The following is a summary of the “Machine learning-based radiomic computed tomography phenotyping of thymic epithelial tumors: Predicting pathological and survival outcomes,” published in the February 2023 issue of Thoracic and cardiovascular surgery by Tian, et al.
Accurately predicting clinicopathological outcomes for patients with thymic epithelial tumors is still difficult. Here, they aimed to examine how well radiomic computed tomography phenotyping using machine learning could predict pathological (World Health Organization [WHO] type and TNM stage) and survival outcomes (overall and progression-free survival) in patients with thymic epithelial tumors.
Participants in this retrospective study had been diagnosed with a thymic epithelial tumor between 2001 and 2022. The preoperative CT scans with no contrast added were used to extract the radiomic features. Pathological and survival outcomes were predicted using random forest and random survival forest models after rigorous feature selection. The area under the curve (AUC) and an internal validation via the bootstrap method were used to evaluate the model’s efficacy. Among the total of 124 participants, the median age was 61. With an AUCWHO of 0.898 (95% CI, 0.753-1.000) and an AUCTNM of 0.766 (95% CI, 0.642-0.886), the radiomics random forest models of WHO type and TNM stage demonstrated satisfactory performance.
Good performance was seen in predicting overall survival and progression-free survival using radiomics random survival forest models (integrated AUCs, 0.923; 95% CI, 0.691-1.000 and 0.702; 95% CI, 0.513-0.875, respectively), with the integrated AUCs increasing to 0.935 (95% CI, 0.705-1.000) and 0.811 (95% CI, 0.647-0.942), respectively, when combined with clinicopathological features. Radiomic computed tomography phenotyping using machine learning has the potential to improve prognostic performance in patients with thymic epithelial tumors when combined with clinicopathological features.