Analysis of Advanced Quantitative Computed Tomography Imaging Features in Predicting Progression-free Survival of Advanced Epithelial Ovarian Cancer

Society of Gynecologic Oncology Presentation
Authors Jessica B DiSilvestro, Paul A. DiSilvestro, Abhishek Pandey, Jacob Gordon, Jonathon Ball, Katherine Moxley

Objective: Recent research efforts have focused on identifying novel clinical and imaging biomarkers capable of predicting optimal surgical resection, tumor response, and disease-specific outcomes. The objective of this study is to demonstrate the utility of quantitative CT image analysis in predicting progression-free survival (PFS) for advanced-stage epithelial ovarian cancer.

Methods: Our initial retrospective cohort identified 40 patients who had CT imaging completed prior to cytoreductive surgery. The images and clinical outcomes data were evaluated using an established imaging platform at Tempus Labs, Inc. Primary and secondary lesions were identified by a trained radiologist. More than 2,000 quantitative imaging (radiomic) features were extracted from these images, and by using the Tempus proprietary imaging platform, feature selection was performed to identify factors predictive of PFS, defined as the interval of time from diagnosis to progression. Area under the curve (AUC) analysis of radiomic features was completed comparing PFS greater or less than 20 months. Statistical significance was calculated using single-sided t test.

Results: We analyzed 40 patients who had their initial debulking surgery with at least 1 year of follow-up. The mean age was 60.2 years (range 31–80 years). Thirty-eight cases were high-grade serous histology (95%), 2 were clear cell, and 1 was carcinosarcoma. The majority of cases were FIGO stage IIIC (73%) with 5 stage IV disease. At the time of the primary debulking surgery, 35% had no gross residual disease, 35% optimal (< 1cm remaining disease), and 30% suboptimal (≥1 cm remaining disease). The univariate analysis identified 66 radiomic features with an AUC greater than 0.7 and P < 0.01 for primary lesions. Similarly, 31 predictive radiomic features were identified for secondary lesions. Within the 4 imaging feature categories, the top 3 predictive radiomic features were selected (Table 1).

Conclusion: Quantitative radiomic features can be used to predict PFS in advanced-stage ovarian cancer. Further development of this concept will be performed in a larger retrospective cohort with the ultimate goal of a prospective validation.