Explainable AI Based Neck Direction Prediction and Analysis During Head Impacts

Abstract

The position and orientation of the human neck need to be measured, analyzed, and monitored during head impacts to provide preventive healthcare measures. The proposed research explores the role of explainable Artificial Intelligence and Machine Learning in predicting the direction of the neck by analysing the muscle forces with appropriate explanations to support any clinical system for correct decision-making. The experimental data includes mild head impacts that were replicated with median American football simulating flexion and lateral movements and was performed on ten subjects including five male and five females. OpenSim software system is used for biomechanical modeling, simulation, and analysis and to visualize the direction of the neck. Different machine learning models including a sequential neural network model were built on the labelled tendon force muscle data. The XGB classifier achieved the best performance in predicting the neck direction with an accuracy of 98 percent. The proposed research work’s accuracy is promising when compared to existing works. Explainable Artificial Intelligence integration deduces the predictions of machine learning models adding meaningful interpretations to the achieved results.

Keywords

• Biomechanical head and neck systems
• machine learning
• explainable AI