Combined CatWalk Index: an improved method to measure mouse motor function using the automated gait analysis system
Crowley et al. BMC Res Notes
Combined CatWalk Index: an improved method to measure mouse motor function using the automated gait analysis system
Samuel T. Crowley 0 1
Kazunori Kataoka 1
Keiji Itaka 0 1
0 Department of Biofunction Research, Institute of Biomaterials and Bioengineering, Tokyo Medical and Dental University (TMDU) , 2-10-3 Kanda-Surugadai, Chiyoda-Ku, Tokyo 113-8510 , Japan
1 Innovation Center of NanoMedicine (iCONM), Kawasaki Institute of Industrial Promotion , 3-25-14 Tonomachi, Kawasaki-Ku, Kawasaki 210-0821 , Japan
Objective: Measuring motor function in mice is important for studying models of spinal cord injury (SCI) or other diseases. Several methods exist based on visual observation of mice moving in an open field. Though these methods require very little equipment, observers must be trained, and the possibility of human error or subjectivity cannot be eliminated. The Noldus CatWalk XT Automated Gait Analysis system assesses mouse motor function by taking high-resolution videos of the mice, with specialized software to measure several aspects of the animal's gait. This instrument reduces the possibility of human error, but it is not always clear what data is important for assessing motor function. This study used data collected during mouse SCI experiments to create a simple mathematical model that combines the data collected by the CatWalk system into a single score, the Combined CatWalk Index or CCI. Results: The CCI system produces similar results to the Basso Mouse Scale or the CatWalk's Step Sequence Regularity Index. However, the CCI has a significantly smaller coefficient of variation than either other method. Additionally, CCI scoring shows slightly better correlation with impact force. The CCI system is likely to be a useful tool for SCI research.
Spinal cord injury; Locomotor function; Mouse; New scoring; CatWalk; Combined Catwalk Index
Spinal cord injury (SCI) research often relies on animal
models. Several methods exist to create animal models
of SCI and to assess motor function recovery after injury.
Open-field locomotion methods are particularly
popular because they require very little equipment. Any flat
surface large enough for the animal to freely walk on is
usually acceptable. The Basso-Beattie-Bresnahan (BBB)
] was developed to measure motor function in
rats. The Basso Mouse Scale (BMS) [
] was developed
later to accommodate for the differences in motor
function recovery between mice and rats. The BMS system
has been further modified and extended to create other
measurement scales, such as the Toyama Mouse Scale
], which is designed to emphasize weight
support and reduce the ambiguity seen in the BMS, which
can assign the same score to mice with different
combinations of stepping frequency, coordination, etc.
When properly performed, these methods are reliable
and repeatable. However, all of these methods require
trained observers. To increase reliability, mice should
be observed by at least 2 different observers, and the
observers must be blinded to the experiment. Not all
laboratories are able to meet these requirements, and the
possibility of human error or subjective measurement
cannot be ruled out.
More objective methods to assess mouse motor
function usually require some form of instrument or other
equipment. The Noldus CatWalk XT Automated Gait
Analysis System [
] uses an instrument with a glass
platform above a high-resolution video camera. Green
light is internally reflected through the glass platform.
As the mouse walks along the platform, light is reflected
down towards the camera wherever the mouse contacts
the glass. The intensity of the reflected light is
proportional to the pressure placed on the glass.
Proprietary software analyzes the videos and produces a large
amount of data to describe several aspects of how the
animal walks, including speed, timing, coordination,
etc. The software produces approximately 25 different
measurements for each paw, plus measurements for the
mouse overall, for a total of approximately 104
parameters. The large amount of data produced by the CatWalk
software is too much to reasonably present everything in
a single publication, so a subset of parameters must be
chosen. This choice can be arbitrary, and it is difficult to
know if the chosen parameters are relevant or adequate
for a given experiment. Researchers may be tempted to
present any parameter that shows a desired statistically
significant result, even if that parameter is not actually
relevant to the study.
This study was carried out to produce a method that
combines all CatWalk parameters into a single score, the
Combined CatWalk Index (CCI), so that the results are
easier to compare and report. The data used in this study
was collected during a series of experiments utilizing a
mouse model of SCI.
Several SCI experiments were performed using female
C57BL/6J mice (Charles River Japan, Yokohama, Japan),
including experiments to test potential therapeutics as
well as impact force optimization experiments. The data
used in this study was collected from approximately 800
measurements taken from 108 mice.
Mice were anesthetized using an intraperitoneal
injection of a mixture of 0.3 mg/kg Medetomidine, 4 mg/kg
Midazolam, and 5 mg/kg Butorphanol. Anesthesia was
confirmed by pinching the hind paw prior to surgery.
