New findings suggest AI can detect pulmonary hypertension for people with IPF

A type of artificial intelligence (AI) can be used to successfully identify pulmonary hypertension (PH) in people living with idiopathic pulmonary fibrosis (IPF), according to a new study.

PH is a common complication in people with IPF. It can make daily activities harder, increase the need for oxygen, reduce quality of life and worsen long-term health. Currently, non-invasive options for diagnosis can be inaccurate or people are required to have an invasive test called right heart catheterisation.

These new results suggest that AI may one day offer a safer, non-invasive way to identify this complication.

What did the study look at?

Researchers looked at medical records of 130 people living with IPF, including 65 people who had PH and 65 similar people who did not. They used results from heart ultrasound images and lung CT scans to train a deep learning computer programme (a type of AI). They also shared blood test and lung health test results to help the computer learn how to spot which people had PH.

What do the results show?

After the computer system was trained, it could very accurately spot cases of pulmonary hypertension. It correctly identified whether a person had PH in an average of 9 out of 10 cases. This result was seen in the first round of testing and also in a second test to confirm the results.

Why is this important?

Spotting the development of PH for people living with IPF is challenging because the symptoms can be vague and caused by other conditions. Finding cases early can help targeted treatment start sooner.

The results of this study suggested that this AI-based approach could become a non-invasive tool to help lung health specialists find PH earlier on and more reliably. This could improve the diagnosis process and care for people living with IPF.

Larger-scale studies including more people are needed to check how effective this tool can be before it is widely used by lung health specialists.

Read the original research paper: Deep learning networks accurately detect pulmonary hypertension in patients with idiopathic pulmonary fibrosis