ARTIFICIAL NEURAL NETWORKS TO ESTIMATE VENTRICULAR ASSIST DEVICE'S SPEED
Marcelo Barboza Silva1; Tarcísio Leão1; Igor Munhoz2; Evandro Drigo3; Paulo Barbosa1; Bruno Silva3; Diolino Santos Filho2; Edinei Legaspe2; Aron Andrade3; Jeison Fonseca3; EDUARDO BOCK1
1 Department of Mechanics - São Paulo Federal Institute of Education, Science and Technology; 2 Department of Mechatronics and Mechanical Systems Engineering - University of São Paulo; 3 Institute Dante Pazzanese of Cardiology
doi:10.20906/CPS/COB-2015-1344
Resumo
In 2014 was reach the record of heart transplantation in Brazil, more than 300 patients received the organ and are no longer in waiting lines. Nevertheless, 259 people were waiting for an available heart yet. According with studies, 25% of components of that list will die before receive the organ. However in all these cases, the heart still works partially, what show that maybe transplant is not mandatory. In some cases, a machine that helps the natural heart to pump is enough to patient survive until the surgery and/or his recover. Based on this principle, a Ventricular Assist Device (VAD) was design. It operates together with natural heart decreasing the mechanical work of myocardia required to pump blood. A prominent research field is control applied in artificial organs. The natural operation of human's organs has many physiological processes to control their functions, including some unknown processes. So the focus of this work is to control the motor speed of a VAD, the Implantable Centrifugal Blood Pump (ICBP), through a control similar to natural control. To determine heart rate (HR), our body already takes into account all the pressure, movements, emotions, and others. If we consider the HR as controller input we will be considering various inputs. For VAD study, each person has different parameter variations in cardiovascular system. So believe that artificial intelligence can fulfill requirements of an artificial control close of natural control. Stenosis in aortic valve was reported in patients supported with VADs, because continuous blood pumps bypass that valve. So a control strategy has to open the aortic valve sometimes to avoid stenosis. During methodology was proposed a routine that analyses the electrocardiogram in real time and through an artificial neural network (ANN), called ANN_QRS, estimates complex QRS moment. A second ANN, called ANN_RPM, estimates speed of VAD´s motor. And a third ANN, called ANN_SIM, was developed to simulate a cardiac system with VAD assistance to validate all ro
Palavras-chave: Implantable Centrifugal Pump Blood; Ventricular Assist Device; Artificial Neural Networks; Artificial Organs; Heart Rate