Development of a football-specific image processing-based reactive agility test system and investigation of its reliability and validity

Reactive agility refers to a player's ability to quickly respond to changing environmental stimuli and position themselves, particularly in sports such as football that require open skills. Reactive agility testing systems developed for this purpose typically utilise light-based stimuli, infrared sensors, and motion capture technologies. These systems enable a multidimensional analysis of athlete performance in terms of both decision-making and motor response measurement. However, significant disadvantages include their high cost, limited portability, low reactive agility capability, low sport-specific characteristics, and the technical knowledge required for their application. With the advancement of technology, projection-based visual cueing systems integrated with image processing techniques are giving rise to ideas that could enable the creation of more realistic and interactive scenarios in reactive agility tests. High-resolution and wide-angle projections enable the creation of a test area for reactive agility, while the idea of developing a new reactive test measurement system has emerged by determining the spatial-spatial positions (x, y, and z) of athletes within this area using a depth camera. In this context, the aim of this study is to develop an image processing-based reactive agility measurement device and to examine this device in terms of validity and reliability. The method of this study consists of two main stages. In the first stage, the hardware, software interface, and sport-specific movement library of the image processing-based reactive agility testing and training system were developed. The system consists of a wide-angle, high-resolution projection device, a depth camera, a control unit, and a projection surface. The projection device projects onto a 12-square-metre area at a 120-degree angle from a height of 5 metres. This test area is divided into 25 squares using the Matlab programme. Stimuli (football images) are projected onto these squares using software. The main logic of the software is based on projecting a football image onto the area, determining the athlete's position within this area using the depth camera, and providing a new stimulus when the athlete distorts this image with their foot. In this study, three different reactive agility test scenarios, three different direction change models, and one planned direction change (PDC) protocol were created in the new system. In the second phase of the study, the system's validity (laboratory and field validity) and reliability (test-retest) were tested. Within the scope of laboratory validity, the time data obtained from the device in different scenario models were compared with the times determined using image analysis software (KINOVIA) via high-speed video recordings. Field validity was tested by comparing reactive agility tests and PYD tests (T-test, Illinois test, and system PYD) between elite and non-elite footballers to determine the relationship between these tests. Accordingly, it was hypothesised that professional footballers would have significantly higher reactive agility values, while this difference would be less pronounced in PYD values. The reliability and validity study included 24 professionals (age: 23.5±4.2 years, height: 176±3.28 cm, body weight: 74.29±2.42 kg, experience: 8.7±3.1 years) and 33 amateurs (age: 21.8±3.7 years, height: 174±4.11 cm, body weight: 74.89±3.76 kg, experience: 5.2±2.4 years) participated as volunteers. An independent t-test was used to determine the difference between methods in laboratory validity. Validity was also tested using the intraclass correlation coefficient (ICC), linear regression analysis, and Bland-Altman limits of agreement (LOA). Field validity was determined by comparing elite and non-elite groups and by the relationship between PYD and reactive agility tests. A paired t-test was used to determine the difference in test-retest reliability, while reliability was also assessed using ICC, linear regression analysis, Bland Altman LOA, and coefficient of variation % (CV%). According to the laboratory validity results, no significant difference was found between the new system and video analyses in all three scenarios and models. However, a very high SKK (between 0.92 and 0.99; p<0.05) was detected. Furthermore, while no systematic bias was found between measurements, random error values were very low, and R2 (between 0.97 and 0.99, p<0.05) was quite high. In field validity; in the planned change of direction test in all scenarios and models (T-test: Illinois Test and Systemic PYD), a statistically significant difference was found in favour of professionals compared to amateur footballers (between 6.42% and 8.9%), while this difference was much higher in reactive agility values (between 13.08% and 19.68%). However, a low-to-moderate positive correlation was found between PYD tests and reactive agility tests (r: between 0.25 and 0.57, p<0.05). According to test-retest reliability data, no significant difference was found between measurements. However, a very high CTT (between 0.92 and 0.99, p<0.05) was detected. Furthermore, while no systematic bias was found between measurements, random error values were very low and R2 (between 0.84 and 0.97, p<0.05) was quite high. In addition, VK% values were found to be within acceptable ranges (between 0.96% and 3.47%). At the end of the study, it was observed that the newly developed reactive agility test measurement system has high validity under laboratory conditions. It also demonstrated the validity of the cognitive factors, which are components of reactive agility, by distinguishing between classes and proving its validity in field conditions. Furthermore, high test-retest reliability was determined. The computer software-based nature of the system and its potential for development and application in reactive measurements across different disciplines are expected to contribute to the emergence of a new measurement model in the literature. Further studies on the applicability of the developed system in different sports disciplines, age groups, and stimulus types are expected to add a new dimension to sports performance testing.