The spinal cord was exposed by laminectomy at the 11th
thoracic vertebra. Contusion SCI was performed using
the Infinite Horizons IH-0400 Impactor [
Systems and Instrumentation LLC, Fairfax, VA, USA)
using peak impact forces between 40 and 70 kdyne. The
animal’s bladders were manually pressed every day to
drain the urine and prevent urinary tract infections.
After all motor function measurements were completed,
mice were anesthetized again and euthanized by cervical
Motor function was monitored using the Basso Mouse
Scale (BMS) and Noldus CatWalk XT (Noldus
Information Technology, Wageningen, The Netherlands) every
week for 6 weeks post-injury.
BMS data was collected by observing a mouse in a
30 × 30 × 15 cm plastic cage for 5 min. Notes were taken
describing several aspects of the animal’s gait, and BMS
scores were calculated according to a flowchart [
CatWalk data was collected using a gain of 0.18, with a
maximum compliant run time of 12.5 s. When possible,
three compliant runs were recorded for each mouse, but
poorly performing mice were often unable to produce
runs faster than 12.5 s. In these cases, three
non-compliant runs were collected. Runs were rejected if the animal
turned around during the run. Post-injury data from each
mouse was compared to a pre-injury baseline made of
three pre-injury measurements to account for naturally
occurring variation in motor function.
The Combined CatWalk Index (CCI) was developed
by correlating 104 CatWalk parameters against observed
BMS scores using linear regression, then combining all
linear regression equations into a single weighted average
(Fig. 1a). The R2 values are used as the weighting values,
so that parameters with strong correlation have strong
weights, while poorly correlating parameters have weak
weights. Calculations were performed using both
Microsoft Excel 2013 and LibreOffice Calc version 5, both
programs produced identical results. Linear regression was
performed on each parameter using the built-in SLOPE,
INTERCEPT, and RSQ functions, which calculate the
slope, Y intercept, and R2 values for a linear regression
equation (Fig. 1c). Using these formulas is much simpler
than preparing a separate scatterplot for each
parameter and linear regression. BMS scores were used as the
Y-axis data and CatWalk parameters were used as the
X-axis data. The CCI coefficients (slope, intercept, and R2
values) for each parameter are listed in Table 1, in order
from highest to lowest R2.
The CCI Score was calculated by combining all
equations into a weighted average. “Adjusted CatWalk
Parameters” were calculated using the original CatWalk data
(See figure on next page.)
Fig. 1 a Schematic of how CCI scores are calculated. Data from N CatWalk parameters is correlated with BMS data using linear regression. Each
parameter produces a slope (M), Y-Intercept (B), and R2, which are listed in Table 1. The CCI score is then determined using a weighted average
of each CatWalk parameter using MX + B linear equations multiplied by R2 as the weighting factor. b Plot of CCI scores against corresponding
BMS scores. CCI scores correlate with BMS scores with an R2 value of 0.7093, slightly higher than the CatWalk parameter with the highest R2
value (Step Sequence Regularity Index, R2 = 0.7048). c Example of how the CCI coefficients are determined from BMS to CatWalk data using a
spreadsheet. Linear regressions are performed for each CatWalk Parameter using the SLOPE, INTERCEPT, and RSQ functions. d Example of how CCI
scores are calculated from CCI coefficients and CatWalk data. Each CatWalk parameter is multiplied by its CCI coefficients to create an “Adjusted
CatWalk Parameter”. The adjusted CatWalk parameters are summed and divided by 104, the number of CatWalk parameters used to create the CCI
Parameters are sorted by R2 value to place parameters with better correlation at the top
and CCI coefficients with this equation: R2 (Slope ×
CatWalk Parameter + Intercept) (Fig. 1d). Adjusted CatWalk
Parameters were then summed and the sum was divided
by 104, the number of CatWalk parameters used in this
analysis. Mock ups of the spreadsheets used to calculate
the CCI coefficients and CCI scores are shown in Fig. 1c,
d, and an example spreadsheet containing the BMS and
CatWalk data from impact force optimization
experiments is available in Additional file 1.
CCI scores were calculated for every mouse at every
time point based on the CatWalk data and the CCI
coefficients in Table 1. The CCI scores were plotted against
the corresponding BMS measurements and linear
regression was used to determine how well the two scores
correlated against each other, and an R2 value of 0.7093 was
obtained (Fig. 1b). This R2 value is not perfect, but may be
a reflection of the BMS system only being
semi-quantitative. For example, if a mouse’s BMS score changes from 1
to 2 (a change from only showing partial ankle movement
to full ankle movement without plantar placement), it is
not the same as a change from 2 to 3 (the mouse shows
plantar paw placement without weight support).
