Patients with suspected spinal cord injuries undergo numerous transfers throughout treatment and care. Effective c-spine stabilization is crucial to minimize the impacts of the suspected injury. Healthcare professionals are trained to perform those transfers using simulation; however, the feedback on the manoeuvre is subjective. This paper proposes a quantitative approach to measure the efficacy of the c-spine stabilization and provide objective feedback during training.
The majority of spinal cord injuries (SCI) occur at the cervical level (c-spine) and are caused by traumatic events (traffic accident, sports, and falls) [
Lately, wearable inertial systems such as attitude and heading reference systems (AHRS) have emerged as an alternative for motion capture systems. Inertial systems are a portable, flexible, and relatively low-cost alternative that is not affected by visual occlusions, offering new possibilities for in-context measurement of kinematic features [
The proposed method of measurement is based on wearable inertial sensors, also called attitude and heading reference systems (AHRS), placed on the forehead and the trunk of a simulated patient (SP). SP here refers to an uninjured individual playing the role of a patient with suspected c-spine injury. AHRS are a specific type of inertial measurement unit comprised of 3-axis accelerometers (measuring linear acceleration), gyroscopes (measuring angular velocity), magnetometers (measuring magnetic field), and a fusion algorithm (estimating the orientation of the platform in a fixed and global reference frame) (Figure
Methods of measurement. (a) Attitude and heading reference systems (AHRS) include a 3-axis accelerometer, gyroscope, and magnetometer to measure, respectively, linear acceleration, angular velocity, and magnetic field. Data are passed on to a proprietary fusion algorithm (included in the measurement system) to estimate the orientation of the platform in a fixed and global reference frame-based gravity and magnetic north. (b) Attached on the forehead and the trunk of a simulated patient, AHRS estimates the orientation of both segments in the same global reference frame. (c) Relative orientation of the head to the trunk can therefore be computed directly from the measurement system. (d) Relative orientation is also decomposed into anatomic motion using a dynamic anatomical alignment process.
AHRS modules on the forehead and the trunk of a simulated patient therefore enable orientation tracking of both the head and the trunk in the same global reference frame (Figure
The selected AHRS modules for this specific study are the MTx from Xsens Technologies (Netherlands) [
The chosen simulation scenario corresponds to the recommended transfer technique performed in prehospital settings in the province of Québec, Canada, and involves transferring the patient onto a vacuum mattress using a log-roll [
Transfer scenario. (a) The lead rescuer immobilizes the head (initial phase). (b) On signal, rescuers roll the patient on his side (roll phase) and maintain the patient in this position while the assistant pulls the vacuum mattress close to the patient (maintain phase) and slowly rolls the patient back onto the mattress (push phase). (c) The final positioning of the patient into the middle of the mattress is performed by pulling gently on the sheet placed onto the mattress.
Throughout the continuum of care, the current clinical guidelines recommend to stabilize the head-to-trunk alignment in order to minimize the possible consequences of a c-spine injury. As such, the main efficacy outcome proposed for performance assessment of c-spine transfer is the peak relative motion, derived from the global change in orientation of the head relative to the trunk (see Figure
Characterization of the quality of the manoeuvre is based on the analysis of the temporal and spatial gap between the actual manoeuvre and the ideal representation of the motion during a log-roll. In a perfect log-roll, both head and trunk segments move at the same time (i.e., temporal synchronicity), following concordant paths or “en-bloc” (i.e., spatial synchronicity). The global change in orientation measured for each segment should therefore be the same throughout the manoeuvre. This concept of an ideal log-roll can be modelled using a straight line along the line of identity (slope = 1) on a 2D graph illustrating the change in global orientation measured at the trunk versus the one measured at the head (Figure
2D graphical representation of a log-roll. Motion of the trunk compared to that of the head during a simulated log-roll for (a) close-to-perfect conditions and (b) head drop during the roll and readjustment at the end of the push.
Motion of the trunk versus the head during the ideal log-roll
Motion of the trunk versus the head during the log-roll with head drop
Temporal synchronization refers to the ability of the rescuers to move the trunk and the head at the same time. Poor communication between the rescuers or difficulty initiating and maintaining smooth motion of the trunk during the roll and push phases (e.g., due to a lack of strength with a large SP) may cause a delay between the motion of the head and the motion of the trunk. Such delay will therefore be investigated. Technically, delays are calculated using both orientation signals (head and trunk) and filtered out using a low-pass 3rd-order Butterworth filter set at a cut-off frequency of 2 Hz. The filter’s cut-off frequency was determined through a residual analysis process, using an acceptable residual accuracy threshold. The residual accuracy threshold refers to the ability of the system to detect movement initiation. It is therefore set according to both the measurement system orientation stability and the initial movement stability (i.e., the stability of the head during initial stabilization). In this specific case, a threshold of 0.1° was determined to be adequate from visual observation of a subset of the available trials. Segments are then considered to be in motion as long as their orientation remains above the predefined threshold. Time points at which the roll and push motions are initiated and terminated are then identified for each segment, and the delay values are derived from it.
