Scientists Simplify the Computational Cardiovascular Model to Achieve Clinical Utility

17:07, February 26, 2020

The model is an analytic tool to assist medical practice and clinical trials

An international group of scientists with representatives of the Ural Branch of the Russian Academy of Sciences and the Ural Federal University has developed a new strategy for reducing the computational complexity necessary to build models of heart mechanics. The researchers found how the volume of the ventricle changes during the contractile cycle, depending on the change in the length of one muscle element of the heart. In the case of the latter, calculations are much simpler than for a 3D model of a whole heart camera. This made it possible to use a lower order model for calculating the contractile function of the ventricle and to obtain the result of the desired accuracy with less training program costs.

The study was supported by the Russian Science Foundation. Project results are presented in PLOS One magazine.

According to Olga Solovyeva, Head of the Laboratory of Mathematical Physiology, IIF Ural Branch of the Russian Academy of Sciences and Laboratory for Mathematical Modeling in Physiology and Medicine using Supercomputer Technologies, the approach is based on well-established data on the similarity of contraction of the isolated muscle and the entire ventricle.

“In particular, we demonstrate that simple linear transformations between muscle stretching and reduction in chamber volume are sufficient to reproduce the general output of pressure and volume in 3D finite element models. We are developing a training procedure for the starting finite element model with increasing its accuracy and give an example of a parameter optimization based on medical images,” explains Olga Solovyova.

The description of the work of the ventricle allows you to solve new problems related to blood flow, pressure and other aspects of the heart. Using the proposed approach, it is possible to analyze large amounts of data from real clinical trials and derive general patterns that are applicable in medical practice. Biophysical models that take into account the characteristics of a particular patient can be useful to increase the efficiency of drug selection or for cardiac resynchronization therapy - the installation of a pacemaker, the electrodes of which control the heart and help its chambers to contract in the right rhythm.

During the cardiac cycle, there is a simultaneous contraction of the two atria, then of the ventricles, and then relaxation. A healthy organ works in a certain rhythm, but almost all of its diseases lead to malfunctions. Because of this, our internal pump loses its ability to pump the blood through veins in the right volume, and chronic heart failure gets developed. It is accompanied by constant fatigue, shortness of breath, swelling of the legs and impaired functioning of the organs - lungs, liver, kidneys, since the body does not receive the necessary amount of oxygen. Blood stasis begins. Despite the wide range of existing treatments, chronic heart failure remains one of the most common death causes worldwide. In Russia, every minute an average of one person dies from it.

A more detailed description of the normal functioning of the heart and with lesions helps scientists increase the effectiveness of existing methods of treatment and develop new ones. One approach - intelligent modeling - allows you to create numerical models that describe the properties and work of the heart. These calculations can be used as an analytical tool applicable in medical practice and in clinical trials.

First, calculations are performed on detailed models of the heart with a large number of variable parameters of the heart muscle (myocardium), such as conduction and contractility. Scientists obtain a so called 'virtual clinical data', which can be compared with the actual results of clinical trials - ECG, ultrasound scan and others. Among them, those that fall into the ranges of observed indicators for a specific group of patients are selected. Then, using machine learning, the correlation between the results of the model and its parameters are found, thus obtaining values ​​that are closest to the patients. Having such correlations, it is possible to assess the state of the heart of a particular patient. However, existing models require complicated calculations and are not applicable for everyday use, therefore, the researchers are simplifying them.

The research was conducted in collaboration with the colleagues from the IBM Thomas Watson Research Center (Yorktown Heights, NY).