CT-based texture analysis predicts BRAF^V600E mutation in calcified papillary thyroid carcinoma

Area of Science:

• Radiology
• Oncology
• Genetics

Background:

• The BRAF gene is crucial in papillary thyroid carcinoma (PTC) development.
• BRAF^V600E mutation is a key driver in PTC pathogenesis.

Purpose of the Study:

• To evaluate computed tomography (CT)-based texture analysis for predicting BRAF^V600E mutation in calcified PTC.
• To assess the diagnostic performance of texture features in identifying BRAF^V600E status.

Main Methods:

• Texture features were extracted from 475 calcified PTC cases using MaZda software.
• Feature selection algorithms (Fisher, POE+ACC, MI) and classification methods (PCA, LDA, NDA) were employed.
• Diagnostic performance was assessed using receiver operating characteristic (ROC) curves.

Main Results:

• Nonlinear discriminant analysis (NDA) demonstrated superior diagnostic performance compared to PCA and LDA.
• The POE+ACC+NDA and MI+NDA methods achieved high accuracy, with areas under the curve (AUC) of 0.969 (training) and 0.964 (validation).
• Texture analysis of both parenchymal and calcified areas proved effective in predicting BRAF^V600E mutation.

Conclusions:

• The POE+ACC+NDA or MI+NDA methods offer high diagnostic performance for predicting BRAF^V600E mutation in PTC.
• CT-based texture analysis, especially of calcified areas, is a viable tool for non-invasively predicting BRAF^V600E mutation status.