A Physics-based Virtual Reality Framework for Medical Training and Simulation
Open AccessMedical simulation offers the opportunity to revolutionize the training of medical professionals, from paramedics to physicians and surgeons, allowing early learning to occur in a safe, controllable, and configurable virtual environment, without putting patients at risk. However, the complexity of the problems involved in the development of medical training systems as well as the spectrum of scientific fields that need to be covered have been the major limiting factors for the achievement of realistic simulations. The current state of the field of medical simulation is characterized by scattered laboratories or commercial entities using a variety of models that are neither interoperable nor independently verifiable, resulting in a steep development curve, duplication of efforts, high cost for simulators, and slow adoption of the technology. There have been some attempts to provide open-source standardization. However, they offer a collage of specialized solvers for different substances (bones, tissues, fluids, etc.), which creates redundant work and does not provide stable and efficient two-way interactions between all object types. Lastly, most of the simulation systems do not provide a means for automated assessment that can record, visualize, and analyze trainees’ performances through quantitative measures, which is an often neglected aspect of medical simulators. This dissertation addresses the issues of medical simulation in an attempt to bridge the gap left by previous works. We propose a practical and efficient virtual reality simulation framework that converts the training of medical procedures to a completely immersive virtual environment where both visual and physical realism are achieved. Our generalizable framework embeds independent dynamics models and interaction devices in separate modules while allows them to interact with each other within the same environment, which offers a flexible solution for multi-modal medical simulation scenarios and enables new simulators to be built efficiently. Our framework includes simulation of all human body constituents, such as bones, soft tissues, and fluids (e.g., blood, secretions), in a unified particle representation using position-based dynamics, which enables different simulated substances to interact with each other seamlessly and allows for efficient modeling of large objects with many different properties in real time. Our system supports inputs of patient-specific anatomies from many sources (serial-section, volumetric, or surface scans), which provides realistic anatomical structures and can be parameterized to allow variations in a range of features that affect the level of difficulty. Moreover, with virtual representation of all the components involved in the procedure, our automated assessment system can capture and visualize a whole set of performance parameters of the instruments in relation to the geometric change of the virtual model for real-time guidance and post-trial assessment. In addition, an interpretable automated scoring system is developed that uses a machine learning algorithm to mimic the evaluation of human raters. Finally, we demonstrate the utility of the framework by developing a test-bed application for neonatal endotracheal intubation. The clinical realism of the VR simulator and the validity of the automated assessment were assessed with a group of neonatologists using qualitative and quantitative measures.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.