‘Virtual 3D heart’ learns to predict heart failure
17 January 2017
Computer software creates a 3D virtual heart from MRI scans and then learns how to predict when patients will die to help treat them effectively.
Computers have learned to predict survival in patients with a heart condition more accurately than is possible today. In the future, this may help doctors to identify patients at greatest risk of death earlier and treat them more effectively.
“This is the first time computers have interpreted heart scans to accurately predict how long patients will live. It could transform the way doctors treat heart patients”, says Declan O’Regan, of the MRC London Institute of Medical Sciences (LMS) and who led the research, published in the journal Radiology.
"The computer is up to 80 percent accurate at predicting survival at one year. A doctor equipped with this new cardiac imaging approach would therefore be able to make more informed judgements about outcome than if they were relying only on current ways to investigate patient data.”
The machine learning, or artificial intelligence software automatically analyses moving images of a patient’s heart captured during an MRI scan. It then uses advanced image processing to build a ‘virtual 3D heart’, which replicates the way over 30,000 points in the heart contract during each beat.
The researchers fed the system historic data from hundreds of previous patients. By linking this data with its models, it learned which attributes of a heart, its shape and structure, put an individual at a given risk of heart failure.
The software was developed using data from over 250 patients with pulmonary hypertension, a condition in which high pressure in the blood vessels that supply the lungs puts strain on the right side of the heart. Over time, this causes progressive damage and can ultimately lead to heart failure.
It affects around 7,000 people in the UK and up to a third of patients die within five years of diagnosis.
“The computer performs the analysis in seconds and simultaneously interprets data from imaging, blood tests and other investigations without any human intervention. It could help doctors to give the right treatments to the right patients, at the right time,” says Tim Dawes, of the LMS, who developed the algorithms that underpin the software.
Until now, radiologists have relied on taking time-consuming measurements by hand and using their experience to identify patients at greatest risk of deteriorating. Innovations in machine learning offer a powerful new approach to make automated predictions, using more detailed and complex data.
“We would like to develop the technology so it can be used in many heart conditions to complement how doctors interpret the results of medical tests. The goal is to see if better predictions can guide treatment to help people to live longer,” says Dawes.
Machine learning is already helping research into cancer and brain diseases, but the analysis of moving images of the heart has been more challenging. This is the first study to use the approach in heart disease. The work has been made possible by collaborations between clinicians at the LMS and computer scientists at Imperial College London.
The next step will be to test the software on data from a different hospital to the one in which it was developed (Hammersmith Hospital, London). The ultimate goal is to develop software to make predictions not only about survival, but also about which type of treatment will work best in each patient.
More information can be found on the Imperial College London website.
Video courtesy of MRC London Institute of Medical Sciences LMS.