Recently, a large amount of research has suggested that artificial intelligence learning or AI and computer algorithms can be extremely helpful when it comes to the world of medicine. For example, a study appearing some months ago discovered the ability of a deep learning algorithm to make an accurate prediction of Alzheimer’s disease’s onset almost 6 years beforehand.
With the help of what is called a “training dataset,” algorithms of deep learning can train themselves to correctly predict the probability and time of an event being likely to occur. Currently researchers are attempting an examination of whether artificial intelligence or machine learning can be used to make accurate predictions about premature mortality caused by chronic diseases.
Stephen Weng works at the assistant professor post in the data science and epidemiology department of the Nottingham University situated in the U.K. and was the lead of this new research. The research team examined health-related data for greater than half million individuals in the age range of 40-69 years. Participants of the study had registration with UK’s Biobank study in the time period of 2006-2010. These individuals were followed by the researchers of the UK Biobank’s study until 2016.
As for this recent study, Weng along with his team came up with a learning algorithm system with the use of 2 models named “deep learning” and “random forest”. These models were in predicting premature death risk caused by chronic diseases. The scientists tested the accuracy of prediction offered by these models, comparing them with the conventional, existing prediction models, like a Cox model working on multiple variables and the analysis of “Cox regression”.
The lead investigator of the study explained that they plotted the resultant predictions to data on mortality of the cohort. They used the death records from the National Statistics Office, the cancer registry of U.K. and statistics of ‘hospital episodes’. According to the study results, Cox regression was the model with least accuracy at prediction of premature death. Cox model using multiple variables could potentially over-predict the risk of death, thus making machine learning-based algorithms a more viable option.