A neural network model for early diagnosis of chronic heart failure 

Vladimir I. Gorbachenko, Doctor of engineering sciences, professor, head of the sub-departmnet of computer technologies, Penza State University (40 Krasnaya street, Penza, Russia), E-mail: gorvi@mail.ru
Vladimir V. Potapov, Candidate of medical sciences, associate professor, assistant of the sub-department of anesthesiology, Donetsk National Medical University named after of M. Gorky (16 Ilyicha avenue, Donetsk), E-mail: x3x3x23@rambler.ru
Oleg K. Zenin, Doctor of medical sciences, professor, professor of the sub-department of human anatomy, Medical Institute, Penza State University (40 Krasnaya street, Penza, Russia), E-mail: zen.olegz@gmail.com
Ilia S. Miltykh, Student, Medical Institute, Penza State University (40 Krasnaya street, Penza, Russia), E-mail: ilia@miltykh.com
Dmitriy N. Gribkov, Student, Penza State University (40 Krasnaya street, Penza, Russia), E-mail: rororol23@mail.ru 

Background. The WHO reports that cardiovascular diseases are the leading cause of disability and mortality. The high prevalence as well as the severity of chronic heart failure requires the development of new methods of early diagnosis and treatment monitoring. Early diagnosis reduces the number of patients that require inpatient treatment, reduces the number of days of disability, and also reduces the probability of adverse outcomes. Materials and methods. Human bodily fluids include surfactants that can change surface tension, accelerate and decelerate the transport of substances and gases. Their amount in human biological fluids may change due to various pathologies, which may be an early marker of the disease. Tensiometry methods have been suggested for early diagnosis of CHF. To select an objective criterion for diagnosing a pathological condition, neural network modeling was applied. Training data for the neural model was obtained using a tensiometer. Results. Neural network allows to diagnose chronic heart failure based on blood tensiometry results with high accuracy. The neural network allows to diagnose CHF at early stages with 98% accuracy. Conclusions. Tensiometric data of patients with CHF and healthy people are a key point in the development of new diagnostic approaches. Analysis of changes in tensiometric parameters of blood plasma and serum with subsequent neural network processing allows an early diagnosis of chronic heart failure with high accuracy. 

heart failure, tensiometry, neural network model training, neural network model quality metrics 

Gorbachenko V.I., Potapov V.V., Zenin O.K., Miltykh I.S., Gribkov D.N. A neural network model for early diagnosis of chronic heart failure. Izvestiya vysshikh uchebnykh zavedeniy. Povolzhskiy region. Meditsinskie nauki = University proceedings. Volga region. Medical sciences. 2022;(4):5–15. (In Russ.). doi:10.21685/2072-3032-2022-4-1