Spatial synchronicity refers to the idea that both segments move along a proportional arc of circle during both the roll and the push phases, represented by the desired line of identity on the 2D motion graph. Potential spatial quality indicators were therefore developed based on the assumptions regarding ideal log-roll and its representation using a 2D graph of the motion (Figure
Overall, a performance measure and seven quality indicators of the log-roll transfer are proposed and summarized in Table
Performance and quality indicators for the log-roll.
Category | Indicator | Equation | Description |
---|---|---|---|
Performance measure | ROMrelpeak | Max(ROMrel) | Peak change in global orientation of the head relative to the trunk |
Temporal quality indicators | Delayroll_ini | Delay at roll initiation | |
Delayroll_end | Delay at roll termination | ||
Delaypush_ini | Delay at push initiation | ||
Delaypush_end | Delay at push termination | ||
Spatial quality indicators | SlopeRoll | Difference between the slope of the best-fit line of the roll curve and the ideal line of identify | |
SlopePush | Difference between the slope of the best-fit line of the push curve and the ideal line of identity | ||
ABCRoll-Push | Area contained between the curves from the roll and the push phases |
In a recent study [
Assessment of the usefulness of the proposed performance measures and quality indicators was performed using all recorded trials from this dataset, regardless of the transfer or immobilization technique used (i.e., LR22 and LR4 and TS and HS). For each trial, the performance measure and quality indicators listed in Table
A global analysis of all trials performed by the 21 paramedics revealed a mean relative peak motion of 22.0° ± 6.5°. Figures
Graphical representation of a good and a bad log-roll. (a, b) Variation in angular motion of the head relative to the trunk during a good (a) and a bad (b) log-roll. (c, d) 2D motion representation of the same good (c) and bad (d) trial.
Global variation in head orientation relative to the trunk (good log-roll)
Global variation in head orientation relative to the trunk (bad log-roll)
Global change in trunk versus head orientation during a good log-roll
Global change in trunk versus head orientation during a bad log-roll
2D graphical representation of the log-roll motion provides an in-depth visual representation of each segment during the transfer. Figures
Table
Recorded values for all potential quality indicators for the log-roll.
Category | Indicator | Mean (Std Dev) | Range (min, max) |
---|---|---|---|
Performance measure | ROMrelpeak | 22.0° (6.5°) | (9.5°, 40.8°) |
Temporal quality indicators | Delayroll_ini | 0.20 s (0.14 s) | (0.02 s, 0.76 s) |
Delayroll_end | 0.29 s (0.24 s) | (0.00 s, 1.12 s) | |
Delaypush_ini | 0.10 s (0.15 s) | (0.00 s, 0.86 s) | |
Delaypush_end | 0.18 s (0.17 s) | (0.00 s, 0.94 s) | |
Spatial quality indicators | SlopeRoll | 0.12 (0.08) | (0.03, 0.34) |
SlopePush | 0.11 (0.08) | (0.02, 0.36) | |
ABCRoll-Push | 396.6 (219.8) | (47.8, 1104.1) |
Details of the incremental model are shown in Table
Hierarchical multiple regression predicting peak relative motion from SlopePush, ABCRoll-Push, and Delaypush_ini.
Variable | Peak motion | |||
---|---|---|---|---|
Model 1 | Model 2 | |||
Constant | 12.053 | 17.63 | ||
SlopeRoll | 35.797 | 0.426 | 31.77 | 0.378 |
ABCRoll-Push | 0.011 | 0.383 | 0.01 | 0.400 |
DelayPush_ini | 12.388 | 0.286 | 10.18 | 0.235 |
Technique | — | — | −0.96 | −0.076 |
Number of assistants | — | — | −2.43 | −0.191 |
0.515 | 0.551 | |||
27.983 | 18.913 | |||
∆ |
0.074 | 0.036 | ||
∆ |
— | 3.088 |
Notes: 1
This study describes a method based on objective measurement of head motion relative to the trunk using AHRS to provide visual and numerical feedback on the performance and quality of c-spine immobilization during log-roll simulation training. Using data from a sample of paramedics performing this specific technique, we used peak motion to assess the variation in the performance of the manoeuvre. In this specific analysis, the identified averaged peak motion of 22.0° was computed over all trials, regardless of the immobilization technique used or the number of assistants. Furthermore, the performance varied between 9.5° and 40.8° among the trials, demonstrating a clear need for objective training. The choice of pooling all trials together for this analysis was based on the fact that regardless of the technique or the number of assistants used, the desired optimal goal of the manoeuvre remains the same: to limit the motion of the head relative to the trunk during the transfer. Details regarding the performance of each technique and number of assistants may be found in [
The method introduced herein proposes the use of a 2D graphical approach to assess, at a glance, the quality of the manoeuvre performed at each segment level. As such, simulated 2D graphical representation of well-conducted and poorly conducted trials was confronted to actual good and bad trials and was shown to be concordant. Although the usefulness of such graphic was not directly tested with the paramedics, it has face validity in the fact that it directly illustrates where and when excessive motion occurred, with minimal explanation of how an ideal trial would appear.
