1. The clinical characteristics, preoperative hematological parameters of patients with epithelial ovarian cancer and patients with benign gynecological diseases, and the pathological stage, grade and features extracted by PET/CT images of patients with epithelial ovarian cancer were recorded.
2. Patients from Renji Hospital were divided into training group and test group at a ratio of 7:3, and patients from Shanghai First Maternity and Infant Hospital were used as external validation group.
3. The training group was used to establish the diagnosis and prognosis prediction model of epithelial ovarian cancer, and the test group and the external validation group were used to verify the model, and the area under the ROC curve, accuracy, specificity, and sensitivity were used to evaluate the effect of the model.
4. For machine learning models, SHAP and LIME algorithms were used for model interpretation.
5. Unsupervised clustering algorithm was used to distinguish the subgroups of epithelial ovarian cancer patients, and KM was used to analyze the overall survival (OS) and progression-free survival (PFS) to predict the survival and recurrence of the subgroups. Overall survival (OS) was defined as the time from the first diagnosis of epithelial ovarian cancer to the confirmation of death or the end of follow-up. Progression-free survival (PFS) was defined as the time from the first diagnosis of epithelial ovarian cancer to the confirmation of disease progression or the end of follow-up.
