Mohsen Esmaeili Sani - PhD student in sports management, Mazandaran University-Babolsar, Iran.

Objective: The aim of this study is to propose a deep learning approach for detecting head injuries in football video data using spatial-temporal features.Methods: The proposed method employs ResNet-۵۰ architecture and the Temporal Shift Module (TSM) for feature learning and classification. The algorithm is trained with a publicly available soccer video dataset labeled with annotated head injuries. The evaluation of the proposed method is done on a test set that includes ۵۰۰ football videos, and the evaluation criteria used include overall accuracy, precision, recall, and F۱ score.Results: The proposed algorithm achieves an overall accuracy of ۰.۹۸۶ in detecting head injuries in the test set, which is a significant improvement compared to previous studies in the same field.Conclusions: The proposed method provides a promising approach for head impact event detection using spatio-temporal features, which could have important implications for sports and medical industries. However, the model requires a large amount of annotated data for training, and future research could focus on addressing limitations such as developing more efficient training methods and incorporating other techniques to identify head injuries outside the camera's field of view.

Brain Concussion, Traumatic Brain Injury, Machine Learning, Neural Networks (Computer), Video Recording