• Researchers have developed a computer program using artificial intelligence (AI) to look for suspicious areas that may be cancer on lung computed tomography (CT) scans.
• The AI program looked at a set of 1,179 lung CT scans and correctly identified 97 percent of suspicious areas that were later confirmed to be cancer.
• The program has the potential to help improve lung cancer screening and find tumors up to one year earlier.

A new study from French researchers has found that an AI program can find lung cancer on CT scans up to a year before it is diagnosed. The researchers hope that this program can be used to make screening for lung cancer quicker and easier to help diagnose people earlier than before. The findings were presented recently at the European Respiratory Society International Congress.

“This work is promising because it shows that AI could help us to review lots of scans quickly and even pick up signs of cancer at an earlier stage. However, before this program can be used, researchers will need to make it better at distinguishing between lung tissue that is abnormal but benign and tissue that is probably cancer,” said Professor Joanna Chorostowska-Wynimko, a consultant in respiratory medicine at the National Institute of Tuberculosis and Lung Diseases in Warsaw, Poland, in a press release.

In the United States, lung cancer is the second-most common cancer in both men and women, according to the American Lung Cancer Association, and it accounts for the most cancer-related deaths each year. Although there are some screening methods, many cases of lung cancer are diagnosed in later stages. By this time, the cancer may have spread and can be difficult to treat.

The U.S. Centers for Disease Control and Prevention (CDC) recommends a low-dose CT scan for lung cancer screening. This method uses low doses of X-rays to take pictures of the lungs to look for tumors or areas of concern in the lungs. However, these images have to be analyzed by a radiologist who looks at hundreds of images from the scan to find signs of cancer.

Enter Artificial Intelligence

For the study, researchers at France’s National Institute for Research in Digital Science and Technology (called the Inria) developed and trained a computer program that uses AI to look at sets of CT scans for signs of lung cancer. The program first looked at scans from 888 participants who’d been examined by radiologists to learn what was cancer and what was not.

Researchers then tested their program with another set of 1,179 participants from a lung screening trial. In this second set, 177 of the participants were biopsied and confirmed to have lung cancer.

The AI program correctly found 172 of the 177 cancer cases, giving it a 97 percent accuracy rate. The five cases that it did not find involved tumors near the middle of the chest, which are difficult to tell apart from healthy tissues. The program also looked at scans that were taken a year before and correctly found 152 areas that were diagnosed later as cancer.

A Work in Progress

While the model shows promise, it still needs some work. Sometimes, it identifies tumors that are not there, giving false positives that could lead to a misdiagnosis and unnecessary testing. These issues will need to be resolved before the program is used in a clinical setting.

However, the researchers are excited about the promise it shows and the help it can give to radiologists. “Screening for lung cancer would mean many more CT scans being taken, and we do not have enough radiologists to review them all,” said Benoit Audelan, a researcher at Inria who worked on the study, in a press release. “That’s why we need to develop computer programs that can help. Our study shows that this program can find possible signs of lung cancer up to a year earlier.”

Audelan and his team from the Inria worked alongside colleagues from Université Côte d’Azur, software company Therapixel, and the University Hospital of Nice.