A novel prediction method for lymph node involvement in endometrial cancer: machine learning.

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning

Background:

• Lymphadenectomy necessity and lymph node involvement (LNI) prediction in endometrial cancer (EC) are critical clinical questions.
• Machine learning (ML) offers advanced estimation capabilities for complex medical predictions.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting lymph node involvement (LNI) in endometrial cancer (EC).
• To identify key histopathological factors associated with LNI in EC patients.

Main Methods:

• Utilized the Naïve Bayes algorithm to construct prediction models for LNI in 762 EC patients.
• Models incorporated histopathological factors: histology, lymphovascular space invasion (LVSI), grade, tumor diameter, myometrial invasion (MI) depth, cervical glandular stromal invasion (CGSI), tubal/ovarian involvement, and pelvic LNI.
• Logistic regression analysis identified independent predictors of LNI.

Main Results:

• The ML models achieved high accuracy rates: 84.2%-88.9% for LNI and 85.0%-97.6% for pelvic LNI (PaLNI).
• Histologic type, LVSI, MI depth, and CGSI were independently significant predictors of LNI (p<0.001).
• LNI was detected in 13.4% of patients, with PaLNI in 7.1%.

Conclusions:

• Machine learning demonstrates potential as a valuable tool in the decision-making process for endometrial cancer management.
• This preliminary study suggests ML algorithms can improve LNI prediction accuracy.
• Future research incorporating sentinel lymph node status and imaging data could further advance EC management.