Sprint Biomechanics Assessment with Low-cost Systems: a Reliability Study

The individual determination of force-velocity and power-velocity pro�les during sprint is of great interest to coaches and sports physiotherapists. As a very short action, sprint evaluation requires a su�ciently accurate and reliable system. The aim of this study was to analyze the reliability of the free software Kinovea®, compared to the MySprint App (Apple Inc, USA). Thirty-one soccer players were evaluated and a comparative study was carried out, where 62 sprints of 30-meters were analyzed by two rates: experienced and non-experienced. Vertical poles were placed at 2.5, 5, 7.5, 10, 15, 20, 25 and 30 meters. All the sprints were recorded in slow motion and HD image quality. Comparisons of partial and total times were made, in addition to force, velocity and power outputs. No differences were shown between the two measurement methods for the different sprint times (ICC = 0.676–0.941, p < 0.001). The intra-rater reliability of total time in the experienced rater was almost perfect: ICC = 0.993 for Kinovea and 0.984 for the MySprint app; the intra-rater reliability for non-experienced one was 0.833 for Kinovea and 0.862 for the MySprint app. Comparing both methods, the ICC was 0.896. There were no signi�cant differences between the variables force, velocity and power (p > 0.05). This study shows that Kinovea + Excel spreadsheet is a reliable method, also an accessible and low-cost option for sport professionals. However, experience using the software is required, but not for the use of the MySprint app, which is an advantage for non-experienced testers.


INTRODUCTION
The assessment of sprint is a frequent measurement in most sports, being decisive for sport performance.
Force-velocity (F-v) and power-velocity (P-v) pro les are useful information and could be estimated with a 30meter sprint.Mostly, these values have been taken using photocells (taking a known distance) or with radars or laser systems.[1] The latter are not always usual, photocells being more accessible.However, chronometers have also been used as a replacement of photocells, as a low budget element but with a proportional relationship between cost and precision.[2,3] Therefore, a simple method has been validated to track the athlete's center of mass (CM) during the sprint acceleration phase.This method allows estimating ground reaction forces in the sagittal plane and determining the F-v and P-v pro les and the effectiveness of its applied force, based on biomechanical models developed by the authors.[4] MySprint app was released in 2015 for iPhone 6 or newer versions.This app allows to obtain partial sprint times every 5 meters over a distance of 30-meter applying the biomechanical formulas included in an Excel spreadsheet proposed by Samozino.[1,4] This app was validated using radar and photocells simultaneously, synchronizing the three devices.[1] Not all coaches and sports scientists have an iPhone device at their disposal, so other types of solutions based on conventional cameras are required.Furthermore, despite the feedback is not in real time, excellent validity and reliability of video recording have been proven in comparison with photocells and laser, especially for distances less than 30 meters.[5] For this reason, high-speed video recording is considered a precise, reliable, low-cost, easy-to-administer method that allows for the collection of multiple timing data from different athletes.[6]Kinovea is a free software available for Windows, which provides accessibility to any user, while other video analysis programs are paid.Kinovea allows 2D video analysis where time and distances can be measured.Despite being simpler, 2D analysis methods have demonstrated validity, reliability, and accessibility.[7][8][9][10] Nevertheless, an adequate organization of the setting should be considered for data consistency.[7] For the reasons mentioned, this current research used two systems: 1) Kinovea plus an Excel spreadsheet, and 2) the MySprint app.As a prior objective, the values obtained from each system were correlated, investigating intra-and inter-rater variations.On the other hand, the values of the variables F0, v0, and P were compared.
Thus, this research highlights the potential of a low-cost or cost-free and easy assessment for the sports community.

Design
An intra and inter-rater reliability study was conducted, analyzing a total of 62 sprints of 30-meters.

Participants
A total of 31 soccer players aged between 17 and 20 years (18 ± 1.1 years, mean ± SD), with and height of 1.69 ± 0.31 meters, and a weight of 70 ± 5 kg were evaluated.Exclusion criteria were receiving physical therapy treatment at the time of evaluation and having had a muscle or joint injury within the previous 6 months.All participants were informed about the procedures involved, risks, and bene ts, and they signed informed consent approved by the Ethics Committee of the University of Gran Rosario.

