Sunday , May 26 2019
Home / argentina / They created an algorithm that says when you will die and they say "it's more effective than a doctor"

They created an algorithm that says when you will die and they say "it's more effective than a doctor"



The development of robots and automated systems is developed in the work world and also shows creative abilities that exceed the only repetitive tasks. Will technical development also succeed in mimicking the effectiveness of healthcare professionals?

New research concludes that trained algorithms are more effective than doctors to predict mortality. But in any case, these advances will not replace staff specialists, but complement their tasks.

From entertainment to health

Machine learning is capable of predicting episodes such as heart attacks and even death. This is one of the most important conclusions of a study led by Dr. Luis Eduardo Juárez-Orosco from the Turku PET Center in Finland, presented at the International Conference on Nuclear Cardiology and Cardiac CT.

One of the most interesting and striking axes in the experiment is the use of LogitBoost, a algorithm similar to that used by some services streaming to recommend movies, series, music, etc., related to users' tastes.

In this case, the algorithm repeatedly analyzed 85 variables in 950 patients over six years. With all that information, managed to identify patterns that serve to predict various episodes of health, illness and death.

According to the researchers, this mechanism had one 90% accuracy in your warnings. A percentage that is much higher than that achieved by human doctors.

"These advances go far beyond what has been done in medicine, where we must be careful about how we evaluate risks and results. We have data, but we do not use them fully," says Juárez-Orozco.

The specialist explained that doctors use risk points (such as those thrown by this system) to recommend treatments. In that sentence, The key to this technology is the very high level of adaptation, it approaches each patient's characteristics, in addition to the large number of variables it analyzes repeatedly.

As the study leader pointed out, "The algorithm gradually learns from data and after many analysis rounds, the high-dimensional patterns that should be used to effectively identify patients who have the event determine. "That way, it delivers individual risk points.


Source link