Anticipating and examining the alignment of the neck in head impacts using Explainable Artificial Intelligence

The objective of this study is to address the lack of research in assessing and measuring the movement of the human neck in connection to the direction and intensity of head impacts, with the goal of enhancing safety measures in healthcare. This work specifically examines the application of explainable artificial intelligence (AI) and machine learning (ML) models to accurately predict the orientation of the neck by analysing muscle forces. The aim is to communicate these findings in a manner that aids doctors in making informed decisions. The experimental setup was designed to replicate four types of fairly severe head impacts, namely flexion and lateral displacement, which are frequently seen in median American football crashes. In total, there were ten participants, consisting of five males and five females. Biomechanical modelling and simulation using the Open Sim program enabled the determination and visualization of the neck direction in inverse dynamics study of the spine. This study utilized the annotated tendon force muscle data obtained from studies to train many machines learning models, including a sequential neural network. Among these models, the XGB classifier achieved superior accuracy, having a neck direction accuracy rate of 98%. Considering the accuracy rates established in previous investigations, the findings of this research have promise for comparison with other studies in the same field of inquiry. By integrating explainable AI techniques, the interpretability of predictions generated by machine learning algorithms is enhanced. The knowledge supplied is valuable for comprehending the generated findings. This work contributes to the expansion of knowledge on the biomechanics of the neck during head impacts and emphasizes the use of an AI-based system to improve early preventive mechanisms and assist in the decision-making process for injury management.