Using AI to predict heart failure

Most illnesses and diseases are treated much more effectively if we can catch them at an early stage. The last year have seen a number of fascinating projects that highlight the growing power of AI to predict our health, and therefore give us that early warning system.

For instance, we’ve seen projects that have been able to predict diseases such as Alzheimers and autism, anemia or even the common cold.  You’ve also got projects such as the Deloitte X-Prize that aims to do similar things for cancer.

The latest effort along these lines is from the UK’s Medical Research Council, and uses AI to predict when patients with a particular heart disorder may die.  The project, documented in a recently published paper, saw an algorithm trained to examine blood tests and heart scans to hunt for signs that it was about to fail.

Supporting the heart

The study was particularly aimed at patients with pulmonary hypertension, whereby high blood pressure causes damage to the heart.  The condition is terminal for around a third of patients, who die within five years of diagnosis.

Current treatments usually involve drugs and/or a lung transplant, but they still make it difficult for doctors to really understand how long a patient might have to live.  Better knowledge of this can help guide doctors towards the appropriate treatment.

The researchers used MRI scans of over 250 patients hearts alongside their blood tests results.  The algorithm then monitored the movement of the heart through 30,000 different points.

Alongside some eight years worth of patient records, the algorithm was capable of detecting the kind of abnormalities that would enable them to accurately predict when the patient might die from their condition.  What’s more, it was able to look approximately five years into the future with its predictions.

As with other algorithm based services, the software was accurate in its predictions around 80% of the time, which compares favorably to 60% for doctors in the study.

Personalized medicine

It’s the kind of predictive capabilities that allow doctors to provide more personalized treatment to each patient based upon their particular stage of the illness.

“So we can tailor getting absolutely the right intensive treatment to those who will benefit the most,” the researchers say.

The next step is to test the system out on a wider range of patients before then hopefully rolling it out as an official service to doctors.  There are also plans to test it on other forms of heart issues, such as cardiomyopathy.


Leave a Reply

Your email address will not be published. Required fields are marked *