Machine learning-based radiomics for predicting BRAF-V600E mutations in ameloblastoma.

Area of Science:

• Oncology
• Radiology
• Genetics
• Machine Learning

Background:

• Ameloblastoma is an aggressive odontogenic neoplasm.
• The BRAF-V600E gene mutation is a key driver in ameloblastoma pathogenesis.
• Accurate identification of this mutation is crucial for treatment.

Purpose of the Study:

• To develop and validate a radiomics-based machine learning method.
• To identify BRAF-V600E gene mutations in ameloblastoma patients non-invasively.
• To correlate radiomics features with mutation status.

Main Methods:

• Retrospective analysis of 103 ameloblastoma patients.
• Radiomics features extracted from CT images.
• Machine learning models (Random Forest, XGBoost) trained and validated.
• Synthetic Minority Over-sampling Technique (SMOTE) used for class imbalance.

Main Results:

• Random Forest model achieved an Area Under the ROC Curve (AUC) of 0.87.
• XGBoost model showed a slightly lower AUC of 0.83.
• Higher radiomics scores correlated with increased susceptibility to BRAF-V600E mutations.

Conclusions:

• A radiomics-based machine learning model can accurately detect BRAF-V600E mutations in ameloblastoma.
• The Random Forest model offers a convenient, cost-effective, non-invasive alternative to molecular testing.
• This approach may guide preoperative/postoperative treatment decisions and improve patient outcomes.