Procedures
Prior to evaluation, participants performed joint mobility warm-up, jogging, and progressive sprints.A body marker was placed on the right greater trochanter of each player to identify the moment when it crossed each distance poles and thus obtain split times.
A 30-meter sprint analysis was conducted, with vertical poles placed at 2.5, 5, 7.5, 10, 15, 20, 25, and 30 meters.Each soccer player executed two sprints with at least a 3 minutes rest between them.The camera was positioned at the midpoint of the sprint distance (15 meters) and 18 meters perpendicular to the running direction.Parallax errors were corrected using calculations provided by the Excel spreadsheet developed by Stenroth et al., a modi cation of Samozino's one.[11] Participants were instructed to start the sprint from a three-point stance and the sprint initiation was determined when the hand was lifted off the ground, as done previously.[1,4] Simultaneous video recordings were conducted using an iPhone 6 and a Xiaomi Mi 9 Lite to obtain the same images in terms of time and space.The iPhone recordings were analyzed using the MySprint app, while the Xiaomi recordings were analyzed using Kinovea software plus an Excel spreadsheet modi ed by Stenroth et al, a modi cation of Samozino's one.[11] The recordings were collected in slow-motion mode on the iPhone − 240 fps as technical features indicate-and at 240 fps on the Xiaomi, both at a resolution of 1280x720 pixels.
Each analysis was performed by two raters: an experienced one with over a year of experience using this methodology, and a non-experienced one using the methodology for the rst time.Each rater analyzed each sprint twice and comparisons between each analysis allowed to calculate intra-rater reliability and inter-rater reliability.
Analyses using MySprint were performed on the same device used for recording, obtaining split times at 5, 10, 15, 20, 25, and 30 meters when the marker set in the greater trochanter was aligned with the distance poles, as well as with values for F0, v0, and P (absolute and relative).The results were transferred to a matrix Excel spreadsheet where all the variables were shown for each player.These analyses were conducted by an experienced and a non-experienced rater.
Analyses in Kinovea were performed by exporting the videos recorded with the Xiaomi Mi 9 Lite, identifying key frames every time the marker on the greater trochanter aligned with the distance poles, and then applying the stopwatch tool.With the identi ed times, the Excel spreadsheet designed by Stenroth et al. [11] was completed with split times for each distance, also considering the temperature and atmospheric pressure data at the time of evaluation.As well, these analyses were conducted by an experienced and a non-experienced rater and also transferred to a matrix Excel spreadsheet where all the variables were shown for each player.

Statistical Analysis
All data from both MySprint and the Excel spreadsheet were transferred to a master spreadsheet containing the variables of interest for this study: times in seconds at 5, 10, 15, 20, 25, and 30 meters; F0, v0, and P (relative).
For correlation analysis of split times at 2.5 and 7.5 meters and the absolute P were omitted due to the missing calculation of this variables with the MySprint app or Excel spreadsheet.Paired Student's t-tests were used to compare the values of maximum theoretical force (F0), maximum theoretical velocity (v0), and power (P) between the MySprint app and the Excel Spreadsheet.Each tester performed two analyses of the same video, allowing to calculate intra-rater reliability.Additionally, each tester's rst analysis was correlated to determine inter-rater reliability considering the evaluator's experience (experienced vs. non-experienced).
Thus, the intraclass correlation coe cient (ICC) was calculated to measure the reliability of the sprints using both evaluators within each measurement method.The reference classi cation values for ICC in this study were described by Landis and Koch as: 0 poor, 0.01-0.20 slight, 0.21-0.40fair, 0.41-0.60moderate, 0.61-0.80substantial, and 0.80-1.00almost perfect.[12] All statistical analyses were performed using IBM SPSS Statistics 22 and Microsoft Excel 2010, with a signi cance level set at p < 0.05.

