44/#219 Predicting response to platinum-based chemotherapy for tubo-ovarian high-grade serous carcinoma using an artificial intelligence histopathology platform. (4th December 2022)

• Record Type: Journal Article
• Authors: Tiu, Ekin; Krishna, Vrishab; Nimgaonkar, Vivek; Krishnan, Rayan; Bhambhvani, Hriday; O'Donoghue, Odhran; Vrabac, Damir; Joshi, Anirudh; Liang, Brooke; Zhang, Xiaoming; Han, Lucy; Wang, Aihui; Krishna, Viswesh; Howitt, Brooke
• Is Part Of: International journal of gynecological cancer. Volume 32(2022)Supplement 3
• Page Start: A46
• Page End: A47
• Publication Date: 2022-12-04
• Subjects: Generative organs, Female -- Cancer -- Periodicals; 616.99465
• DOI: 10.1136/ijgc-2022-igcs.88
• Languages: English
• ISSNs: 1048-891X

Abstract

Abstract : Objectives: Platinum-based chemotherapy is the standard of care first-line systemic treatment for patients diagnosed with advanced stages of tubo-ovarian high-grade serous carcinoma (HGSC). While the majority of patients respond, roughly 15% of patients are platinum-resistant. We aimed to develop an artificial intelligence-based platform leveraging routine pre-treatment histopathology specimens to predict platinum-based chemotherapy response. Methods: 87 patients from The Cancer Genome Atlas (TCGA) and 19 patients from Stanford Hospital with HGSC who received platinum-based chemotherapy post resection were included in this study. Using scanned hematoxylin and eosin-stained (H&E) images, we extracted nuclei images from tissue regions using segmentation models and computed geometric features of these nuclei. In the TCGA cohort, quantitative features of the nuclear geometry were correlated with Progression Free Survival (PFS) using a multivariable Cox Proportional Hazards (CPH) model in order to construct a signature associated with platinum treatment benefit. The signature was assessed with a Kaplan Meier Estimator and log rank test by comparing the PFS between the high and low cohorts stratified by the signature in the internal TCGA and external Stanford cohorts. Results: The artificial intelligence derived histological biomarker is able to stratify patients into high and low responders to platinum-based chemotherapy with statistical significance (logrank test – internal: p=0.000556, external: p=0.00571), achieving hazard ratios of 0.227 (95% CI: 0.092, 0.559) on the internal TCGA test cohort and 0.132 (95% CI: 0.025, 0.704) on the external Stanford Hospital validation cohort. Conclusions: An artificial intelligence derived histological biomarker utilizing only routine whole-slide histopathology images can robustly predict responders and non-responders to platinum-based chemotherapy.