Oran Muduroglu is executive chairman of Median Technologies and president of Eyonis, Inc., maker of eyonis LCS.
Lung cancer screening saves lives by finding cancer when it is more likely to be treatable. Yet the U.S. is still far from realizing its full potential. Only 18.2% of people eligible for screening were actually screened in 2022, even as expanded guidelines have nearly doubled the population that could benefit.
Getting more eligible Americans screened is an urgent public health goal. But expanding screening also creates a practical challenge for health systems: increasing demand for imaging at a time when the radiologist workforce is already stretched thin. According to the American College of Radiology (ACR), “Across every practice setting, imaging volumes continue to rise while the workforce pipeline expands slowly and unevenly.”
Accelerating that trend for lung cancer, the U.S. Preventive Services Task Force (USPSTF) in 2021 expanded screening eligibility by lowering the starting age from 55 to 50 and the smoking threshold from 30 to 20 pack-years. The change is estimated to increase the eligible population to 14.5 million.
Meanwhile, workforce research from the ACR found that annual radiologist attrition more than doubled from 1.1% in 2014 to 2.5% in 2022. A separate ACR survey found that 70% of radiologists worked in understaffed practices.
And it’s no wonder. Radiologists are required to compare new scans to prior scans to determine the malignancy risk of each nodule detected. An abnormal finding requires nodule measurement, risk classification and follow-up recommendations. More suspicious findings can lead to short-interval CT, PET/CT, tissue sampling or specialist referral. Moreover, each additional screening exam can create downstream work—follow-up recommendations and tracking patients over time.
Why False Positives Affect More Than Workflow
False positives in lung cancer screening create downstream burdens not only for patients but also for radiologists and health systems. After a suspicious nodule is found, the appropriate follow-up depends on the size, appearance, growth and risk profile. Under Lung-RADS, for instance, some findings return to annual screening while others call for six-month or three-month re-imaging or additional diagnostic evaluation.
For many patients, this period of watchful waiting creates prolonged uncertainty and distress while they wait for follow-up. Separately, maintaining adherence to annual screening remains a significant challenge. A 2025 multicenter study found adherence fell from 61% in the first year after screening to 51% in the second.
This matters because screening saves far more lives if patients return consistently for recommended repeat imaging and follow-up. The hidden cost of false positives, therefore, may be the patients who never come back for their next annual scan.
Advancing From Detection To Diagnosis
Lung cancer prognosis varies dramatically by stage. For non-small cell lung cancer, five-year relative survival is about 67% when the disease is localized, compared with 12% when it has spread to distant sites.
Lung cancer outcomes have improved with advances in both treatment and detection. In particular, low-dose computed tomography (LDCT) screening can detect the disease earlier in people at high risk. In the National Lung Screening Trial, participants screened with LDCT had a 15% to 20% lower risk of dying from lung cancer than those screened with chest X-rays.
Pulmonary nodules can be small and difficult to distinguish within the complex architecture of the lung. Radiologists have traditionally depended on their own expertise and eyesight to identify potentially cancerous findings. More recently, artificial intelligence tools have been developed to augment LDCT interpretation.
The first of these AI tools were for detection only (CADe)—finding suspicious nodules that might otherwise be missed. A 2024 systematic review found that AI-assisted reading improved sensitivity by up to 20 percentage points, but often with lower specificity. Yet detection alone—simply identifying more abnormalities without computer-aided diagnosis (CADx)—risks multiplying workload rather than improving care.
More recent AI approaches can combine detection with characterization, analyzing features such as size, shape, texture, density and growth to help estimate a nodule’s likelihood of malignancy. Compared with false-positive rates reported in a 2021 JAMA evidence review of low-dose CT lung-cancer screening, newer AI approaches have the potential to substantially reduce false positives. Evidence is still evolving, however, and performance varies by technology, population and study.
The future of screening is not simply finding more nodules. It’s empowering radiologists to identify those that truly matter and to act upon them before they advance in stage. Better characterization can give radiologists greater confidence to escalate suspicious findings to PET/CT, biopsy or specialist referral rather than risking growth during surveillance. Small-cell lung cancer grows extremely rapidly, with documented tumor doubling times as short as 25 to 30 days.
For healthcare leaders, the relevant question is whether an AI tool can support both detection and characterization, integrate into existing workflows and demonstrate meaningful clinical or operational value. The goal is not to replace radiologists, but to help them determine which findings deserve closer attention and which are likely benign.
After all, better screening is not just about finding more disease. It’s about reducing uncertainty while preserving the benefits of early detection.
Lung Cancer Screening 2.0
AI’s broader potential is to help make increasingly strained screening programs more efficient and sustainable while preserving radiologist oversight.
The larger opportunity extends beyond workflow efficiency to making lung cancer screening itself more sustainable and more effective. AI has the potential to support earlier and more consistent assessment while reducing unnecessary follow-up. But whether individual systems deliver those benefits depends on the evidence for each technology and how it performs in clinical practice. By shortening the path from detection to action, AI may also reduce losses to follow-up and improve adherence.
As shown, LDCT has an established mortality benefit for people at high risk for lung cancer. The next opportunity is improving screening itself by making it more precise, more efficient and more sustainable for the millions of patients and clinicians who rely on it.
For healthcare leaders, the path forward is less about detecting ever more abnormalities than about improving the balance between early detection, diagnostic precision and health-system capacity. AI’s value should ultimately be judged by whether it helps radiologists achieve that balance in real-world practice and move appropriate patients toward earlier diagnosis.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?