Development of prediction model to estimate future risk of ovarian lesions: A multi-center retrospective study

Bilin Jing; Gaowen Chen; Miner Yang; Zhi Zhang; Yue Zhang; Jingyao Zhang; Juncheng Xie; Wenjie Hou; Yong Xie; Yi Huang; Lijie Zhao; Hua Yuan; Weilin Liao; Yifeng Wang

doi:10.1016/j.pmedr.2023.102296

Development of prediction model to estimate future risk of ovarian lesions: A multi-center retrospective study

Prev Med Rep. 2023 Jun 23:35:102296. doi: 10.1016/j.pmedr.2023.102296. eCollection 2023 Oct.

Authors

Bilin Jing¹, Gaowen Chen¹, Miner Yang², Zhi Zhang³, Yue Zhang⁴, Jingyao Zhang⁴, Juncheng Xie⁴, Wenjie Hou⁵, Yong Xie⁶, Yi Huang⁷, Lijie Zhao⁸, Hua Yuan⁹, Weilin Liao³, Yifeng Wang¹

Affiliations

¹ Zhujiang Hospital of Southern Medical University, Guangzhou 510280, China.
² Guangzhou Women and Children's Medical Center, Guangzhou 510620, China.
³ Geography and Planning of Sun Yat-sen University, Guangzhou 510275, China.
⁴ Second Clinical Medical College, Guangzhou 510599, China.
⁵ Soochow University Medical Center, Suzhou 215125, China.
⁶ Foshan First People's Hospital, Foshan 528010, China.
⁷ Nanhai District People's Hospital, Foshan 528099, China.
⁸ Foshan Maternal and Child Health Hospital, Foshan 528099, China.
⁹ Wuxi Maternal and Child Health Hospital, Wuxi 214002, China.

Abstract

Background: To develop the preoperative prediction of ovarian lesions using regression-based statistics analyses and machine learning methods based on multiple serological biomarkers in China.

Methods: 1137 patients with ovarian lesions in Zhujiang Hospital and 518 patients in others hospital in China were randomly assigned to training, test and external validation cohorts. Five machine learning classifiers, including Random Forest (RF), Extreme Gradient Boosting (XGB), Support Vector Classifier (SVC), K-nearest Neighbor (KN), Multi-Layer Perceptron (MLP) and the Lasso-Logistics prediction model (LLRM) were used to derive diagnostic information from 23 predictors.

Results: The RF model had a high diagnostic value (AUC = 0.968) in predicting benign and malignant ovarian disease. Age and MLR were also potential diagnostic indicators for predicting ovarian disease except tumor indicators. The RF model well distinguished borderline ovarian tumors (AUC = 0.742). The RFM had a high predictive power to identify ovarian serous adenocarcinoma (AUC = 0.943) and ovarian endometriosis cysts (AUC = 0.914).

Conclusions: The RF models can effectively predict adnexal lesions, promising to be adjuncts to the preoperative prediction of ovarian cancer.

Keywords: AUC; Disease Prediction; Lasso Regression; Machine Learning; Ovarian Disease.