NHS doctors want ChatGPT AI to write patient heart reports so they can see more people

NHS heart experts want to use artificial intelligence program ChatGPT to write vital patient reports – because they say it would free up time and allow them to see more people.

The free-to-use software is already being used by students and office-workers to carry out some tasks.

And, according to Dr Samer Alabed, a cardiac radiologist at Sheffield Teaching Hospitals NHS Foundation Trust, the AI program could be trained to interpret and write notes on heart scans.

‘It takes radiologists 45 minutes to analyse these scans and then write a report,’ says Dr Alabed. ‘We estimate that NHS clinicians spend 115,000 hours putting heart scan reports together each year.

‘If we were able to use software like ChatGPT to do this work, we could free up an incredible amount of time which could be used to treat more patients.’

Doctors say it would free up time and allow them to see more people

The software may also be able to do this in non-medical language so patients could understand their results without the help of a doctor.

Dr Alabed and his team are already experimenting with the program and hope to soon set up a clinical trial involving a large group of heart disease patients. The project is viewed as the next step in the increasing use of artificial intelligence in the NHS.

In December, the MoS revealed that clinicians at Sheffield Teaching Hospitals had begun using a specially designed AI program that could analyse MRI scans of the heart and carry out complicated measurements to work out if there were any signs of disease.

These measurements usually take doctors 20 minutes to complete but the program is able to complete them in under a minute. However, doctors still have to write up their findings.

Dr Alabed believes that the entire process could be carried out entirely by AI. However, he adds that the team are still encountering problems with ChatGPT which need to be addressed before the software can be used on the NHS.

‘When we input measurements into ChatGPT we’re finding that it sometimes adds in extra details which are made-up or not accurate,’ he says.

‘The next step is to work out how to train it to be accurate every time, so patients get the correct diagnosis.’