The neural network analyzes information on a patient’s anamnesis vitae, pulse, blood pressure, and even relatives’ diseases - 69 indicators altogether.
Scientists of the Perm State National Research University (PSNRU) have created an intelligent system for diagnosing and predicting cardiovascular diseases. It is based on neural networks.
The research team leader Leonid Yasnitsky, professor of the Applied Mathematics and Informatics Department of the Mechanics and Mathematics Faculty of PSNRU reported about the invention.
"Doctors hardly ever use the math modeling method, but it can help you find out the interconnection of many things, as well as eliminate the risks that affect the dangerous prognosis. This will help partially or completely relieve the patient from the necessity of going to the hospital," – Mr. Yasnitsky said.
According to him, the neural network with its software and hardware implementation makes a mathematical model based on the principles of networking nerve cells of a living organism. The neural network analyzes information on a patient’s anamnesis vitae, pulse, blood pressure, and even relatives’ diseases - 69 indicators altogether. By changing one or more input parameters, one can observe the pathogenic pathway. It is called the "freezing method” by mathematicians.
"Such networks do not require constant programming, but they learn and “grow up” themselves. In total, more than a thousand Perm residents participated in the study. Our system collected and analyzed statistical information and successfully diagnosed every case", - said Fyodor Cherepanov, one of the developers. According to him, the initial databases that the system "studied" on were provided by the doctors of the emergency cardiology department of Perm City Clinical Hospital No. 4.
As part of the study, scientists were able to identify a number of patterns. For example, analysis of the neural network has showed that diabetes mellitus has no effect on predisposition to myocardial infarction. The system has established that the increase or decrease in weight is not related to the probability of arrhythmia, slowing or stopping the pulse transmission along the cardiac muscle, i.e. heart block. In addition, the neural network has recorded that the heredity factor increases the risk of infarction by 6.25 percent and hypertension by 7.4 percent.
The invention has already received support from the regional authorities, as well as a grant in the amount of 1 million 350 thousand rubles from the Russian Foundation for Basic Research. According to scientists, the mathematical system of diagnosing diseases has come to be used by doctors of the Perm Medical Academy. The introduction of the neural network system into the medical facilities of the Perm Territory has not yet been announced. "The research is still going on," - the university concluded.