Chemotherapeutic response prediction in high-grade serous ovarian cancer using histopathological image analysis

High-grade serous ovarian cancer (HGSOC) is a subtype of epithelial cancer originating from thefallopian tubes and it is a the most abundant and lethal subtype of epithelial ovarian cancer.The standard treatment for HGSOC consists of surgery and platinum-taxane chemotherapy,and even though 80% of patients have an excellent initial response, the majority relapse within18 months leading to less than 45% 5-year survival rate. Thus, it is important to develop toolsto predict the patient’s response and identify which patients who benefit from the treatment.This study focuses on using prospectively collected samples from 37 HGSOC patients, specificallyhematoxylin and eosin stained(H&E) histopathological images corresponding to primarytreatment phase, in order to predict the response to chemotherapy in HGSOC.The study is mainly composed by two phases. Firstly, tools have been developed to extractfeatures, as Haralick and local binary patterns(LBP) features, by means of texture methods,and an existing pipeline has been modified to obtain 185 morphological features related to thedensity and the heterogeneity of the cells, starting from H&E images. Secondly, machine learningmethods were used to predict the response to chemotherapy based on the herein extractedfeatures, considered both individually and in combination with each other.The accuracy of the response to the chemotherapy prediction obtained in the end is >78% andthe results demonstrate that H&E images allow prediction of chemotherapy response in HGSOCand shed light on the role of tumor infiltrating lymphocytes in HGSOC.