Deep Learning Model Helps Predict Lung Cancer
Could a single, routine X-ray fed into a deep learning model be the first line of defense in spotting lung cancer? You might be interested to know that NCI-funded researchers at Massachusetts General Brigham and Harvard Medical School in Boston have a model they think will do just that.
Their deep learning model, called CXR-Lung-Risk, was helpful in identifying people at risk for lung cancer or other lung disease, based on a single chest X-ray image.
For details and to learn whether you can get the code, see “Deep Learning to Estimate Lung Disease Mortality From Chest Radiographs.” See also the recent Radiological Society of North America’s annual meeting.
The deep learning model identified people at high risk of lung cancer, even after adjusting for other risk factors like smoking, age, sex, etc.
The CXR-Lung-Risk also worked using X-ray images from nonsmokers (e.g., obtained as part of routine practice for assessing patients with cough, fever, etc.), identifying nonsmokers at risk of developing lung cancer over the next 6 years.
“Lung cancer is increasingly common in nonsmokers, but we currently have only a limited number of tools to tell who’s at risk of lung cancer and who might benefit from further tests, like chest CT screening,” said senior author Dr. Michael Lu.
He added, “Chest X-rays are one of the most common medical tests. The underlying idea here is that there’s information on those images about the individual’s health and risk of cancer. We can extract that information using AI, which adds to the value of these existing chest X-rays, and helps us make more personalized decisions about our patients’ health.”