Machine learning-based prognostic model for metastatic breast cancer and its interpretability: a multicenter retrospective study

Description

Jie Wang, Si-Li Jin, Jing-Hua Wu, and Kai Zhu build a survival prediction model for metastatic breast cancer. Their multicentre study from twenty twenty-five uses a large national database. Standard scoring systems predict survival with limited accuracy. Machine learning models can be accurate but hard to explain.

Over a thousand patients trained and validated the models internally. A small hospital cohort supplied external validation data. Key survival factors came from adjusted regression analysis first. Four learning algorithms then competed on prediction quality. A forest-based model performed best in training and validation. The authors stress interpretability alongside accuracy for clinical use. Doctors gain a tool that aims to be both strong and understandable.

Catalog record

Source: https://europepmc.org/articles/PMC12771207