Computers able to teach themselves to predict a strategic death can significantly improve future preventive healthcare, reflecting a new survey by specialists at Nottingham University.
The team of health data experts and doctors has developed and tested a system of algorithms & computer learning # 39; computer-based to predict the risk of early death as a result of long-term illness in a large, medium-aged population.
They found that this AI system was accurate in its forecasts and performed better than the usual way of producing prediction by human experts. The review is published PLOS ONE preparation of special collections of "Machine Learning in Health and Biomedicine."
The team used health data from just over half a million people between 40 and 69 who were recruited to the UK Biobank between 2006 and 2010 and beyond to 2016.
Dr Stephen Weng, the project's pioneer of Psychology and Data Science, said: "Healthcare is an important priority against the fight against serious diseases and so on. T we have been working on several years to improve a computerized health risk assessment process but the majority of applications focus on a range of single disease diseases but premature deaths are the result of a number of t different factors, especially because of environmental and individual factors that may affect them.
“We have taken a big step forward in this area by developing a more unique and holistic approach to predicting the possible risk of premature death through learning. This uses computers to construct new risk forecasting models which will take account of a wide range of demographic people; biotrophic, clinical and lifestyle reasons for every person assessed, even to eat the food of fruit, vegetables and meat each day.
"We submitted the following data from death data from the organization, using the National Statistics Office death records, UK cancer statistics statistics, and hospital statistics and a human expert."
The AI modules used in the new survey are referred to as “random woodland” and “deep education”. These were opposed to the “Cox regression” intelligence model which was traditionally based on age and gender – which was confirmed as the least popular at predictability – and also a multi-faceted module. Cox worked better but thought that it was a risk of prediction.
Dr Joe Kai, one of the clinical scholars who is working on the project, said: "This is currently a great interest in the ability to use; In this particular case, we have shown that these algorithms can easily improve forecasting by careful handling.
“These approaches can be new to many in health research, and are difficult to follow. We believe that by communicating clearly on these ways in a clear way, this would help with scientific analysis and future development of health care. "
This new study builds on the work of the previous Nottingham team who showed that there were four different AI algorithms: 'random woodland', 'logistic regression', & # 39; raising the gradient; use of nuclear networks, better to predict the use of cardiovascular disease of the established algorithm to guide current cardiology. This earlier study is available here.
The Nottingham inspectors conclude that AI will play a vital role in the development of tools that are capable of delivering personal treatment, managing risk for individual patients. Further research seeks to test and prove these algorithms in other population groups and to explore ways of putting these systems into conventional healthcare.
Story source: t
Materials donated by University of Nottingham. Note: Content may be edited for style and length.