Peritoneal cytology predicting distant metastasis in uterine carcinosarcoma: machine learning model development and...

Area of Science:

• Gynecologic Oncology
• Computational Pathology
• Machine Learning in Medicine

Background:

• Uterine carcinosarcoma is an aggressive malignancy with a high risk of distant metastasis.
• Predicting distant metastasis is crucial for effective patient management and treatment planning.

Purpose of the Study:

• To develop and validate a machine learning model for predicting distant metastasis in uterine carcinosarcoma.
• To integrate peritoneal cytology findings into predictive models for enhanced accuracy.

Main Methods:

• Utilized SEER database data including clinical information and peritoneal cytology from uterine carcinosarcoma patients.
• Applied and compared eight machine learning algorithms (Logistic Regression, SVM, GBM, etc.) to predict distant metastasis.
• Evaluated model performance using AUC, calibration curves, DCA, and SHAP values for interpretability.

Main Results:

• Peritoneal cytology, T stage, age, and tumor size were identified as key predictors of distant metastasis.
• The logistic regression model achieved an AUC of 0.882 (training) and 0.881 (testing).
• GBM emerged as the top-performing model, with significant clinical utility indicated by calibration and DCA curves.

Conclusions:

• This study presents the first model integrating peritoneal cytology to predict distant metastasis in uterine carcinosarcoma.
• The developed tool facilitates early identification of high-risk patients, optimizing follow-up and personalized treatment.
• This approach supports improved monitoring and tailored therapeutic strategies for uterine carcinosarcoma.