Diagnostic performance of artificial intelligence for pediatric pulmonary nodule detection on chest computed tomography: comparison of simulated lower radiation doses

The combination of low dose CT and AI performance in the pediatric population has not been explored. Understanding this relationship is relevant for pediatric patients given the potential radiation risks. Here, the objective was to determine the diagnostic performance of commercially available Compu...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:European journal of pediatrics 2023-11, Vol.182 (11), p.5159-5165
Hauptverfasser: Salman, Rida, Nguyen, HaiThuy N., Sher, Andrew C., Hallam, Kristina, Seghers, Victor J., Sammer, Marla B. K.
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The combination of low dose CT and AI performance in the pediatric population has not been explored. Understanding this relationship is relevant for pediatric patients given the potential radiation risks. Here, the objective was to determine the diagnostic performance of commercially available Computer Aided Detection (CAD) for pulmonary nodules in pediatric patients at simulated lower radiation doses. Retrospective chart review of 30 sequential patients between 12–18 years old who underwent a chest CT on the Siemens SOMATOM Force from December 20, 2021, to April 12, 2022. Simulated lower doses at 75%, 50%, and 25% were reconstructed in lung kernel at 3 mm slice thickness using ReconCT and imported to Syngo CT Lung CAD software for analysis. Two pediatric radiologists reviewed the full dose CTs to determine the reference read. Two other pediatric radiologists compared the Lung CAD results at 100% dose and each simulated lower dose level to the reference on a nodule by nodule basis. The sensitivity (Sn), positive predictive value (PPV), and McNemar test were used for comparison of Lung CAD performance based on dose. As reference standard, 109 nodules were identified by the two radiologists. At 100%, and simulated 75%, 50%, and 25% doses, lung CAD detected 60, 62, 58, and 62 nodules, respectively; 28, 28, 29, and 26 were true positive (Sn = 26%, 26%, 27%, 24%), 30, 32, 27, and 34 were false positive (PPV = 48%, 47%, 52%, 43%). No statistically significance difference of Lung CAD performance at different doses was found, with p -values of 1.0, 1.0, and 0.7 at simulated 75%, 50%, and 25% doses compared to standard dose. Conclusion : The Lung CAD shows low sensitivity at all simulated lower doses for the detection of pulmonary nodules in this pediatric population. However, radiation dose may be reduced from standard without further compromise to the Lung CAD performance. What is Known: • High diagnostic performance of Lung CAD for detection of pulmonary nodules in adults. • Several imaging techniques are applied to reduce pediatric radiation dose. What is New: • Low sensitivity at all simulated lower doses for the detection of pulmonary nodules in our pediatric population. • Radiation dose may be reduced from standard without further compromise to the Lung CAD performance.
ISSN:1432-1076
0340-6199
1432-1076
DOI:10.1007/s00431-023-05194-8