A graduate of the Faculty of Applied Mathematics, Informatics, and Mechanics, PhD in Technical Sciences Maria Demchenko developed a decision support system, which can be used to choose and adjust treatment for patients with cardiovascular diseases. The system takes into account the dynamics of the treatment process.
Maria Demchenko developed the system as a result of her PhD thesis, which she defended last year in Nizhny Novgorod. Her research advisor was Irina Kashirina, DSc in Technical Sciences, professor at the Department of Mathematical Methods of Operations Research of the Faculty of Applied Mathematics, Informatics, and Mechanics of VSU. At the moment, Maria Demchenko is an expert on machine learning at ITentika.
The system she developed is based on machine learning methods and reinforcement learning.
“The system involves a model acting as an agent which interacts with the environment in the trial and error mode. Based on its actions and the obtained feedback (rewards for those actions), the agent learns to see the difference between effective strategies and the ones that can lead to undesirable consequences”, said Maria Demchenko.
As a result, the agent develops optimal strategies for each state of the environment. For patients, depending on their physical state, this will mean an optimal treatment.
Maria Demchenko believes that the system still needs to be elaborated to function stably.
“Basically, we need to incorporate special medical reference books and rules, obtain validation by medical experts, and develop algorithms for the processing of unique cases.”
The system was tested using the patient data from the MIMIC III international deidentified database. The database is freely-available and contains deidentified health-related data associated with over forty thousand patients who stayed in critical care units of the Beth Israel Deaconess Medical Centre (a teaching hospital of Harvard Medical School in Boston) between 2001 and 2012. The database includes information such as demographics, vital sign measurements made at the bedside, laboratory test results, procedures, medications, caregiver notes, imaging reports, and mortality (including post-hospital discharge). All patients in the study had been diagnosed with atherosclerosis.
A computational experiment demonstrated that, if the system developed by Maria Demchenko had been introduced, most of those patients would have received more effective treatment and would have spent less time at hospital. The method suggested by Maria Demchenko is much more effective than the existing common treatment strategies. Therefore, it will soon be introduced in medical practice.
