15% of lung-resection surgeries are performed on benign nodules, with AI a potential remedy
About 15% of lung-resections surgeries are performed on nodules that are actually benign, researchers estimated in a new study published in Clinical Imaging [1].
The specialty has made great strides in addressing this number. As recently as the mid-1990s, about half of such surgeries were conducted on benign nodules, a number that dropped to 20% by 2005, one study found.
The American College of Radiology introduced the Lung CT Screening Reporting and Data System, or Lung-RADS, in 2014 to help standardize the process. Harvard Medical School researchers recently set out to see how the system has impacted surgical rates in the years since. The rate appears to remain on a downward trajectory, though radiologists still can strive further toward zero.
“Surgical resection of benign nodules is unavoidable despite application of Lung-RADS guidelines in a modern screening program, with approximately 15% of surgeries being done performed for benign lesions,” corresponding author Mark M. Hammer, MD, with the Department of Radiology at Brigham and Women's Hospital, and colleagues concluded. “More conservative approaches for equivocal cases, such as a multidisciplinary team approach, lung nodule biopsy, or short-term follow-up, should be considered before direct referral to surgery, which may help prevent unnecessary operations on benign pathologies. Future work in artificial intelligence may help triage nodules better and avoid unnecessary surgeries.”
For the study, researchers analyzed data from a large tertiary healthcare network’s lung cancer screening program, spanning from 2015 to 2021. Their search uncovered more than 21,000 CT exams performed on 9,000-plus patients, with 260 undergoing a resection. Pathology results revealed 220 lung cancer cases (85%), two other malignancies (1%) and 38 benign findings (15%). Of the benign nodules, 12 had scarring/fibrosis, five with benign neoplasms, 14 infection/inflammation and seven were classified as “other.”
Hammer et al. highlighted “substantial overlap” between benign and malignant nodules in terms of size, FDG uptake and Brock risk score. About 74% of resected benign lung nodules were either growing or new and 78% of the very low-risk nodules were growing.
“In our study, the benign resected lung nodules shared overlapping features with malignant nodules, making differentiation between these categories difficult, even in retrospect,” the authors noted. “Machine learning and deep learning artificial intelligence techniques are promising for nodules characterization … However, validation of these models is yet to be finalized.”
Read much more, including potential study limitations, in Clinical Imaging at the link below.