Multiparametric MRI for differentiating idiopathic granulomatous mastitis from invasive breast cancer:Improving radiologists' diagnostic accuracy

Purpose: To develop a model integrating multiparametric MRI and clinical data to distinguish idiopathic granulomatous mastitis (IGM) from invasive breast cancer (IBC) and assess its potential to improve clinical decisionmaking in ambiguous cases. Methods: A retrospective study was conducted on 255 female patients (135 with IGM and 120 with IBC) from two hospitals, divided into training (n = 161), internal validation (n = 41), and external validation (n = 53) cohorts. All patients underwent multiparametric MRI (including DCE and DWI) within two weeks prior to histopathological exam. Multiparametric MRI-based radiomics and clinical features were extracted and then selected using a two-staged method. The logistic regression was applied to construct DCE-model, DWI-model, Fusion_rad-model and Fusion_rad + cli-model. Diagnostic performance was evaluated using the area under the receiver operating characteristic curve (AUC). The model' ability to assist radiologists in differential diagnosis was also analyzed. Results: The Fusion_rad + cli-model achieved the highest diagnostic performance with AUCs of 0.946, 0.923, and 0.845 in the training cohort, the internal cohort and external validation cohort, respectively. It surpassed the other three models for differentiating IGM from IBC in all validation cohorts. Additionally, the Fusion_rad + climodel improved radiologists' diagnostic capabilities, increasing the average accuracy from 0.732 to 0.805 in the internal validation cohort and from 0.717 to 0.792 in the external validation cohort. Conclusion: The radiomics-clinical model can differentiate IGM from IBC and improve radiologists' diagnostic capabilities on MRI. Further studies are needed to validate these findings in larger, diverse populations and to explore the model's integration into routine diagnostic workflows.