Computer analysis of the severity of heart disease

Artificial intelligence predicted the risk of death from heart failure

Kristina Ulasovich, N+1

Scientists have created a system that has learned to predict the risk of death for patients with cardiovascular diseases with high accuracy. Testing of its work on patients with pulmonary hypertension showed that the accuracy of predictions reaches 73 percent. The researchers' work is published in the journal Radiology (Dawes et al., Machine Learning of Three-dimensional Right Ventricular Motion Enables Outcome Prediction in Pulmonary Hypertension: A Cardiac MR Imaging Study).

Pulmonary hypertension is characterized by increased arterial pressure in the pulmonary artery and increased load on the right ventricle of the heart. As a result, this leads to the development of heart failure and premature death. In order to determine whether patients are at high risk, as well as prescribe treatment, doctors conduct complex diagnostics: they do magnetic resonance imaging of the heart, electrocardiography, ultrasound of the heart, lung X-rays and other procedures. However, they are not always able to accurately predict how high the probability of deterioration of patients' condition is.

The authors of the new work proposed using artificial intelligence in order to estimate the possible life expectancy of patients with pulmonary hypertension. Scientists have created a system that, after preliminary preparation, was trained to analyze heart scans and predict the likelihood of death of patients. As a training sample, the researchers used heart MRI data for 224 patients. Based on them, the program built a three-dimensional model of the organ and measured the movement of 30 thousand individual points during each heartbeat. Then the artificial intelligence combined the results obtained with medical records of patients' health for eight years and made a forecast.

The scientists tested the program on 32 patients. The computer was able to determine those who will live for more than a year, with an accuracy of 73 percent (during the study, 93 of 256 patients died). For doctors who relied on heart images, hemodynamic data and clinical markers, the accuracy of predictions was only 60 percent. At the same time, the authors of the work state that artificial intelligence can make a forecast for five years ahead.

At the moment, scientists whose work took about four years in total, plan to test the system on other patients in different hospitals. In the future, the program may probably be used to diagnose other types of cardiovascular diseases that lead to heart failure, such as for cardiomyopathy.

Recently, more and more new methods of recognizing diseases have been appearing. In December 2016, American scientists also used image analysis and machine learning to conduct an early diagnosis of melanoma. A team of scientists from Sweden and the USA, in turn, recently presented a nozzle for a Nokia smartphone in the form of a portable fluorescent microscope. Using the new system, the researchers analyzed DNA samples and identified mutations associated with certain types of cancer.

Portal "Eternal youth" http://vechnayamolodost.ru  20.01.2017