Comparison of nodule volumetric classification by using two different nodule segmentation algorithms in an LDCT lung cancer baseline screening dataset

This study compared the performance of two nodule segmentation algorithms—Syngo.via VB30A and AVIEW v1.1.39.14—for volumetric classification in LDCT lung cancer screening. Using data from 300 participants, two independent radiologists evaluated nodules with each software according to the NELSON2.0 protocol, classifying nodules as indeterminate-positive or negative. AVIEW, a deep learning-based system, used Hounsfield unit thresholds and an asymmetric deformable model for segmentation. The inter-software agreement was high (κ = 0.88). AVIEW resulted in 5 misclassifications (2 PM, 3 NM), while Syngo.via had 12 (11 PM, 1 NM). Although two authors were affiliated with Coreline Soft, they were not involved in the data analysis. Overall, both software tools demonstrated similar performance in participant/scan-level classification.