Anatomical dissection plays a fundamental role in medical education, offering hands-on experience that deepens students’ understanding of human anatomy. Virtual Reality (VR) applications have emerged as powerful tools for medical learning, enabling immersive and interactive learning experiences that supplement traditional methods. In these VR environments, precise and intuitive interactions remain a challenge due to limitations in movement recognition and unintended object selections. This study introduces a transformer-based neural network approach to enhance object selection in VR-based learning anatomical dissection by analyzing user movement sequences. Our AI model achieves a 95.5% validation accuracy, significantly improving interaction precision and reducing recognition errors. Our experiments confirm that transformer-powered sequence recognition enhances user interactions, leading to a more natural and effective learning experience. In this paper we present the results from training a model with a dataset collected from over 200 medical and biomedical engineering students, immersing them in custom-made VR scenarios that mimics real-world anatomical dissections experiences. This work serves as a foundation for the development of more sophisticated anatomical dissection simulations, paving the way for next-generation VR-based medical learning environments with enhanced interactivity and precision.