Virtual patients are able to represent patients in realistic clinical scenarios and engage learners in doctor-patient conversations about the patient's health, interpret laboratory results and medical images, and form a diagnosis. Currently, due to the conditions created by the COVID-19 pandemic, the interaction of medical students with patients is limited, therefore virtual doctor-patient interaction environments become a safe and practical solution for training medical skills, including blended learning. With the current advances in artificial intelligence, virtual patients can be equipped with advanced functions, such as the integration of different types of conversational agents or support for the automatic prediction of the patient's evolution.
In this context, the main objective of the project is to develop a virtual environment to simulate the interactions between the learner (medical student or doctor/resident) and a virtual clinical patient for the diagnosis and treatment of acute and chronic heart diseases. The virtual environment, supported by a software platform and artificial intelligence technologies, simulates real-life clinical scenarios in which the student emulates the role of the doctor who performs an examination of the virtual patient and obtains a history, performs anamnesis, physical examination, paraclinical investigations, establishes a diagnosis and recommends a therapeutic plan.
Starting from our already developed system, the challenges of this project are to advance virtual patient technology by developing a flexible and integrated cardiology training environment that incorporates advanced artificial intelligence techniques, especially case diagnosis and treatment planning simple and complex, a conversational agent capable of dialogue in natural language and voice, as well as the generation of synthetic data based on deep neural networks.
The project consortium consists of:
Patent of Invention Application: Patent Invention Application submitted to the State Office for Inventions and Trademarks (OSIM registration number: A/00323 from 13 June 2024) - Artificial intelligence-based system for training in the diagnosis and treatment of cardiovascular diseases.
Data Set: Extending the data set which contains data for 7 cardiovascular cases: 5 cases - unique pathology and 2 cases - multiple pathologies.
Data Set: Creation of a data set which contains data for 7 cardiovascular cases: 5 cases - unique pathology and 2 cases - multiple pathologies.
The project developed an e-learning platform that simulates the interactions between a medical student or resident and a virtual clinical patient for the diagnosis and treatment of acute and chronic cardiovascular diseases. The platform simulates real-life clinical scenarios in which the student plays the role of the doctor who initiates a dialogue with the virtual patient to find out his history, performs an anamnesis and physical examination, requests paraclinical investigations, establishes a diagnosis and recommends a treatement plan. The interface of the virtual patient allows interaction in written and spoken natural language, in Romanian and English. The interaction of the student with the virtual patient can be viewed by the teacher, both synchronously and asynchronously, with the teacher giving notes and feedback for the actions performed by the student.
Data were collected and anonymized for a number of 5 clinical cases for patients with unique conditions, respectively: Pulmonary Thromboembolism, Aortic Stenosis, Myocardial Infarction, Mitral Regurgitation and Cardiac Insufficiency; and a number of 2 clinical cases of patients who have two or more conditions, respectively: the first case in which the patient simultaneously suffers from acute inferior myocardial infarction, severe functional mitral regurgitation and heart failure; and the second case where the patient suffers from heart failure and mitral stenosis simultaneously.
An automatic scenario generation algorithm was developed for the different clinical cases (mentioned earlier), with data consistency checking, and synthetic data were generated to expand the datasets.
The platform was validated with 5 groups of students from UMFCD in several stages, totaling a number of 210 users: 178 students from UMFCD, as direct beneficiaries of the platform, 25 students from UNSTPB to evaluate the technical aspects of the platform (total students 203) and 7 teachers from UMFCD. Feedback was collected from them. The results of the feedback were analyzed and used to improve the functionality and performance of the platform. Students considered the Virtual Patient platform as a significant aid in learning, and teachers shared the same opinion with students.