The AI (Artificial Intelligence) Physical Exam, developed for the Uniform Services University of Health and Science (USUHS), offers students an innovative way to practice efficient physical exams.
The AI Physical Exam virtual environment with a patient in an exam room.
Summary
This interactive platform provides USUHS students with valuable, interactive experience in managing diverse situations and medical incidents during physical exams. Through realistic scenarios using patients from incidents, the game challenges the user to diagnose problems by selecting appropriate actions to test on patients.
Features and Design
Software
Unreal Engine 5 (5.7)
Contributions
Team:
Aathi Anandan, 1st LT, USAF
Zach Galpin
_
Client:
Uniform Services University of Health and Science
Dr. Cole
Outcomes and Additional Information
The physical examination is a cornerstone of clinical medicine, allowing healthcare professionals to gather objective data to complement the patient's history. The goal of the interactive virtual trainer is to provide a realistic training experience that is mobile, accessible, and does not rely on recruiting other people to act as patients.
The user must analyze a patient's background, medical history, and personal attitudes to take specific actions to arrive at the final solution efficiently. Their performance is rated on both the quality and efficiency of their approach.
The system can take a user's question, send it to the AI chatbot, receive a response, and display that response directly within the virtual environment. The patient scenario information (documentation) is connected to the prompt being sent to the AI so that responses can be tailored to the specific patient case rather than acting as a generic chatbot.
The prototype tests the student in three scenarios. The AI was trained to answer questions related to patient symptoms, medical history, current medications, allergies, and other pertinent information for each scenario.
Scenario 1 tests the user’s ability to target the lower GI and pelvic systems on a young female patient while avoiding time-wasting neurological or heavy musculoskeletal exams.
Scenario 2 tests if the user can recognize fluid overload and target the cardiovascular and pulmonary systems for an older male patient.
Scenario 3 tests the user’s ability to assess vascular emergencies and recognize the risk of pulmonary embolism in a middle-aged male.
The completed virtual training would include general survey and vital signs; assessment of the cardiovascular and pulmonary systems; and neurological, musculoskeletal, genitourinary & rectal exams.
Future work includes improving the user interface, refining the patient prompts to produce more realistic and consistent patient behavior, and then creating a grading system to produce an efficiency score based on the user's actions during the exam.
The AI Physical Exam patient creator (top left) and scenario creator (top right).