The school magazine PLOS ONE reproduced this week's incredible exploration by a team of scientists led by Stephen Weng (Univ. Of Nottingham), which led the way in t the development of an algorithm to accurately predict the risk of premature death of a patient (who understands prematurely to die before average life expectancy).
Weng, professor in epidemiology and data science, assessed between 2006 and 2016 medical records of 502,628 British patients with ages between 40 and 69 years of age. 14,418 of the participants in the survey died in that decade, in most cases due to respiratory diseases and coronary disease, or cancer.
Using all of these data, researchers could use a false understanding Look for patterns that were common to those who died too soon, with the aim that they'll be in the future allowing reliable estimates of the life expectancy of a living patient.
3 modes, 3 results
For this they studied two types of AI: "in-depth learning" twhere there are neutral networks processing information to learn from the examples; and the "speculative forest" ta more simple type of AI that links several models in a tree format to predict predictions.
They then compared the results that the two modules of AI showed with results common algorithm called 'Cox model';, who risks the risks that need to affect the lives of individuals.
Factors were identified by the 3 ways as would be appropriate when assessing risk of death in advance: age, gender, history of smoking and previous judgment in cancer. However, they were different in other places.
So, the Cox model relied heavily on things such as ethnicity and physical activity… but it just could do t 44% of the dead are clearly identified participants in the survey.
The speculative forestry model was then based on items such as the number of body fat, tone skin or using vegetables, and t up to 64%.
Finally, by trying to learn in depth, research authors identified a pattern which has given greater weight to issues such as alcohol consumption and specific drugs, and emerged to pollution and work accidents. In this way, they correctly identified over three quarters of the fatal death cases (in particular, 76%) t.
Weng says that what caused this research was that it was preventive medical care it grows, with strength growing, "more priority in the fight against serious diseases. "
That's, and most of the technological ways it was in order to predict some areas or diseases in a patient's health center, but t none note the relationship of several diseases or different environmental and individual factors 39 may affect patients.
Joe Kai, another of the researchers involved, says that is currently the ability to AI is likely to be predicted in this field:
“In some cases, it may be useful, in other cases, to not, in this particular case, to show that these algorithms can develop with this. carefully change. "
The next step is to prove and confirm the findings of this inspection with other population groups, and explore ways to apply these systems in everyday medical care.