Machine-learning-based computed tomography radiomic analysis for histologic subtype classification of thymic...

Area of Science:

• Radiology
• Artificial Intelligence
• Oncology

Background:

• Thymic epithelial tumors (TETs) are classified by WHO into high-risk and low-risk subtypes.
• Accurate differentiation is crucial for appropriate clinical management and treatment strategies.
• Computed tomography (CT) is a primary imaging modality for evaluating TETs.

Purpose of the Study:

• To assess the efficacy of machine learning-based CT radiomic analysis in distinguishing high-risk from low-risk TETs.
• To compare the performance of different machine learning classifiers for this task.

Main Methods:

• Retrospective analysis of 155 patients with high-risk (n=72) or low-risk (n=83) TETs.
• Radiomic features extracted from unenhanced CT (UECT) and contrast-enhanced CT (CECT) datasets.
• Nested leave-one-out cross-validation with feature selection and classifiers (GLM, KNN, SVM, RF) evaluated using ROC curves and AUC.

Main Results:

• Combining UECT and CECT radiomic features yielded the best differentiation performance across all classifiers.
• The Random Forest (RF) classifier achieved the highest Area Under the Curve (AUC) of 0.87.
• Generalised Linear Model (GLM) and K-Nearest Neighbor (KNN) classifiers showed strong performance with AUCs of 0.86, while Support Vector Machine (SVM) had an AUC of 0.84.

Conclusions:

• Machine learning-based CT radiomic analysis demonstrates excellent performance in differentiating high-risk and low-risk TETs.
• This AI-driven approach serves as a valuable tool to support clinical decision-making in TET management.
• Further integration of radiomics into clinical practice could improve patient outcomes.