[ad_1]
Computers capable of learning by themselves to predict premature deaths could dramatically improve preventive health care in the future, suggests a new study by experts at the University of Nottingham.
The team of scientists and physicians in the health sector has developed and tested a system of computer-based algorithms for machine learning to predict the risk of premature death from chronic disease. within a large population of middle age.
They discovered that this AI system was very accurate in its predictions and gave better results than the current standard forecasting approach developed by human experts. The study is published by PLOS ONE in a special edition of Machine Learning in Health and Biomedicine, devoted to collections.
The team used the health data of just over half a million people aged 40 to 69 recruited from the UK Biobank between 2006 and 2010 and monitored until 2016.
Dr. Stephen Weng, Assistant Professor of Epidemiology and Data Science, said, "Preventive health care is a growing priority in the fight against serious diseases, so we have been working for several years to improve the accuracy of Computerized badessment of health risks In the general population, most applications focus on a single disease area, but the prediction of death due to a number of different diseases is extremely complex, due in part to environmental and individual issues that may affect it.
"We have taken a big step forward in this area by developing a unique and comprehensive approach to predicting the risk of premature death of a person through machine learning." This method uses computers to create new ones. Risk prediction models that take into account a wide range of demographic, biometric, clinical and lifestyle factors for each individual badessed, even their dietary intake of fruits, vegetables and meat per day.
"We mapped the resulting forecasts with mortality data from the cohort, using the Office of National Statistics death records, the UK cancer registry, and" hospital episode "statistics. We found that machine-learned algorithms were much more accurate in predicting deaths than standard prediction models developed by a human expert. "
The AI learning models used in the new study are known as "random forest" and "deep learning". These were opposed to the Cox regression model, traditionally used, based on age and bad – judged to be the least accurate predictor of mortality – and also on a Cox multivariate model that worked better but tended to overestimate the risk.
Professor Joe Kai, one of the clinical academics working on the project, said: "There is currently a keen interest in the potential of using" AI "or" l ". In some situations, this may help In other cases, we have shown that with careful tuning, these algorithms could usefully improve the prediction.
"These techniques may be new to many in health research and difficult to follow, and we believe that clearly reporting these methods in a transparent manner could help in the scientific audit and future development of this exciting area for health care. . "
This new study builds on previous work by the Nottingham team that showed that four different IA algorithms, "random forest," "logistic regression," "gradient amplification," and "neural networks," were significantly better at predicting cardiovascular disease than an established algorithm. used in current guidelines in cardiology. This previous study is available here.
The Nottingham researchers believe that AI will play a vital role in developing future tools capable of providing personalized medicine, tailoring risk management to each patient. Further research requires checking and validating these AI algorithms in other population groups and exploring ways to implement these systems in routine health care.
Source of the story:
Material provided by University of Nottingham. Note: Content can be changed for style and length.
Source link
