Reproducibility of temporal volume change in CT of lung cancer: comparison of computer software and manual assessment

¹ Department of Radiology, Osaka University Graduate School of Medicine, Osaka, Japan, 2Department of Radiology, Osaka National Hospital, Osaka, Japan, 3Department of Radiology, Osaka Rosai Hospital, Osaka, Japan, 4Department of Diagnostic Radiology, Osaka Medical Center for Cancer and Cardiovascular Diseases, Osaka, Japan, 5Department of Radiology, Kinki Central Hospital, Hyogo, Japan, and 6CT Advance Applications, GE Healthcare Technologies, Tokyo, Japan

	Abstract

The purpose of this study was to investigate the reproducibility of volumetric software evaluation and manual evaluation of tumour growth. Three observers manually evaluated whether tumour volume was increasing, if it was unchanged, or if it had decreased in size in two serial CT examinations of 45 solid lung cancers. The tumour volumes were calculated three times using volumetric software and evaluated using the same classifications as for manual evaluation. Both data sets were divided into three groups: growth or reduction with consistency among all three evaluations (group A), growth or reduction with consistency between only two evaluations (group B), and others (group C). The volume variation and relative volume variation were calculated from the median volumes measured by volumetric software. Although all 45 tumours were categorized in group A by volumetric software, only 21 tumours were categorized in group A by manual assessment. The relative volume variation of the manual assessment was 88.5 ± 76.5%, 20.8 ± 28.3% and 12.9 ± 12.8% in group A, B and C, respectively. Significant differences were found between groups A and B (p < 0.01) and between groups A and C (p < 0.001). Inconsistency is often seen in manual assessment; in contrast, evaluation using volumetric software has good reproducibility, even when the relative change in tumour volume is small.