lungs under xray

AI equals experts in lung disease diagnosis

Image credit: Dreamstime

A team of scientists in Japan has developed an artificial intelligence (AI) model able to identify idiopathic pulmonary fibrosis as accurately as medical experts.

The AI model developed at Nagoya University makes its diagnosis based only on information from non-invasive examinations, including lung images and medical information collected during daily medical care.  

Idiopathic pulmonary fibrosis, a potentially fatal disease that can scar a person’s lungs, is famously difficult to diagnose, particularly in the early stages. 

This AI model, developed in collaboration with RIKEN and Tosei General Hospital, was able to analyse data from the hospital's patients and diagnose the disease with a similar level of accuracy to that of a human specialist, which doctors often have to consult when dealing with pulmonary fibrosis. 

“Idiopathic pulmonary fibrosis has a very poor prognosis among lung diseases,” said Taiki Furukawa, assistant professor of the Nagoya University Hospital. "It has been difficult to diagnose even for general respiratory physicians.

"The diagnostic AI developed in this study would allow any hospital to get a diagnosis equivalent to that of a specialist. For idiopathic pulmonary fibrosis, the developed diagnostic AI is useful as a screening tool and may lead to personalised medicine by collaborating with medical specialists."

The AI suggested that the red area is idiopathic pulmonary fibrosis and blue area is non-idiopathic pulmonary fibrosis.

The AI suggested that the red area is idiopathic pulmonary fibrosis and blue area is non-idiopathic pulmonary fibrosis./ Tosei General Hospital, Reiko Matsushita

Image credit: Tosei General Hospital, Reiko Matsushita

Currently, there are no therapies to cure idiopathic pulmonary fibrosis, although some drugs are able to delay the disease's progression if it is identified early. To do so, patients often have to be subjected to invasive diagnostic techniques, such as lung biopsies.

Unlike current diagnostic techniques, the AI model can identify the disease without the risk of exacerbating it or increasing a patient's risk of dying.  

“The practical application of diagnostic AI and collaborative diagnosis with specialists may lead to a more accurate diagnosis and treatment," Furukawa added. "We expect it to revolutionise medical care.” 

The team's findings were published in the journal Respirology, where the scientists stressed that the AI should be considered a support tool, rather than a substitute for medical specialists.  

Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.

Recent articles