A modified strike index for detection of foot strike pattern in barefoot running

Full text

A MODIFIED STRIKE INDEX FOR DETECTION OF FOOT STRIKE PATTERN IN BAREFOOT
RUNNING
Eveline S. Graf, Michael J. Rainbow, Cynthia D. Samaan, Irene S. Davis
Spaulding National Running Center, Department of Physical Medicine and Rehabilitation,
Harvard Medical School, Cambridge, MA, USA
email: egraf@partners.org, web: http://www.runsnrc.org
INTRODUCTION
The strike index (SI) is used to identify the foot
strike pattern in runners. It describes the location of
the center of pressure (CoP) at initial foot-ground
contact with respect to the long axis of the foot. SI
is traditionally expressed as the percentage of the
total foot length [1]. Based on the SI, runners can be
classified as rearfoot (with a SI between 0 and
33%), midfoot (34 to 67%), or forefoot strikers (68
to 100%). This concept was developed for shod
running [1] but has also been applied to barefoot
running [2].
When forefoot striking, barefoot runners typically
contact the ground with the distal area of the
metatarsals, not the toes; therefore, the underutilized
portion of the forefoot region (from the toes to the
distal metatarsals), may reduce the ability of the SI
to properly categorize forefoot strikes (Figure 1).
Therefore, the purpose of this study was to define a
modified strike index (SImod) that neglects the toes,
and to compare the sensitivity and specificity of SI
and SImod for each strike pattern. It was
hypothesized that SImod would identify a larger
proportion of forefoot strikes than SI.
METHODS
Eight recreational runners (27.8±7.0 years, 1.7±0.1
m, 74.6±10.6 kg) participated in the study and
signed informed consent. Five retro-reflective
markers were attached to the left foot: three on the
heel (used to track the foot position in dynamic
trials), one on the first metatarsal head, and one on
the distal end of the second toe. Each subject
performed seven barefoot running trials at their self-
selected pace. The marker trajectories were
collected with a ten-camera Vicon system (250 Hz)
while ground reaction forces were simultaneously
collected using two AMTI force platforms (1000
Hz) embedded in the floor. Additionally, a high
speed video camera (Basler GigE) recorded the
steps on the force plates from lateral at 125 Hz.
The strike pattern of each trial was determined as
forefoot (FFS), midfoot (MFS), or rearfoot (RFS)
based on visual inspection of the high-speed video
recordings. Custom code (Matlab, MathWorks) was
used to calculate the center of pressure location
relative to the foot at touchdown (vertical ground
reaction force > 20 N), which was then used to
determine the SI and SImod. For the SI, the foot
length was defined by the distal marker on the heel
and the marker on the second toe. The foot length
used to determine SImod was defined as the
distance between the distal marker on the heel and
the marker on the first metatarsal head. The strike
index regions for SI and SImod were defined by
dividing the foot length into equal thirds (Figure 1).
Figure 1: strike index regions for SI (solid) and
SImod (dashed)
The validity of SI and SImod to determine the strike
pattern was determined by calculating the
sensitivity and specificity of both methods for each
of the landing patterns.

RESULTS AND DISCUSSION
Out of 56 steps that were analyzed with visual
classification there were 37 forefoot, 7 midfoot, and
12 rearfoot strikes. The SI correctly identified 6
forefoot, 5 midfoot, and 12 rearfoot strikes. This
resulted in a high sensitivity for rearfoot and
midfoot strikes but low sensitivity for the forefoot
strikes (Table 1). Therefore, the traditional SI fails
at detecting forefoot landings in the barefoot
condition. The SI specificity was high for forefoot
and rearfoot strikes but low for midfoot which is a
result of a large number of false positive ratings for
the MFS (Table 1). This can be directly related to
the way midfoot strikes are defined by the SI. The
SI midfoot contains most of the distal area of the
metatarsals which is the region where ground
contact during forefoot landing occurs.
Using the SImod, 37 forefoot, 1 midfoot, and 12
rearfoot strikes were correctly classified. This
resulted in high sensitivity values for forefoot and
rearfoot strikes but a low value for midfoot landings
(Table 1). The specificity was high for MFS and
RFS and slightly reduced for FFS. Only one step
with an index of over 100% (indicating a toe
landing) was found (101.4%).
SImod had a higher validity for forefoot strikes,
which verified the hypothesis, but a lower validity
for midfoot strikes than SI. A large portion of the
forefoot area of the SI is made up of the toe region,
which is typically not used during forefoot landing.
Therefore, forefoot landings where the distal area of
the metatarsals touches the ground first are often
incorrectly classified as midfoot strikes. This issue
is resolved by using the SImod.
SImod is less sensitive at detecting midfoot strikes
compared to SI. Midfoot landings are defined as
landings where the foot lands flat on the ground.
There can be large variability in the CoP location
depending on the pressure distribution at
touchdown. A midfoot landing where more pressure
is exerted on the forefoot compared to the rearfoot
causes the CoP to shift distally and into the FFS
region of the SImod. This leads to a reduced
sensitivity for MFS and a decreased specificity
value for FFS and RFS.
Another cause for the decreased specificity of FFS
for SImod could be limitations in the strike pattern
detection using visual inspection. MFS were more
difficult to determine. It is possible that during
landings which were classified as midfoot strikes,
the forefoot touched the ground just before the
rearfoot but that this instance was missed due to the
camera angle and the limited frame rates.
CONCLUSIONS
The modified strike index described in this study is
more valid for detection of forefoot landing patterns
during barefoot running compared to the traditional
strike index. Additional research is needed to
examine the validity of the SImod during shod
running. In order to define the strike index areas, the
foot length is divided into equal thirds. Using a
different definition of the strike index regions (e.g.
MFS area larger than FFS and RFS) may increase
the sensitivity of the SImod for MFS. Therefore,
further research with an increased sample size is
currently being conducted to test this hypothesis.
REFERENCES
1.
Cavanagh PR, Lafortune MA. J Biomech13, 397-
406, 1980.
2.
Altman AR, Davis IS. Gait Posture35, 298-300,
2012.
Table 1: Outcomes of diagnostic test (true positive (tp), false positive (fp), false negative (fn), true negative
(tn)), sensitivity and specificity of SI and SImod to determine FFS, MFS, RFS
SI SImod
tp, fp, fn, tn Sensitivity Specificity tp, fp, fn, tn Sensitivity Specificity
FFS 6, 0, 31, 19 0.16 1.00 37, 5, 0, 14 1.00 0.74
MFS 6, 31, 1, 18 0.86 0.37 1, 0, 6, 49 0.14 1.00
RFS 12, 1, 0, 43 1.00 0.98 12, 1, 0, 43 1.00 0.98