Development and validation of a CT-texture analysis nomogram for preoperatively differentiating thymic epithelial...

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Thymic epithelial tumors (TETs) are common anterior mediastinum neoplasms with significant histologic variability.
• Accurate preoperative classification of TET subtypes is crucial for treatment planning.

Purpose of the Study:

• To develop and validate a predictive model for TET pathological classification using computed tomography (CT) and texture analysis (TA).
• To create a nomogram for preoperative risk stratification of thymic epithelial tumors.

Main Methods:

• Retrospective analysis of 172 TET patients, divided into training (n=120) and validation (n=52) cohorts.
• Development of prediction models using clinical factors, CT signs, and texture features via LASSO regression.
• Evaluation of model performance using ROC curve analysis (AUC) and DeLong test; clinical utility assessed by DCA.

Main Results:

• A combined model integrating 2 CT signs and 2 texture parameters demonstrated the highest predictive efficiency (AUC 0.88 in training, 0.94 in validation).
• This combined model significantly outperformed models using only clinical or CT features (p < 0.05).
• A validated nomogram was developed based on the combined model, showing good calibration for predicting TET classifications.

Conclusions:

• A CT and texture analysis-based nomogram provides a reliable tool for preoperative prediction of TET histological subtypes.
• This nomogram facilitates individualized risk assessment and treatment strategies for thymic epithelial tumor patients.