Abstract
Introduction: The size of nodules in the lungs is an important criterion in deciding what further tactics to use. There are several measurement approaches: determination of maximum size, average size, nodule volume and volume doubling time (VDT). Each of these requires validity, reproducibility and accuracy, which are not always achieved in routine practice. Aim: To evaluate the performance of different programs in the automated measurement of nodules and solid masses in lung tissue and the reproducibility of the results. Materials and methods. To conduct a comparative evaluation of automated nodule size estimation, we used four software products: Program A — Vitrea (manufactured by Canon, version 6.3.2047.65), Program B — Vitrea (manufactured by Canon, version 7.14.2.227), Program C — AW Server (manufactured by GE, version 3.2 Ext.4.0), and Program D — Hiveomics Platform (Hiveomics Ltd., version 1.2). We selected 62 nodules and solid masses for statistical processing. Results. During the study, all software products had technological defects (unable to process the study technical/other reasons) that prevented the measurement of all nodules and masses accepted for participation. Statistically significant differences were found when comparing the mean effective diameters obtained from the measurements. These differences were observed between automatic nodule measurement systems from different manufacturers and between older generation software products and the AI algorithm. There were no statistically significant differences observed in the measurements of mean effective diameters between the AI-based automatic nodules analysis system and the results obtained after physician delineation correction on earlier versions of automatic volumetry systems. Conclusion. The study of average volume values in all groups showed no statistically significant differences, confirming the universality of the use of this parameter in clinical practice.