Article Text

Download PDFPDF

#880 Impact of ascites and peritoneal metastatic lesions, measured by newly developed, deep learning-based algorithm, on survival outcomes in advanced epithelial ovarian cancer
  1. Hyunji Lim,
  2. Se Ik Kim,
  3. Soon Ho Yoon,
  4. Taek Min Kim,
  5. Maria Lee,
  6. Jeong Yeon Cho,
  7. Jae-Weon Kim and
  8. Hyun Hoon Chung
  1. Seoul National University College of Medicine, Seoul, South Korea


Introduction/Background We developed an auto-segmentation algorithm using deep learning to identify peritoneal metastatic (PM) lesions in advanced epithelial ovarian cancer (aEOC). We investigated the impact of ascites and PM lesion volumes on survival outcomes in aEOC.

Methodology We measured ascites and PM lesion volumes in abdominopelvic cavity of 195 patients with aEOC using our algorithm and pre-treatment computed tomography (CT) images. Patients were divided into high- and low-volume groups based on the median value for each factor. Survival outcomes were compared between the groups.

Results Of the 195 patients, 127 (65.1%) had FIGO stage III and 68 (34.9%) had stage IV disease. The most common histologic subtype was high-grade serous carcinoma (78.5%). Primary cytoreductive surgery was performed in 69.2% of patients, with 56.4% achieving complete cytoreduction. The median volumes of ascites and PM lesions were 714.5 cm3 and 341.1 cm3, respectively.

There was no difference in progression-free survival (PFS) between the high- and low-volume ascites groups (P=0.338). However, the high-volume ascites group had worse overall survival (OS) compared to the low-volume group (5-year PFS rate, 68.7% vs. 46.1%, P=0.08). In multivariate analyses adjusting for histologic subtypes and residual tumor after surgery, high-volume ascites was an independent poor prognostic factor for OS (adjusted hazard ratio [aHR] 1.801; 95% confidence interval [CI] 1.147–2.828; P=0.011).

No differences in PFS (P=0.120) and OS (P=0.279) were observed between the high- and low-volume PM groups. However, in a subgroup analysis of patients who achieved complete cytoreduction (n=110), the high-volume PM group was an independent poor prognostic factor for OS (aHR 2.231; 95% CI 1.066–4.669; P=0.033) after adjusting for histology and neoadjuvant chemotherapy use.

Conclusion Our study demonstrates the successful measurement of ascites and PM lesion volumes using a deep learning-based auto-segmentation algorithm. Volumetric measurements of ascites and PM lesions could serve as novel prognostic factors for survival outcomes in aEOC patients.

Disclosures I have no conflict of interest to declare

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.