AI programme can predict when heart patients would die with 80 per cent accuracy

Scientists say they had developed an artificial intelligence (AI) programme that can predict when patients with a serious heart disorder would die, with an accuracy rate of 80 per cent.

According to researchers from the MRC London Institute of Medical Sciences (LMS), the software will allow doctors to better treat patients with pulmonary hypertension by determining how aggressive their treatment needed to be.

In pulmonary hypertension, a rare blood vessel disorder, pulmonary arteries of the heart harden, making them less efficient at pumping blood to the lungs. Around 7,000 people in the UK suffer from the condition, which if left untreated could cause fatal heart failure.

Pulmonary hypertension is an incurable condition that worsens with time, with a third of patients dying within five years of being diagnosed.

This meant that assessing how advanced the condition was in a patient was vital for selecting the right treatment.

The researchers' programme assessed the outlook of 250 patients on the basis of blood test results and MRI scans of their hearts and then used the data to create a virtual 3D heart of each patient which, combined with the health records of "hundreds" of previous patients, allowed it to learn which characteristics indicated fatal heart failure within five years.

On the basis of the data, the software could then predict patient mortality rates up to five years in the future. The system accurately predicted which patients would be alive in one year with about 80 per cent accuracy, while earlier human predictions using the same dataset had only achieved an accuracy rate of 60 per cent.

Speaking to BBC News, Dr Declan O'Regan claimed the system was designed to help doctors target their treatment programmes for specific patients, "So we can tailor getting absolutely the right intensive treatment to those who will benefit the most."