Machine learning is a subset of artificial intelligence, where a computer system performs a specific task by relying on patterns and inferences, rather than a fixed set of instructions. These patterns and inferences are made based on large amounts of data and information that the computer is presented. Therefore, the more data that the computer has, the more information and patterns it will be able to detect. This technology can be incredibly powerful and effective in medicine. For instance, machine learning can be used to integrate each patient’s clinical, personal, and socioeconomic factors to make reliable predictions about disease risk or health outcomes. Before its widespread application, however, comprehensive studies will need to assess whether predictions made by machine learning are truly accurate and reliable.
In a recent study, published in the European Heart Journal, researchers investigated how well machine learning could predict the risk of heart attack and death. To test this, data from 1,912 individuals in the prospective, randomized clinical trial called EISNER was used. Based on the clinical data and the quantitative imagining data available, machine learning was used to predict the risk factor for heart attacks and cardiac death over a 15-year period.
During the 15-year follow-up period, there were 76 cases of heart attack or cardiac-related deaths. The analysis revealed that machine learning had more accurately predicted risk of heart attacks and death when compared to predictions made based on scores and assessments that cardiologists use daily in their clinical practice. More precisely, cardiologists overestimated the risk of heart attack and death, when compared to machine learning.
These results demonstrate the potential value of using machine learning to successfully integrate large amounts of clinical, imaging, and personal data to make accurate predictions regarding disease risk and health. Similar machine learning strategies can be applied to predict a variety of health complications other than heart disease.
Written by Haisam Shah
Reference: Commandeur, F. C., Slomka, P. J., Goeller, M., Chen, X., Cadet, S., Razipour, A., … & Achenbaclh, S. (2019). 30 Machine learning to predict the long-term risk of myocardial infarction and cardiac death based on clinical risk, coronary calcium and epicardial adipose tissue: a prospective study. European Heart Journal, 40(Supplement_1), ehz747-0002.
Image by 200 Degrees from Pixabay