Distinguishing granulomatous lobular mastitis from breast cancer using a clinical-radiomics nomogram to improve diagnostic accuracy.

Background: Granulomatous lobular mastitis (GLM) is often misdiagnosed clinically as breast cancer (BC). Therefore, it is crucial to differentiate between GLM and BC. Methods: This study used 259 samples from 129 patients with GLM and 130 with BC. A total of 874 radiomics and 11 clinical features were obtained. The least absolute shrinkage and selection operator algorithm was used to select radiomics features. Univariate and multivariate analyses were performed to screen clinical features. Three machine learning algorithms were applied to assess the efficiency of the radiomics, clinical, and combined models, which were compared to select the optimal model. Finally, a nomogram based on the optimal model was developed. Decision curve analysis (DCA) and calibration curves were used to assess the clinical utility of the nomogram. Results: Twelve radiomics features were identified as the most relevant for distinguishing GLM from BC, including the original gray level co-occurrence matrix autocorrelation feature. Important clinical features included age, nipple inversion, and C-reactive protein levels. The combined model demonstrated superior performance in terms of accuracy, specificity, and sensitivity compared with the clinical and radiomics models. A nomogram was constructed based on the combined model. The calibration curve and DCA further confirmed the superior clinical value of the nomogram. Conclusion: A combined model incorporating 12 radiomics and 3 clinical features is potentially valuable for distinguishing GLM from BC.