Radiomics based on F-FDG PET/CT could differentiate breast carcinoma from breast lymphoma using machine-learning approach: A preliminary study.

Purpose: Our study assessed the ability F-fluorodeoxyglucose (FDG) positron emission tomography (PET)/computed tomography (CT) radiomics to differentiate breast carcinoma from breast lymphoma using machine-learning approach.

Methods: Sixty-five breast nodules from 44 patients diagnosed as breast carcinoma or breast lymphoma were included. Standardized uptake value (SUV) and radiomic features from CT and PET images were extracted using local image features extraction software. Six discriminative models including PETa (based on clinical, SUV and radiomic features from PET images), PETb (SUV and radiomic features from PET images), PETc (radiomic features only from PET images), CTa (clinical and radiomic features from CT images), CTb (radiomic features only from CT images), and SUV model were generated using least absolute shrinkage and selection operator method and linear discriminant analysis. The areas under the receiver operating characteristic curve (AUCs), accuracy, sensitivity, and specificity were calculated to evaluate the discriminative ability of these models.

Results: PETa and CTa models showed the best ability to differentiation in training and validation group (AUCs of 0.867 and 0.806 for PETa model, AUCs of 0.891 and 0.759 for CTa model, respectively).

Conclusion: Models based on clinical, SUV, and radiomic features of F-FDG PET/CT images could accurately discriminate breast carcinoma from breast lymphoma.