Early detection and intervention of injuries, lesions, or brain anomalies can significantly improve athletes recovery process, reducing long-term impact and unexpected health side effects. AI-supported health anomaly detection systems provide high accuracy and consistency in real-time image analysis, out-performing human counterparts, especially in high-throughput situations. Moreover the scalability of AI allows the rapid processing of large amounts of data, making comprehensive screening of athletes feasible. Motivation-wise, AI's ability to integrate multiple data sources, like game statistics and wearable sensor data, offers a holistic approach to understanding and managing, or even preventing the head injury risks in time. The novelty in this field lies in the application of Neural Architecture Search for optimizing model architectures, transfer learning for model enhancement, multimodal learning for comprehensive analysis, as well as explainable AI to provide intelligible insights, thereby building confidence in applied AI ecosystems. We conducted a study to find a feasible, machine learning-based pipeline for sports safety, which could identify and detect brain injuries and tumors early on to help doctors and athletes reduce the risk of serious complications and disorders caused by severe collisions and concussions.