Radiomics models using machine learning algorithms to differentiate the primary focus of brain metastasis

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Brain metastases are common in adults, with distinct MRI features based on primary tumor type.
• Machine learning (ML) and radiomics offer advanced tools for medical image analysis and tumor differentiation.

Purpose of the Study:

• To develop and evaluate radiomics models using ML algorithms to differentiate lung cancer from breast cancer brain metastases.
• To assess the diagnostic performance of these models on post-contrast T1-weighted MRI images.

Main Methods:

• Retrospective analysis of 180 patients with lung or breast cancer brain metastases.
• Radiomic features extracted from T1-weighted MRI, with feature selection using the least absolute shrinkage and selection operator.
• Development of ML models (logistic regression, SVM, KNN, MLP, LightGBM) for differentiation.

Main Results:

• The LightGBM radiomics model achieved the highest diagnostic performance.
• Area under the curve (AUC) was 0.875 in the training set and 0.866 in the validation set.
• The model demonstrated strong accuracy in predicting primary lesion types.

Conclusions:

• Enhanced MRI radiomics models, particularly LightGBM, accurately predict primary tumor origins for brain metastases.
• This approach aids in differentiating lung and breast cancer metastases, improving diagnostic capabilities.