Phase II Amount
$1,098,879
One of the major challenges in healthcare training using patient simulators such as manikins, task trainers, screen-based virtual patients is the lack of natural language communication capability between trainers/trainees and the patient simulators, which makes the communicating skill training almost impossible and the overall training experience unreal. Another challenge is that while conducting a training, the trainer has to observe learners and run the simulator at the same time, which can be overwhelming. Recent advancements in AI technology on natural language processing, and machine learning have made it possible to create close-to-real-life healthcare training experiences by making the simulator more intelligent. The objective of this topic is to develop a platform that enables a patient simulator to carry out a conversation in natural language with an instructor and trainees in a healthcare training context, as well as enable the instructor to use voice commands to control the simulator for setting up training scenarios and changing physiology parameters. During the Phase I effort for this topic, this proposing firm, Sharp Vision Software, designed a proof-of-concept architecture using deep machine learning algorithms on an industry-leading, cloud-based AI platform on multiple devices including a small device that can be easily embedded inside a manikin and a med-sized device with audio and 3D animated video. A working prototype was then developed, integrated and tested with a high-fidelity full-body manikin to assess feasibility, performance, and areas for improvements for the architecture proposed. The prototype has been successfully demonstrated to the DOD team at USUHS with encouraging results. Building upon the success and lessons learned from Phase I, a formal development is proposed for developing the platform to meet the Phase II objective, which is to deliver a product prototype of the Intelligent Patient Simulation Platform. The development activities will include functional requirement, architecture design, component design, coding, testing, and delivering to ensure the final product has a solid base in quality and stability as a platform. Agile development methodology is expected to be used to adapt to changes in technologies. The key features of the platform to be developed in Phase II are expected to accurately understand instructor/learners questions in various ways in natural conversations; 1. respond to instructors commands to change the physiological condition of the patient simulator for different training scenarios; 2. answer and ask sensible questions in the context of medical training; 3. learn to understand questions asked in different ways using machine learning; 4. build a library of trained patient simulators with different personalities for various training scenarios; 5. expand to handle multiple languages in the future; 6. be compatible with selected DOD funded platforms, such as AMM and BioGear.