Table 1 shows that the CatWalk parameter that most
closely correlates with BMS scores is the Step Sequence
Regularity Index (SSRI), with an R2 of 0.7048, slightly
lower than the R2 value for the correlation between BMS
score and CCI score, but this difference is likely
insignificant. The SSRI measures coordination by determining if
the order of footprints falls into one of six regular
patterns. Mice with poor motor function have poor
coordination, and do not follow these regular patterns well,
producing low SSRI scores. SSRI is often reported in
studies that use the CatWalk system [
], and the
high correlation with BMS scores supports this practice.
SSRI was chosen to represent CatWalk data when
comparing CCI to CatWalk data in Fig. 2.
Data from the impact force optimization experiments
was used to compare CCI scores to BMS scores and
SSRI scores. Mean scores and sample standard
deviations for each impact force were plotted for every time
point (Fig. 2a–c). CCI and SSRI scores are presented
as a percent of pre-injury baseline to account for
naturally occurring variation between mice. BMS scores are
presented directly, because all non-injured mice have
a BMS score of 9, so there is no pre-injury variation.
All three methods showed similar trends, with higher
impact forces producing lower scores. All three
methods show fairly large standard deviations,
demonstrating the difficulty in producing consistent levels of injury
with the contusion SCI model. This is in part due to the
difficulty of controlling the Infinite Horizons impactor’s
peak impact force. Actual impact forces were usually
higher than the desired impact force, with substantial
Average scores for each impact force were estimated
by calculating the mean score across weeks 1–6 and were
plotted against impact force (Fig. 2d–f ). Linear regression
was used to determine how well each method correlated
with impact force. The CCI scores had a slightly higher R2
value (0.8854) than BMS (0.8636) or SSRI (0.8557). This
indicates that the CCI score correlates well with the peak
impact force in this contusion SCI model.
Coefficients of variation (CV) at each impact force
and time point were determined by dividing the sample
standard deviation by the mean. Each method’s CVs were
averaged and compared using an unpaired, two-tailed
T Test (Fig. 2g). The CCI method showed significantly
smaller CV than either the BMS method or the SSRI
method (P < 0.0001). This implies that the CCI method
may be more precise than either other method.
One advantage of the human-observation based BMS
method is that it produces a single score that can be easily
be compared between mice, but suffers from the
potential for human error and the requirement for training.
The CatWalk system has the advantage of greater
objectivity, but the large number of measurements can
complicate several things, such as choosing parameters for
publication or making comparisons between mice. For
example, if one of set of mice has better coordination, but
another set of mice has higher speed, which set of mice
has better overall motor function? The Combined
CatWalk Index appears to combine the advantages of both
system by creating a single number based on objectively
determined data. In addition, the CCI scores have slightly
better correlation with BMS scores than any individual
CatWalk parameter (Fig. 1b, Table 1), slightly better
correlation with impact force than BMS scores or SSRI
scores (Fig. 2d–f ), and significantly smaller coefficients of
variation than BMS scores or SSRI scores (Fig. 2g).
Although the CCI method requires a specialized
instrument, the CatWalk system is fairly simple, and users can
be quickly trained to measure mice. The BBB, BMS, and
TMS systems require more extensive training, and steps
must be taken to remove human bias or interrater
variability. Additionally, the CCI method could potentially
be modified and applied to any disease model that can
be studied using the CatWalk system, such as chronic
], arthritis, or vestibular disease [
]. The main
requirement is to have some semi-quantitative method to
rank mice so that correlation between rank and CatWalk
parameters can be determined.
The CCI was created by using linear regression of 104
CatWalk parameters against observed BMS scores. The
BMS scores were collected by an untrained observer, so
the quality of the BMS data might not be optimal, but
the decent correlation between BMS score and impact
force in Fig. 2e suggests that they are probably sufficient
for this analysis. A more significant limitation may be the
linear regression model. More sophisticated multivariate
regression models exist, but can be more difficult to use
with large data sets. For example, the multiple regression
method built into Microsoft Excel 2013 can only accept
up to 16 variables. The method presented here is a simple
extension of linear regression that can be understood and
used by almost any researcher using common, or even
open source, software.
Finally, good SCI scoring systems should also correlate
with spinal cord damage. Systematic histological studies
were not performed to verify a link between CCI score
and tissue damage, but the close agreement between CCI
and the more thoroughly studied BMS and CatWalk
systems suggests that CCI is likely to predict spinal cord
Additional file 1. Example spreadsheet with CCI coefficients. An
annotated example spreadsheet containing CatWalk and BMS data from
impact force optimization experiments. This spreadsheet shows how to
calculate CCI coefficients and scores.