The 2D graphical representation of motion was also used as the baseline for the definition of the potential quality indicators, in both the temporal and the spatial domains. For the specific context tested, a univariate regression model revealed that Delaypush_ini has a significant relationship to peak motion. In all cases, the greater the delay, the greater the peak motion (i.e., the worse is the performance). Delay at initiation of the motion for the roll phase was not shown to be significant, and therefore the relationship between the communication and timing does not seem to be the major issue. Univariate regression also showed that all the three spatial indicators proposed (SlopeRoll, SlopePush, and ABCRoll-Push) have a moderate linear relationship with the recorded peak motion. The larger the deviation of the estimated slope from the ideal, the worse the performance. Similarly, the smaller the area between the roll and the push curves, the better the performance. This is consistent with the hypothesis that paramedics with sufficient control over the motion will more likely execute an efficient log-roll. Control may be influenced by different factors, including training and force. The limit to this study is that the participants’ force was not measured. It is therefore impossible, at this stage, to evaluate the impact of strength on the results. The multiple regression model identified the roll slope, the spread between the roll and the push curves (ABCRoll-Push), and the delay at push initiation as the main explanatory variables associated with the log-roll performance which explained 52% of the variance. Adding the technique and the number of assistants in the model increased that rate to 55% of the variance explained with the model.
The chosen statistical analysis considered the different parameters as absolute values as the purpose was to look at the influence of synchronicity (temporal and spatial) on the quality of the manoeuvre performed. However, with regard to providing constructive feedback to healthcare professionals, the sign of the delay and its slope difference are also interesting as they reveal which segment moved before or faster than the other. For example, a negative delay at push initiation reveals that the trunk started earlier than the head for that specific phase.
The proposed set of quality indicators is a first attempt at characterizing the quality of a log-roll manoeuvre from a motion coordination-based point of view. The authors feel that this level of feedback can be sufficient in improving the technique during training sessions. However, we recognize that not all specific cases can be assessed using those indicators and that a more in-depth analysis would be required to do so. For example, oscillations during the motion will not be captured directly with the current set of indicators but will appear in the 2D motion graphical representation and could be addressed as such during debriefing. Thus, this study is considered the first step towards the general objective to develop a simulation-based tool to objectively assess c-spine motion.
Furthermore, the techniques used to derive the indicators are straightforward. Although advanced techniques could be investigated to improve, for example, the accuracy of the delay at initiation, it shall be kept in mind that the goal of the present project is to provide a quick and user-friendly approach to assess c-spine motion in training scenarios and provide efficient real-time feedback on the manoeuvre performed in order to improve the technique. As such, the proposed indicators with their current accuracy were able to explain 55% of the variation in performance observed in the study results and are therefore believed to be sufficient to provide an efficient feedback to the trainee.
Indicators were developed using the data from wearable AHRS. Wearable AHRS offer accurate orientation measurements that capture motion of the head relative to the trunk in relatively constraint-free conditions at an affordable price compared to more traditional motion capture technology. Their form factor allows researchers to collect data in varying clinical context scenarios, avoiding problems such as positioning limitations that might occur when an optical marker is blocked from the camera lens view. It shall, however, be noted that the developed indicators appear to be independent of the motion capture measurement system and could therefore be used with other systems such as magnetic motion capture systems [
This study proposes a methodology which provides objective feedback to participants during c-spine transfer simulated scenarios using AHRS. The concept was applied to the specific case of prehospital transfers for which three specific metrics (slope of the motion performed by the trunk compared to the motion performed by the head during roll, delay at push initiation, and area between the roll and the push phases) have been identified as explaining 55% of the performance variation when combined with the technique and number of assistants used. The proposed approach has a potential to be used for personalized feedback during training and could even be directly embedded into simulation mannequins in order to provide a complete training solution.
The authors declare that they have no conflicts of interest.
This study was supported by a grant from the Canadian Institute of Health Research (CIHR). The authors would like to acknowledge the contributions of Ian Shrier to the design and data collection of the study. They would also like to thank Urgence-Santé, L.C. Vacon, and E. Segal for their support in the recruitment of paramedic’s participants and the access to a testing location.