Radiogenomics of high-grade serous ovarian cancer: Multireader multi-institutional study from the cancer genome atlas ovarian cancer imaging research group

Hebert Alberto Vargas, Erich P. Huang, Yulia Lakhman, Joseph E. Ippolito, Priya Bhosale, Vincent Mellnick, Atul B. Shinagare, Maria Anello, Justin Kirby, Brenda Fevrier-Sullivan, John Freymann, C. Carl Jaffe, Evis Sala

Research output: Contribution to journalArticlepeer-review

40 Scopus citations


Purpose: To evaluate interradiologist agreement on assessments of computed tomography (CT) imaging features of highgrade serous ovarian cancer (HGSOC), to assess their associations with time-to-disease progression (TTP) and HGSOC transcriptomic profiles (Classification of Ovarian Cancer [CLOVAR]), and to develop an imaging-based risk score system to predict TTP and CLOVAR profiles. Materials and Methods: This study was a multireader, multi-institutional, institutional review board-approved, HIPAA-compliant retrospective analysis of 92 patients with HGSOC (median age, 61 years) with abdominopelvic CT before primary cytoreductive surgery available through the Cancer Imaging Archive. Eight radiologists from the Cancer Genome Atlas Ovarian Cancer Imaging Research Group developed and independently recorded the following CT features: characteristics of primary ovarian mass(es), presence of definable mesenteric implants and infiltration, presence of other implants, presence and distribution of peritoneal spread, presence and size of pleural effusions and ascites, lymphadenopathy, and distant metastases. Interobserver agreement for CT features was assessed, as were univariate and multivariate associations with TTP and CLOVAR mesenchymal profile (worst prognosis). Results: Interobserver agreement for some features was strong (eg, α = .78 for pleural effusion and ascites) but was lower for others (eg, α = .08 for intraparenchymal splenic metastases). Presence of peritoneal disease in the right upper quadrant (P = .0003), supradiaphragmatic lymphadenopathy (P = .0004), more peritoneal disease sites (P = .0006), and nonvisualization of a discrete ovarian mass (P = .0037) were associated with shorter TTP. More peritoneal disease sites (P = .0025) and presence of pouch of Douglas implants (P = .0045) were associated with CLOVAR mesenchymal profile. Combinations of imaging features contained predictive signal for TTP (concordance index = 0.658; P = .0006) and CLOVAR profile (mean squared deviation = 1.776; P = .0043). Conclusion: These results provide some evidence of the clinical and biologic validity of these image features. Interobserver agreement is strong for some features, but could be improved for others.

Original languageEnglish
Pages (from-to)482-492
Number of pages11
Issue number2
StatePublished - Nov 2017


Dive into the research topics of 'Radiogenomics of high-grade serous ovarian cancer: Multireader multi-institutional study from the cancer genome atlas ovarian cancer imaging research group'. Together they form a unique fingerprint.

Cite this