BBB: Basso-Beattie-Bresnahan Rat Scale; BMS: Basso Mouse Scale; CCI:
Combined CatWalk Index; CV: coefficient of variation; SCI: spinal cord injury; SSRI:
Step Sequence Regularity Index; TMS: Toyama Mouse Scale.
STC: Study design, SCI surgeries, data collection and analysis, drafted the
manuscript. KI and KK: Study design. All authors read and approved the final
The authors would like to gratefully acknowledge Satoshi Uchida and Satomi
Ogura for training STC to perform SCI surgical methods.
The authors declare that they have no competing interests.
Availability of data and materials
The authors declare that the data supporting the findings of this study are
available within the article and its supplementary information files.
Consent for publication
Ethics approval and consent to participate
All animal experimental procedures were approved by the Animal Experiment
Committee at Innovation Center of NanoMedicine (iCONM), Kawasaki Institute
of Industrial Promotion, 3-25-14 Tonomachi, Kawasaki-Ku, Kawasaki, 210-0821,
Japan. The procedures were approved as Animal Experiment Plan A16-008-4
on May 23rd, 2017.
This work was financially supported in part by JSPS KAKENHI Grant
Number JP15H03017, 16K15642 (K.I.), JSPS Postdoctoral Fellowship for Foreign
Researchers (Grant Number 17F17410 (S.C. and K.I.), and Center of Innovation
Program (COI) from MEXT.
Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
1. Basso DM , Beattie MS , Bresnahan JC . A Sensitive and reliable locomotor rating scale for open field testing in rats . J Neurotrauma . 1995 ; 12 : 1 - 21 .
2. Basso DM , Fisher LC , Anderson AJ , Jakeman LB , Mctigue DM , Popovich PG . Basso Mouse Scale for locomotion detects differences in recovery after spinal cord injury in five common mouse strains . J Neurotrauma . 2006 ; 23 : 635 - 59 .
3. Shigyo M , Tanabe N , Kuboyama T , Choi S-H , Tohda C. New reliable scoring system, Toyama mouse score, to evaluate locomotor function following spinal cord injury in mice . BMC Res Notes . 2014 ; 7 : 332 .
4. Hamers FPT , Lankhorst AJ , van Laar TJ , Veldhuis WB , Gispen WH . Automated quantitative gait analysis during overground locomotion in the rat: its application to spinal cord contusion and transection injuries . J Neurotrauma . 2001 ; 18 : 187 - 201 .
5. Hamers FPT , Koopmans GC , Joosten EAJ . CatWalk-assisted gait analysis in the assessment of spinal cord injury . J Neurotrauma . 2006 ; 23 : 537 - 48 .
6. Scheff S , Roberts KN . Infinite horizon spinal cord contusion model . In: Chen J , Xu ZC , Xu XM , Zhang JH , editors. Animal models of acute neurological injuries . Totowa: Humana Press; 2009 .
7. Lankhorst AJ , ter Laak MP , van Laar TJ, van Meeteren NL , de Groot JC , Schrama LH , et al. Effects of enriched housing on functional recovery after spinal cord contusive injury in the adult rat . J Neurotrauma . 2001 ; 18 : 203 - 15 .
8. Koopmans GC , Deumens R , Honig WMM , Hamers FPT , Steinbusch HWM , Joosten EAJ . The assessment of locomotor function in spinal cord injured rats: the importance of objective analysis of coordination . J Neurotrauma . 2005 ; 22 : 214 - 25 .
9. Singh A , Murray M , Houle JD . A training paradigm to enhance motor recovery in contused rats: effects of staircase training . Neurorehabil Neural Repair . 2011 ; 25 : 24 - 34 .
10. Chen Y-J, Cheng F-C, Sheu M-L , Su H-L , Chen C-J , Sheehan J , et al. Detection of subtle neurological alterations by the Catwalk XT gait analysis system . J Neuroeng Rehabil . 2014 ; 11 : 62 .
11. Kopecky B , DeCook R , Fritzsch B. Mutational ataxia resulting from abnormal vestibular acquisition and processing is partially compensated for . Behav Neurosci . 2012 ; 126 : 301 - 13 .
12. Vrinten DH , Hamers FFT . ' Catwalk' automated quantitative gait analysis as a novel method to assess mechanical allodynia in the rat; a comparison with von Frey testing . Pain . 2003 ; 102 : 203 - 9 .