New artificial intelligence technology is helping diagnose the U.S.’s most deadly cancer sooner, greatly improving the chances of lung cancer patients’ survival.
New artificial intelligence technology is helping diagnose the U.S.’s most deadly cancer sooner, greatly improving the chances of patients’ survival.
The key? Detecting tiny lung nodules when doctors aren’t screening for cancer, and automating and streamlining follow-up care. While most incidentally-discovered lung nodules turn out to be noncancerous, some become malignant over time.
The importance of catching lung cancer early is clear: The five-year survival rate for non-small cell lung cancer when detected in localized Stage 1 is 67%. However, most lung cancer is diagnosed after it has spread to other organs, when the five-year rate is 12%, according to the American Cancer Society.
Inova Schar Cancer Institute, based in Fairfax, Virginia, is one forward-thinking cancer center harnessing the power of AI to flag incidental lung nodules that often go unnoticed, during an emergency room CT scan or MRI for pneumonia or a broken bone. The Eon Lung Cancer Screening system uses computational linguistics and natural language processing to scan radiology reports.
The company says it identifies high-risk patients with 98.3% accuracy by analyzing imaging data and integrating with electronic health records in real time.
‘The patient is still in the ER, we call and tell them to come right to the clinic’
Amit “Bobby” Mahajan is the medical director of interventional pulmonology in the Inova Health System. (Disclosure: He is also the doctor who did my bronchoscopy in November 2022 and told me I had lung cancer. After four months of one-pill-a-day targeted therapy and a robotic-assisted lobectomy, I was declared cancer-free in May 2023 and have remained that way while continuing my daily pill.)
AI-powered technology is enabling Schar’s interventional pulmonologists and surgeons to get patients with found-by-accident nodules into cancer care months or years earlier. Mahajan heads the Incidental Pulmonary Nodule Clinic, as part of the Inova Saville Cancer Screening and Prevention Center.
“Whether it be an MRI, a chest CT, or abdominal CT, it takes that data, comprises it into a finding, and then makes a risk score of that being cancer,” said Mahajan, during a recent WTOP visit and demonstration of the Eon technology.
With the AI system scanning electronic health records as data is entered, “We’re able to call the patient and say, ‘Look, I know you just had a CT scan in the ER for your abdominal pain, but we also caught a lung nodule in the bottom of your lung that is suspicious,’” Mahajan said.
“For better or worse, we’ve had more than a handful of people who we’ve said, ‘We need to send you over to the clinic right now, because you came in for something that’s nothing to worry about, but we did find something that needs to be addressed today,’” he added.
Traditionally, reaching a cancer diagnosis for a patient with a persistent cough or other symptoms can take weeks and requires patients and doctors to coordinate follow-up scans and labs.
“From an AI perspective, the system will learn more from our CT scans and image reports every time it sees one, and starts picking out the word ‘spiculated,’ the word ‘nodule,’ and where it’s located,”‘ Mahajan said.
While benign nodules usually have smooth borders, a spiculated nodule’s edges appear irregular, or spiky, which often suggests the lesion is malignant.
“It takes that data to the very well known Brock Model for risk of lung cancer, and it will actually calculate the risk of cancer in those patients, and give us a percentage,” Mahajan said. “Anyone over 5%, we call, and get them into the clinic right away, most of the time in the same week.”
After being notified of an incidental nodule found in ER imaging, some patients prefer to check with their primary care physician.
“Totally reasonable,” Mahajan said.
Streamlining the follow-up process helps reduce the risk of patients “falling through the cracks.”
“We’ve biopsied them two days later, and gotten a diagnosis of cancer,” Mahajan said. “Luckily, most have been early stage disease and they’ve been resected afterward.”
With lung cancer, resection is a surgical procedure to remove lung tissue affected by cancer and is regarded as the most effective treatment for cancer that hasn’t spread to other organs.
“Our goal is to get a patient with a newly-diagnosed lung cancer evaluated as soon as possible, to get them into surgery,” Schar thoracic surgeon Melanie Subramanian said. “It’s not only better for treating the disease, but it also gives patients a peace of mind too, knowing that they have a treatment plan and a treatment team.”
The AI system creates guideline-based care plans, and sends alerts to doctors and nurse navigators, helping patients stay on schedule for future screenings.
‘It’s as close to an Xbox controller as you get’
Artificial intelligence is also enabling robotic bronchoscopy procedures.
“Previously, when we had to biopsy these small nodules in the lung, we had to use a handheld camera, to drive down as far as we could, but the lungs and airways get smaller and smaller the further out you go,” Mahajan said.
“Now, we have robotic platforms,” Mahajan added. “The patient is completely asleep, and we drive about a four millimeter camera down to these nodules, using a handheld controller that’s as close to an Xbox controller as you can get.”
And AI helps navigate through the airways: “There’s advanced imaging associated as well, and with the robotic platform, we can pretty much reach anything in the lung nowadays,” he said.
Inova Schar says 69% of lung cancers are now being detected at Stage 1 or 2, compared to only 34% without low-dose CT screening and proactive follow-up of incidental nodules.
Get breaking news and daily headlines delivered to your email inbox by signing up here.
© 2025 WTOP. All Rights Reserved. This website is not intended for users located within the European Economic Area.