RESULTS
Descriptive statistics (mean ± SD) of inter and intra-rater measurements using both measurement methods are shown in Table 1.
The distribution of the data at different split times measured with both methods is shown in Fig. 1, where no differences were found, showing a very good correlation between the two measurement methods (ICC = 0.676-0.941,p < 0.001).Regarding the evaluation of maximum theoretical force (F0), maximum theoretical velocity (v0), and power (P), Table 2 shows that there are no signi cant differences between the average force, velocity, and power values for the two measurement systems (p = 0.053-0.695).
No signi cant differences were found by analyzing the average of F, v and relative P using the MySprint app and Kinovea.Moreover, if we focus on the rst 5 meters covered, the differences between the raters increase.(Figs. 6 and 7).

DISCUSSION
This study is aimed at determining the reliability of the software Kinovea combined with an Excel spreadsheet for sprint biomechanical calculations, compared to the MySprint App, identifying differences between experienced and non-experienced testers.In consequence, an analysis of 62 30-meter sprints was conducted in young football players, comparing the data obtained from both methods' partial times.
Regarding the set of characteristics of the developed methods based on video recording, it allows greater applicability in sports eld, being a valuable instrument for coaches and physiotherapists while maintaining ecological validity.As for the recording features of the smartphones used, both videos were shot at 240 frames per second (fps) with a resolution of 720p.
Sprints mechanics has been analized in football players to follow up season changes [13] and identify hamstrings injury risks.[14] Also, MySprint app has been used to test 30-m sprint performance among different maturational stages.[15] Analyzing sprint biomechanics provides valuable information for sports professionals, even more so in the case of sports performance.
Total times analyzed in the 30 meters showed an almost perfect intra-rater correlation, with a ICC of 0.99 and MAE 0,011 for the experienced rater while an ICC of 0.83 and MAE 0,018 for the non-experienced one using the Kinovea method.For MySprint app, the ICC was 0.98 and MAE 0,013 for the experienced and 0.86 and MAE 0,036 for the non-experienced.These results suggest that some experience should be necessary to get reliable data using this methods, although correlation is almost perfect.Dispersion data can be observed in Figs. 1 to 5, with very good correlation (ICC = 0.676-0.941,p < 0.001).
However, this was different for the rst 5 meters, where practice and visual acuity are required to identify the moment of the start, considered as the hand leaving the ground.Here, almost perfect values were found for the experienced rater using Kinovea (ICC = 0.94) and ICC = 0.73 for the experienced rater using MySprint, indicating a substantial correlation.For non-experienced evaluators, an ICC of 0.65 was found for Kinovea system and 0.696 for the MySprint app, wich is still a substancial correlation.Many errors could have been made by the non-experienced evaluators: identifying the start of the sprint with the hand movement, being crucial for the split times obtained; identifying the body marker on the greater trochanter, setting the key image on the ideal frame and the use of the stopwatch.
To date, this is the rst study that has compared these two measurement systems.The Excel spreadsheet has been used by some authors in the eld of ice skating, but a paid software was used for video analysis.[11] Also, in another study, a high-speed camera at 240 fps was used in an ice hockey player's acceleration test. [6] However, the parallax corrections were not detailed in relation to the lens used and the visual eld.At same point, it could modify the times measured.
Another aspect to consider regarding other assessment methods is that, due to the different heights of the athletes, there is a possibility that the photocell may be triggered at different moments, altering the criterion for starting the timer (without being the starting line), making di cult its implementation with great precision.[5] In this study, a marker was placed at the level of the greater trochanter to determine partial split times so it makes possible to always locate the same point when passing through the marks; as well as it would also allow to compare evaluations at different times of the season.
