Researchers at Geisinger, a medical care centre based in Danville, Pennsylvania, found that a computer algorithm developed using a video heart echocardiogram could predict deaths within the previous year.
Machine learning algorithms, which are part of artificial intelligence (AI) technology, outperform other tools clinically in predicting mortality risk, including the pooled cohort equations and the Seattle Heart Failure score.
Chris Haggerty, assistant professor in the Department of Data Science and Translational Informatics at Geisinger, says machine learning can take advantage of unstructured data sets such as medical images and videos that are useful for improving various clinical prediction models.
Such imagery is essential for decision-making in medical medicine and has become the most data-rich component of the electronic health record. For example, says Haggerty, one ultrasound of the heart produces about 3,000 images and cardiologists have limited time to interpret these images in the context of many other diagnostic data.
The researchers used specialized computing hardware to train machine learning models on 812,278 video echocardiograms collected from 34,362 patients over the past 10 years. The results will be analyzed by a specialist before making further decisions.
“We are pleased that our algorithm can help cardiologists improve their predictions about patients because decisions about treatment and interventions are based on this type of clinical prediction,” said Alvaro Ulloa Cerna, senior data scientist in the Department of Translational Data Science and Informatics at Geisinger.
Thus, machine learning algorithms that are part of AI technology have increasingly useful benefits, including for predicting very crucial things such as heart health in humans.