Artificial intelligence tools have proven to be beneficial in detecting pulmonary nodules on chest CTs of adults, but less is known about their utility in pediatric populations.
Authors of a new paper published in Clinical Imaging explored this topic recently, when they put a commercially available pulmonary nodule detection AI tool to the test on chest CTs from pediatric patients between the ages of 12 and 18.
“As greater utilization of artificial intelligence for the detection of pulmonary nodules in adults has been advocated to aid radiologists who may miss the finding on initial interpretation, it is reasonable to assume a similar need for accurate detection of pulmonary nodules in children,” corresponding author Marla B.K. Sammer, MD, with the Edward B. Singleton Department of Radiology, Division of Body Imaging, at Texas Children's Hospital, and colleagues suggested. “However, to date, none of the commercially available software for detection of pulmonary nodules is intended for use in pediatric patients.”
The team tested Syngo CT Lung Computer Aided Detection (CAD)—a Seimens Healthineers AI software used to detect pulmonary nodules on adult chest scans—on 30 pediatric chest CTs containing a total of 109 nodules. Images were retrospectively reconstructed at 3 mm and 1 mm slice thickness and were reviewed by two pediatric radiologists.
This revealed the tool to be less effective in pediatric patients, as it yielded significantly lower sensitivity in kids compared to the tool’s performance in adults. However, the team did note that they were able to identify some adjustments related to slice thickness that improved its utility.
The tool’s performance improved with smaller slices, the group noted. For the 1 mm slices, it achieved a sensitivity of 39% and a positive predictive value (PPV) of 62%. In comparison, the 3 mm slices resulted in a sensitivity of 26% and PPV of 48%. Its performance also improved when excluding smaller nodules (solid < 3 mm and subsolid < 5 mm).
The team noted that the differences in performance between slice thickness and nodule size are likely related to the AI software’s constraints, as it was developed for use in adults, not children. Although more research with a larger sample size is needed, the group suggested that their findings could “guide ongoing research focusing on algorithm improvements in children with the end goal of the same or better performance of Lung CAD AI in pediatric patients of all ages as in adults.”
The study abstract is available here.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She joined Innovate Healthcare in 2021 and has since put her unique expertise to use in her editorial role with Health Imaging.