are more susceptible to associated errors. It is generally accepted that error must be accounted for when using such a technique for kinematic studies and this may be more important when using inertial sensors, due to the intrinsic weight and increased sensitivity to movement. Skin-mounted sensors may be able to be utilised in a relative way when comparing gait within groups of horses, such as carried out by Pfau et al. (2007), but error from skin fixation may be unacceptable when absolute motion patterns are to be established.

In using the more accurate values from the bone fixated sensors, there was less motion at the ilium and sacrum recorded during the trot in all orthogonal planes. It has been shown in man that a relative stiffening of the SIJ occurs due to facilitation of muscle groups (Richardson et al. 2002). It could be extrapolated from this that the decreased movement of the ilium and sacrum during trot could be related to greater muscular control during the trot in the horse. Robert et al. (2001) suggested that equine trunk muscles contribute to the decreased range of back flexion and extension in trot, as they had recorded increases in EMG activity of these muscles, associated with increasing speed of trot on a treadmill.

In summary, the orientation sensors provided a measurement of the orientation of the given bony landmark in space at an appointed time and at present this gives us an insight into the direction and amount of movement that occurs at segments of the vertebral column and pelvis during 2 different types of gait.


Motion sensors recorded movement in 3 orthogonal planes from sensors attached to the tuber sacrale and sacrum in 6 Thoroughbred horses. There were significant effects of gait, mounting and sensor location. Movements were smaller at the trot with bone-fixated sensors compared to walk, suggesting increased muscular control of movement at the trot. There were significant differences between the degree of motion of the equine ilium and the sacrum measured with skin-mounted orientation sensors vs. bone-implanted (pin-mounted) orientation sensors during walk and trot and there was poor correlation between skin- and pin-mounted data for walk and trot for all planes of motion. Use of skin-mounted sensors was associated with unacceptable error in the current study.

Conflicts of interest

The authors declare no potential conflicts.

Manufacturers' addresses

1 Kagra AG, Berne, Switzerland.

2 Intersense, Bedford, Massachusetts, USA (http://www.intersense.com/InertiaCube_Sensors.aspx).

3 National Instruments, Austin, Texas, USA.