Previous studies have been conducted using video recording for sprint analysis, using softwares such as Dart sh or QuickTime.In one of them, the number of fps used in the recording was not speci ed, but based on the description and publication year, it is estimated to be a standard 30fps video.In the other, a recording at 240fps was used.[6,16] Alternatively, Kinovea offers the advantage of being free cost and available for any Sports and Exercise professional, same as the Excel spreadsheet.Kinovea was created in 2009 and is a free, open-source software where people from around the world collaborate in its creation and de nition of new versions.Due to its advantages, it is a great tool for those working in sports and enables video analysis without interfering with the athlete's actions.[17] One of the main areas of interest when using high-speed cameras has been the assessment of contact times in running or sprinting and jumping, which allows for quantifying performance.This can also be applied to identify the phases or the different timing of a movement, a sports technique, or a motor skill.
[18] Time quanti cation using the stopwatch tool in Kinovea is equally valid and reliable compared to measurement with infrared technology.The inter-rater reliability using Kinovea was ICC = 1; the ICC for the Kinovea vs. OptoJump inter-rater comparison was 0.9997, with a difference of 2.2ms in ight time and 0.31cm in jump height between the calculations of both methods.
[8] Measurements of countermovement jumps (CMJ) have also been conducted, comparing them with a 3DMA system, which included 112 subjects.It was concluded that the use of a smartphone and Kinovea software is a valid, reliable, and useful method for measuring CMJ height and its derivatives.The ICC values found were high, demonstrating excellent reliability between both methods: 0.98 for height, 0.98 for take-off velocity, and 0.99 for impulse.The minimum detectable change (MDC) was also calculated for different variables: jump height, take-off velocity, and impulse, which were 1.34cm, 1.15m/s, and 2.93Ns, respectively.[19] Lastly, Kinovea was also used to correlate with other methods of measuring jump height (based on the mentioned formula), nding good ICC values (> 0.85), although underestimating the jump height, but with consistent results).[20] In relation to the raters experience, more consistence data in the experienced one, with fewer differences between rst and second analysis can be observed; while in the non-experienced, this variation was greater, with both methods used.Regarding the test method between experienced vs. non-experienced, MySprint app had better correlation results, with a con dence interval of 95% between 0.95-0.981over the Kinovea method with the Excel spreadsheet (0.902-0.963).This suggests that My Sprint app might be useful when testers are non-experienced, although future research may shed light on this topic.
Team sports require short accelerations and decelerations, so the rst 5 meters are decisive.According to this study, the analysis of the rst 5 meters was more consistent with the Kinovea method (Figs. 6 and 7).Probably, using a computer rather than an iPhone to do sprint reviews and analyses can offer an advantage in this regard, since a bigger screen is used.
To calculate F, v and P, no signi cant differences were found between the two methods (Table 3), so both are a good option to get these variables in a biomechanical analysis of the 30-meter sprint.
The F-V pro le is expected to be useful for both researchers and sports professionals since it allows to thoroughly know the characteristics of the athletes and, based on this, the most effective training prescription according to their needs.[21] CONCLUSION This investigation demonstrates that Kinovea is a reliable method for performing a biomechanical analysis of the 30-meter sprint, as well as being an accessible and low-cost option for sports professionals.However, some experience using the software is required for these tests.On the other hand, this experience does not seem necessary for using My Sprint app, which is an advantage for non-experienced testers.

AKNOWLEDGEMENTS
The authors thank Paola Giménez from Gran Rosario University and Lisandro Leon for their collaboration with this project.Also, thanks to Club Atlético Rosario Central for the loan of facilities and players for data collection.
Bland Altman plot for experienced intra-rater reliability using Kinovea Bland Altman plot for experienced intra-rater reliability using My Sprint app Page 13/16  Bland Altman plot for non-experienced intra-rater reliability using MySprint app

Figures
Figures

Figure 1 Data
Figure 1

Table 1
Split times (mean ± SD in seconds) using the two different methodologies (kinovea vs. MySprint) by raters (experienced vs. non-experienced) and analysis ( rst vs. second)

Table 2
is lower than MySprint.Moreover, for non-experienced raters, MAEs are higher than in experienced ones.This is also shown in Figs.2 to 5. a ICC: Intra-class Correlation Coe cient b MAE: mean absolute error c Lin CCC: Lin's Concordance Correlation Coe cient