Scientific Reports (Jun 2024)

Construction and validation of a clinical risk model based on machine learning for screening characteristic factors of lymphovascular space invasion in endometrial cancer

  • Fang Wang,
  • Rui Pang,
  • Shaohong Shi,
  • Yang Zhang

DOI
https://doi.org/10.1038/s41598-024-63436-7
Journal volume & issue
Vol. 14, no. 1
pp. 1 – 11

Abstract

Read online

Abstract This study aimed to identify factors that affect lymphovascular space invasion (LVSI) in endometrial cancer (EC) using machine learning technology, and to build a clinical risk assessment model based on these factors. Samples were collected from May 2017 to March 2022, including 312 EC patients who received treatment at Xuzhou Medical University Affiliated Hospital of Lianyungang. Of these, 219 cases were collected for the training group and 93 for the validation group. Clinical data and laboratory indicators were analyzed. Logistic regression and least absolute shrinkage and selection operator (LASSO) regression were used to analyze risk factors and construct risk models. The LVSI and non-LVSI groups showed statistical significance in clinical data and laboratory indicators (P < 0.05). Multivariable logistic regression analysis identified independent risk factors for LVSI in EC, which were myometrial infiltration depth, cervical stromal invasion, lymphocyte count (LYM), monocyte count (MONO), albumin (ALB), and fibrinogen (FIB) (P < 0.05). LASSO regression identified 19 key feature factors for model construction. In the training and validation groups, the risk scores for the logistic and LASSO models were significantly higher in the LVSI group compared with that in the non-LVSI group (P < 0.001). The model was built based on machine learning and can effectively predict LVSI in EC and enhance preoperative decision-making. The reliability of the model was demonstrated by the significant difference in risk scores between LVSI and non-LVSI patients in both the training and validation groups.

Keywords