Comparison of AI software tools for automated detection, quantification and categorization of pulmonary nodules in the HANSE LCS trial

Authors
Rimma Kondrashova, Filip Klimeš, Till Frederik Kaireit, Katharina May, Jörg Barkhausen, Susanne Stiebeler, Jonathan Sperl, Sabine Dettmer, Frank Wacker, Jens Vogel-Claussen
Journal
Scientific Reports
Related Product

LCS

Date Published
2024.11
Summary

Kondrashova et al. (2023) compared two AI-based software tools, Aview (Coreline Soft, Seoul, Korea, S1) and ChestCTExplore (Siemens Healthineers, Germany, S2), for lung nodule detection, quantification, and classification in the HANSE Lung Cancer Screening (LCS) trial. The study analyzed 946 low-dose CT scans and compared AI results with radiologists’ final readings. The correlation of true positive nodule volumes between the two tools was strong (r > 0.95), but S2 consistently measured larger volumes than S1. S1 outperformed S2 in sensitivity, positive predictive value (PPV), and overall agreement with radiologists' assessments. Moreover, S1 showed higher accuracy in Lung-RADS classification, with a 75% agreement with radiologists compared to 55% for S2. Notably, 38% of participants had different Lung-RADS scores depending on the AI tool, potentially affecting patient management. The study highlights the need for high performance and agreement among accredited AI tools to ensure consistency in a national screening program.

Contact

Please leave your inquiry if you have any questions regarding our products, recruitment, investment, or any other matters.

Contact us