1626P - Monitoring of hepatic tumors on computed tomography: Quantifiable target lesions and variability of volume-based assessments

Date 28 September 2014
Event ESMO 2014
Session Poster Display session
Topics Staging procedures (clinical staging)
Translational Research
Basic Principles in the Management and Treatment (of cancer)
Presenter Hubert Beaumont
Citation Annals of Oncology (2014) 25 (suppl_4): iv546-iv563. 10.1093/annonc/mdu358
Authors H. Beaumont1, S. Egels2, S. Zaim3, A. Iannessi4, O. Lucidarme2
  • 1Sciences, MEDIAN Technologies, 06560 - Valbonne/FR
  • 2Radiology Department, La Pitié-Salpêtrière hospital, 75013 - Paris/FR
  • 3Radiology Department, MEDIAN Technologies, San Francisco/US
  • 4Radiology, Centre Antoine Lacassagne, Nice/FR



Tumor volume (TV) is an emerging quantitative imaging biomarker that is expected to overcome some limitations of the longest axial diameter measurements in objective response evaluation criteria. Several features of the TV biomarker are still investigated, notably its application to hepatic lesions where the minimum detectable change has not been established. Another undocumented aspect concerns the kind of hepatic lesions that can be reliably quantified. This study aims at measuring inter reader variability of hepatic TV assessments on computed tomography.


82 patients enrolled from four different hospitals were included in our retrospective study. A total of 142 target lesions that included 32 hepatocellular carcinomas (HCC) were analysed. Each of the three involved experienced radiologists performed volumetric measurements of all lesions using a semi-automated segmentation system allowing manual corrections. Each reader documented lesion complexity when found relevant. Bland Altman analysis was performed between pairs of readers. Wilcoxon test reported association between inter-reader relative difference and factors as pathology, lesion pattern, localisation or phase enhancement.


39 out of 142 (27.5%) lesions were reported as complex and unsuitable for quantification by, at least, two radiologists. 14 out of 103 (13.6%) lesions that were segmented by all readers were inconsistently interpreted, generating a variability in the range of [100%; 190%]. Association was found between variability and disease (p=0.014) HCC/non-HCC. Considering the subset of consistent segmentations, the Limit of Agreement by each pair of readers was: 49.1%, 54.9% and 53.1%. No association was found regarding factors as subcapsular/isolated (p=0.71), portal/arterial scans (p=0.58), hypo/hyper attenuated lesions (p=0.37) or diffuse/focal lesions (p=0.21).


Our results indicate that volumetry was not applicable to a subset of lesions and that HCC assessments featured higher variability. Approximate bounds for minimal detectable change of hepatic TV are found in the range of [45%; 55%]. We did not identify evidence of significant associations between variability and specific imaging or lesion features.


All authors have declared no conflicts of interest.