Groundbreaking AI can spot lung cancer a year early – with up to 97 percent accuracy

Artificial intelligence could become the key to lifesaving early lung cancer diagnoses, as new research claims the technology can catch early warning signs up to a year early.

Researchers at a university in Nice, France, claim to have developed an artificial intelligence (AI) tool that can help doctors and radiologists diagnose lung cancer a year earlier than typical methods.

Currently, spotting lung cancer typically involves a human radiologist manually examining around 300 images taken from chest scans, and then performing a small surgery to confirm whether a tumour is malignant.

This time-consuming and difficult process can sometimes mean that doctors miss the telltale signs of lung cancer in their early stages, leading to much greater risk for patients further down the line.



The AI successfully identified 172 of 177 tumours in a trial – giving it 97 percent effectiveness at detecting cancer

However, the researchers built an AI software trained on thousands of CT scan images from hundreds of patients that had already been examined by doctors. This data helps to ‘teach’ the AI common patterns. The software was then tested on 1,179 patients in a three-year trial, including 177 patients who were diagnosed with lung cancer.

The AI successfully identified 172 of the 177 malignant tumours in those scans – giving it 97 percent accuracy in detecting cancer.



Male Radiologist Examine At An X-Ray Of Thorax
Typically, lung cancer diagnoses are performed by a radiologist who must manually sift through hundreds of images – which means small details are sometimes missed

Despite the success, scientists warned that the software tends to identify many areas that are not cancerous. The technology will therefore need to be improved drastically before we start to see it in hospitals.

“The aim of our research is not to replace radiologists but to assist them by giving them a tool that can spot the earliest signs of lung cancer,” said researcher Benoît Audelan.

He added: “Screening for lung cancer would mean many more CT scans being taken and we do not have enough radiologists to review them all. That’s why we need to develop computer programs that can help.”



CAPTION: AI could help diagnose lung cancer a year early https://eandt.theiet.org/content/articles/2021/09/artificial-intelligence-could-help-diagnose-lung-cancer-a-year-earlier/
A research diagram illustrating how the AI is trained to recognise lung cancer.

Leave a Comment