Page 32 - JSOM Fall 2018
P. 32

FIGURE 4  Placement of accelerometer.              FIGURE 5  Acceleration data points in x, y, and z axes for trials
                                                             1 and 2.
























                                                             0.001 seconds, could lead to moderate injury for healthy sub­
                                                             jects.  The disagreement between the extreme Sked­mounted
                                                                 5
                                                             acceleration magnitudes and the lack of injury of the human
          Data Preparation and Statistical Analysis          patient suggests inaccuracy with that particular sensor place­
          Raw data were converted by code provided by the accelerome­  ment. Due to this suspected inferiority of data from the lit­
          ter vendor using MATLAB software (The Mathworks, https://  ter­mounted trial, we limited our analysis to the data collected
          www.mathworks.com). We performed additional calculations   while the accelerometer was mounted on the mannequin.
          for statistical analysis and data representation. We used the
          generalized extreme value (GEV) distribution using maximum   Acceleration of Litter Patient
          likelihood estimation in MATLAB to model the maximum   During Military Dismounted Movement
          values for the mannequin­mounted trial. The 75­minute, man­  Figure 6a shows the triaxial acceleration profile in the man­
          nequin­mounted trial was split into 75 individual, 1­minute   nequin­mounted trial. Due to the amount of data contained
          intervals to standardize the time interval and create a sample   in Figure 6a and unavoidable resolution issues due to space
          population for statistical analysis. Identifying the maximum   constraints, Figure 6b presents the measured accelerations
          acceleration  value for  each 1­minute interval created a dis­  from 2000 seconds to 2500 seconds in an effort to represent
          tribution of maximum values, which was then modeled with   the data with greater clarity. There are a substantial number
          GEV parameter analysis. The confidence interval for the 90th   of baseline readings resulting from rest or movement with
          percentile of the maximum acceleration value distribution   near­constant velocity, sparsely interspersed with significant
          was determined using log­likelihood calculations of the GEV   accelerations. The volume of baseline readings of the litter at
          parameters.                                        rest makes commonly used mean and median analyses of this
                                                             type of data less valuable.
          Results
                                                             Dividing the mannequin­mounted trial into 75 1­minute in­
          Evaluation of Accelerometer Placement              tervals enabled an analysis of maximum values that avoids in­
          To devise a convenient, executable protocol for evaluation of   terpreting data when the litter is at rest. In addition, division
          dismounted movement exercises, we sought to determine the   of the data into 1­minute intervals allows comparison with
          validity of affixing the accelerometer directly on the Sked ver­  results of other studies by overcoming the variability in differ­
          sus directly on the patient, as depicted in Figure 1. Although   ent protocols; this is a necessary treatment given the constraint
          the literature suggests application of accelerometers to the lit­  of nonintervention with the Rough Terrain Evacuation Course
          ter, these studies  used rigid, framed stretchers,  whereas  this   curriculum. The mean maximum acceleration for the manne­
          study used  the  semiflexible  plastic Sked.  Because  of  the   quin­mounted trial after dividing it into 75 1­minute intervals
                                           2,3
          potential for instantaneous deflections in the flexible plastic   was 1.9g. The GEV distribution, using maximum likelihood
          upon point impacts, we anticipated anomalous readings at    estimation, creates a model that represents the distribution
          the accelerometer that would not represent the forces felt by   of maximum values for the 1­minute intervals. GEV model­
          the litter patient.                                ing resulted in the following parameters: 1.4700 for location
                                                             (μ), 0.3736 for scale (σ), and 0.4133 for shape (κ). The GEV
          More than 20 million acceleration data points were collected   model produces a 90th percentile maximum acceleration value
          for each trial across the combined x, y, and z axes. These data   of 2.9g, signifying that given an infinite number of 1­minute
          points are depicted in Figure 5 as a comparison between tri­  intervals similar to those gathered in the mannequin­mounted
          als. The data from the mannequin­mounted trial demonstrate a   trial, only one  in 10 of their  maximums would breach  this
          tighter spread of points compared with the Sked­mounted trial.   value. The 95% confidence interval for the 90th percentile
          The maximum recorded accelerations for trials 1 and 2 were   of the maximum acceleration value, based on log­likelihood
          178.3g and 5.5g, respectively. Importantly, it has been reported   estimation of the GEV parameters (μ, σ, κ), is between 2.4g
          that forces greater than 100g, even at durations as short as   and 3.7g.


          30  |  JSOM   Volume 18, Edition 3 / Fall 2018
   27   28   29   30   31   32   33   34   35   36   37