V.N. Zhurman1,2
1Primorsky Regional Oncological Dispensary, Vladivostok
2Pacific State Medical University, Vladivostok
Zhurman Varvara N. — Cand. of Sci. (Med.), oncologist of the Primorsky Regional Oncological Dispensary
63А Russkaya Str., Vladivostok, 690105, Russian Federation, tel. +7-904-622-25-77, e-mail: varvara2007@yandex.ru, ORCID ID: 0000-0002-6927-3336
Abstract
The purpose of this study was to use a machine learning algorithm to determine prognostic parameters, to predict the survival of patients with ovarian cancer.
Information on 910 ovarian cancer patients from the Primorsky Regional Oncological Dispensary was analyzed using the Random Forest machine learning method and Cox regression. The Kaplan ― Meyer method was used to visualize the overall survival of patients.
Results. It was revealed that the use of platinum-containing agents in second-line chemotherapy, primary cytoreduction (complete and optimal and suboptimal), histological types of serous high-grade, low-grade and mucinous cancer make the greatest contribution to predicting the overall survival of patients with ovarian cancer. The proposed combination of a machine learning algorithm with Cox regression analysis and the Kaplan ― Meyer method makes it possible to isolate from the many factors accompanying the development of ovarian cancer those indicators that have prognostic value for determining the overall survival of patients. This algorithm can be useful for more accurate assessment of individual results and selection of the best treatment options for ovarian cancer patients.
Key words: ovarian cancer, survival, correlation, Random Forest, Kaplan ― Meyer method.