As scientists strive to create machine learning algorithms that can save us all
Artificial intelligence researchers have discovered that computers are beyond the capabilities of human physicians in many important areas. We have already seen the ability of artificial intelligence to detect phenomena such as cancer, and a new study reveals that a digital brain can also better predict overall mortality and specific conditions such as a heart attack with precision superior to that of a trained individual.
The research, which was presented to the International Conference on Nuclear Cardiology and Cardiac Computed Tomography suggests that we may be fast approaching the day when artificial intelligence works hand in hand with healthcare professionals to anticipate life-threatening issues.
The researchers, led by Dr. Luis Eduardo Juarez-Orozco of the Turku PET Center in Finland, have formed a machine learning algorithm on a set of data from nearly 1 000 patients. The data, which covered six years for each patient, included dozens of variables that the computer needed to digest to correlate deaths and heart attacks with data from various heart and blood flow readings.
learns from the data and after many analyzes, it identifies the high dimensional patterns that should be used to effectively identify patients presenting the event, "said Dr. Juarez-Orozoc in a statement . "The result is an individual risk score."
With each precision taken into account, the predictive accuracy of the AI makes it possible to anticipate a cardiac event or a considerably increased death. Once the system has analyzed all available data, it has managed a predictive score of about 90%, which is significantly higher than most physicians are able to rate based on the typical amount of information that they have on each patient.
to collect a lot of information about patients, for example those who suffer from chest pain, "said Dr Juarez-Orozco" We have found that machine learning can integrate this data and accurately predict individual risks. This should allow us to personalize the treatment and, ultimately, to achieve better results for patients. "
This article appeared first (in English) on BGR