Constructing a prediction model for lymph node metastasis in patients with incidental finding of endometrial cancer based on Fully-Connected Network.

Objective: Rare studies focused on patients with incidental diagnosis of endometrial cancer (EC) after hysterectomy. We intended to construct a prediction model of lymph node metastasis (LNM) based on Fully-Connected Network (FC Network) for these patients.

Methods: A total of 3,920 cases of EC that met the criteria from Obstetrics & Gynecology Hospital of Fudan University between January 2016 and February 2023 and 1995 cases from Fudan University Shanghai Cancer Center between January 2013 and October 2020 were retrospectively included for the construction of a predicting model which was based on FC Network. At the same time, 572 cases were prospectively collected for external validation.

Results: The sensitivity of the model was 0.946. Lympho-vascular space invasion, myometrial invasion, tumor grade, microcystic elongated and fragmented invasion, progesterone receptor, and cancer antigen 125 were used to construct a simplified nomogram. The area under the curve of the nomogram was 0.890 and 0.885 in validation and prospective cohorts, respectively.

Conclusion: The model we proposed has good sensitivity and can be used to predict the risk of LNM in patients with incidentally found EC. The simplified nomogram can be used as a substitute in certain situations. Based on another study, the threshold of 5% and 25% can be used for risk stratification.