قالب وردپرس درنا توس
Home / Health / AI is good (maybe too good) in predicting who will die prematurely

AI is good (maybe too good) in predicting who will die prematurely



  AI is good (maybe too good) in predicting who will die prematurely

Can AI predict when you will die?

Credit: Shutterstock

Medical researchers have released a troubling ability in artificial intelligence (AI): predicting a person's early death.

Scientists recently trained an AI system to evaluate a decade of general health data submitted by more than half a million people in the UK. They then asked the AI ​​to predict if there was a risk of people dying prematurely, that is, earlier than the average life expectancy, of chronic disease, they said in a new study.

The predictions about the early death of AI algorithms were "significantly more accurate" than predictions of a model that did not use machine learning, senior study author Dr. Ing. Stephen Weng, Assistant Professor of Epidemiology and Data Science at the University of Nottingham (UN) in the United Kingdom. said in a statement. [Can Machines Be Creative? Meet 9 AI ̵

6;Artists’]

To assess the likelihood of premature mortality of subjects, researchers tested two types of AI: deep learning, in which layered information processing networks help a computer learn from examples; and Random Forest, a simpler type of AI that combines several tree-like models to account for possible outcomes.

They then compared the conclusions of the AI ​​models with the results of a standard algorithm called the Cox model. 19659005] Using these three models, the researchers evaluated data from the UK Biobank – a freely accessible database of genetic, physical and health data – submitted by more than 500,000 people between 2006 and 2016. During this time, nearly 14,500 of the participants died primarily from cancer, heart disease and respiratory diseases.

All three models determined that factors such as age, gender, smoking, and previous cancer diagnosis were the most important variables for assessing the likelihood of a person's early death. However, the models diverged over other key factors, the researchers found.

The Cox model relied heavily on ethnicity and physical activity, although this was not the case with machine learning models. In comparison, the forest model for random animals emphasized the body fat percentage, the waist circumference, the amount of fruit and vegetables that people ate, and the skin tone, according to the study. Key elements of the deep learning model included occupational hazards and air pollution, alcohol use and the use of certain medications.

After all the number crunching, the deep learning algorithm provided the most accurate predictions. 76 percent of the subjects who died during their studies were identified correctly. In comparison, the Random Forest model correctly predicted about 64 percent of premature deaths, while the Cox model identified only 44 percent.

This is not the first time that experts have used AI's predictive power for health care. In 2017, another team of researchers showed that AI could learn to detect early signs of Alzheimer's disease. In their algorithm, brain scans were evaluated to predict whether a person is likely to develop Alzheimer's disease, with an accuracy of about 84 percent, Live Science previously reported.

Another study found that AI could predict the onset of autism in 6 months-old babies who had a high risk of developing the disorder. Another study detected signs of diabetes by analyzing retinal scans. and even more – even using data derived from retinal scans – predicted the likelihood of a patient suffering a heart attack or stroke.

In the new study, scientists have shown that machine learning can be used "with careful tuning" According to a successful prediction of mortality outcomes over time, co-author of the study, Joe Kai, a UN professor for Primary care, in the statement.

Using AI, this route may be unfamiliar to many healthcare professionals as it presents the methods used in the study "could help with the scientific verification and future development of this exciting field," said Kai.

The results were published today (March 27) online in the journal PLOS ONE.

Originally published on Live Science Document.addEventListener ("DOMContentLoaded", function () {if (document.getElementById ("comments")) {var listener = function () {var rect = document. getElementById ("comments"). if (rect.top <window.innerHeigh t) {loadAPI (); window.removeEventListener ("scroll", Listener)}}; window.addEventListener ("scroll", listener)}}; Function loadAPI () {var js = document.createElement ("script"); js.src = "http://connect.facebook.net/en_US/sdk.js#xfbml=1&appId=131734303545872&version=v2.4"; document. body.appendChild (js)}
var Purch = Purch || {}; Purchasequeue = Purch.queue || []; Purch.queue.push ([[“jquery”,”Purch/UI/Poll”] Function ($, Poll) {$ ("[data-poll]"). Each (function () {var e = this; var oconf = $ (this) .data ("poll") ? $ (this) .data ("poll"): {}; oconf [“element”] = e; var poll = new Poll (oconf)})}]);

Source link