Integrated radiogenomics models predict response to neoadjuvant chemotherapy in high grade serous ovarian cancer

Abstract

High grade serous ovarian cancer (HGSOC) is a highly heterogeneous disease that often presents at an advanced, metastatic state. The multi-scale complexity of HGSOC is a major obstacle to measuring response to neoadjuvant chemotherapy (NACT) and understanding its determinants. Here we propose a radiogenomic framework integrating clinical, radiomic, and blood-based biomarkers to measure and predict the response of HGSOC patients to NACT, showing how quantitative imaging data can serve as the backbone of multi-scale data integration. We developed and validated our approach in two independent highly-annotated multi-omic multi-lesion data sets. In a discovery cohort (n=72) we found that different tumour sites present distinct response patterns, and identified volumetric response assessment as a better predictor of overall survival (OS) than RECIST 1.1 status. We trained an ensemble machine learning approach to predict tumour volume response to NACT from data obtained prior to treatment, and validated the model in an internal hold-out cohort (n=20) and an independent external patient cohort (n=42). Benchmarking integrated models against models built on single data types highlighted the importance of comprehensive patient characterisation. Our study sets the foundation for developing new clinical trials of NACT in HGSOC.