Predictive Features of Thymic Carcinoma and High-Risk Thymomas Using Random Forest Analysis

Area of Science:

• Oncology
• Radiology
• Machine Learning

Background:

• Thymic tumors, including thymic carcinomas and high-risk thymomas, present diagnostic challenges.
• Accurate differentiation is crucial for appropriate treatment and patient management.

Purpose of the Study:

• To identify predictive computed tomography (CT) features for distinguishing thymic carcinomas from high-risk thymomas.
• To evaluate the efficacy of the random forest algorithm in this diagnostic task.

Main Methods:

• A cohort of 137 patients with pathologically confirmed thymic tumors was analyzed.
• Clinical and 20 CT features were assessed for their association with tumor type.
• Univariate analysis and random forest modeling were employed for feature selection and prediction.
• Receiver operating characteristic (ROC) curve analysis validated the model's performance.

Main Results:

• Random forest identified tumor shape, lymphadenopathy, and pericardial effusion as key differentiating features.
• The model achieved high predictive accuracy: 94.73% for test data and 96.35% for the entire dataset.
• ROC analysis demonstrated strong diagnostic power with an area under the curve of 0.957.

Conclusions:

• The random forest model offers a highly efficient tool for the predictive diagnosis of thymic carcinomas and high-risk thymomas.
• Irregular tumor shape, presence of lymphadenopathy, and pericardial effusion are strong indicators of thymic carcinoma.