¸ñÀû: Mucoepidermoid carcinoma (MEC) is the most common type of malignancy
originating from the parotid gland. The aim of this study was to develop
and establish an effective survival prediction model based on a large
population-based dataset of patients with advanced parotid gland MEC
using a machine learning (ML) algorithm. ¹æ¹ý:A total of 607 patients with advanced T stage parotid gland MEC were
identified from the Surveillance, Epidemiology, and End Results (SEER)
database. The dataset was randomly assigned to a train or test dataset
at a 7:3 ratio. The Cox proportional hazard model (CoxPH), conditional
survival forest model (CSF), random survival forest model (RSF), and
DeepSurv model were used to predict disease-specific survival in
patients with advanced parotid gland MEC. The predictive ability was
evaluated using the concordance index (C-index) and the Integrated Brier
Score (IBS). °á°ú:The best prediction performance was achieved with CSF (0.79) and
DeepSurv (0.79) based on the C-index, and CSF (0.02) and CoxPH (0.02)
based on IBS. In our survival prediction ML models, age, tumor size, and
grade were important variables, as in the multivariate CoxPH model. °á·Ð:We developed and compared ML-based survival prediction models with good
performance in patients with advanced T-stage parotid gland MEC.
Compared with conventional statistical methods, ML techniques have
satisfactory results in predicting disease-specific survival in real-
world data. |