Perm neural networks diagnose and forecast infarctions
29 May '17
Scientists at the Perm State University (PSNRU) in the West Urals have developed an intelligent cardiovascular diseases diagnostics and prognostication system built on advanced artificial neural networks (ANNs), the Russian news agency TASS reported, citing Prof. Leonid Yasnitskiy of PSNRU’s applied mathematics and informatics chair, the research leader in the team.
The Perm-originating ANN is said to be able to analyze information on all the diseases a patient and his relatives have had in their life, as well as on heartbeat, blood pressure, etc.—69 parameters all told.
“Such networks don’t have to be constantly re-programmed as they learn and grow all on their own. More than a thousand Perm residents have taken part in the research. Our system gathered and screened stats and has successfully produced a diagnosis in each of the cases,” said Fyodor Cherepanov, one of the developers. He also said that initial databases, which the system “used” to learn, came from cardiac physicians at Perm City Hospital 4.
Some consistent patterns have reportedly been identified in the research. For example, the ANNs showed that suffering from diabetes has no impact on myocardial infarction predisposition; gaining or losing weight is not tied to developing arrhythmia; slowing or stopping pulse transmission in the cardiac muscle does not necessarily lead to heart block. The networks also revealed that the hereditary factor increases risk of infarction development by 6.25% and hypertension by 7.4%.
The project has been supported by regional authorities; the Russian Fundamental Research Fund has also backed the effort with a $24,000 grant. The math-based diagnostics system is reported to have already been used by physicians at the Perm Medical Academy.