Artificial intelligence can predict death threats
The technology of artificial intelligence or Artificial Intelligence (AI) has long been discussed in the technology world. There is debate about whether the proliferation of artificial intelligence will bring relief to the world in the coming days, or whether it will increase suffering.
Various studies have been going on for a long time to improve the subject. However, researchers now heard of another ability of artificial intelligence. They say artificial intelligence can predict when people can die. In this case, modern technology capable of producing better results than the physician.
Researchers in the United States are claiming that in many cases, physicians treat patients as normal, with the use of artificial intelligence that may produce different results. In this case, it is possible to predict when the patient will die by analyzing the result of one year of cardiac examination. However, how artificial intelligence can predict this is still mysterious.
A group of researchers, led by researcher Brendan Fronwalt, at the University of Pennsylvania-based health care organization, conducted research on AI. They try out the results of the test of 1 lakh and 4,000 electrocardiogram (ECG) of 5 lakh people and try to find out who are most at risk by next year.
The researchers created two versions of artificial intelligence to test them. To give the AI of a version just the primary ECG information, which measures the voltage over time. Age and gender information is given along with ECG information in other versions of AI. Researchers determine the performance of artificial intelligence using a method called 'AUC'. It expects who is likely to die within a year and who survives.
Researchers claim that in their model, the death toll is far from perfect. Many times, cardiologists cannot even make such a prediction. Three cardiologists consider the results of the ECG test to be normal, considered AI risk.
Researcher Fronwalt claims that experiments with artificial intelligence have shown that humans cannot see what the model does. What we neglect as normal, can appear as risky in this test. However, it is not yet known whether AI identifies cardiovascular risk factors at a particular pattern. That is why many doctors are not yet ready to apply such a procedure.
The dissertation article will be presented at the American Heart Association's Scientific Session in Dallas, USA tomorrow.
Researcher Christopher Haggerty says the study is based on historical data. It is important to examine how effective such algorithms are for the patient.