Using Machine Learning to Predict the Prognosis of Cervical Cancer Patients with Lymph Node Metastasis: An Analysis Based on the SEER Database.

Abstract

Cervical cancer is one of the most common malignant tumors in women worldwide, and patients with lymph node metastasis have a poor prognosis. This study aimed to develop an effective machine learning model to predict the prognosis of these patients. Data from the SEER*Stat database (version: November 2021) was used, including 1016 female patients diagnosed with cervical cancer and lymph node metastasis from 2000 to 2020. Various machine learning models, including XGBoost, random forest, SVM, ANN, and the Cox proportional hazards model, were constructed and evaluated using metrics such as C-index, AUC, accuracy, and precision. Additionally, to validate model stability, a random sample of 200 patients from 8 registries between 1975 and 2019 was used as a validation set. XGBoost outperformed other models with an AUC of 0.787 in the validation set and C-index values of 0.900 and 0.773 for the training and testing sets, respectively. Cox regression analysis showed that surgery at the primary site significantly improved survival outcomes and reduced mortality. XGBoost demonstrated superior performance in predicting the prognosis of cervical cancer patients with lymph node metastasis, providing new support for personalized clinical management.