Ultrasound-based radiomics nomogram for distinguishing invasive breast cancer (IBC) from invasive breast cancer with intraductal component (IBC-IC)

Objective: The objective was to devise a nomogram integrating radiomics features, sonographic characteristics, and immunologic indices to distinguish between IBC and IBC-IC. Methods: The preoperative ultrasonography (US) findings and pathological data from 244 BCE patients (115 IBCIC and 129 IBC) diagnosed with IDC through surgical confirmation were retrospectively gathered. The study utilized the BI-RADS lexicon from the ACR to interpret the findings. 1125 radiomics features were derived from the regions of breast lesions, coupled with pathological data, to establish a predictive model employing the LASSO technique. The model's performance was examined by the AUC, along with sensitivity, and specificity measures, to evaluate its effectiveness in differentiating between the two groups of subjects. Results: Four prognostic models for IDC-IC were developed, comprising a conventional ultrasound (CUS) signature, pathology signature nomogram, radiomics signature, and combined signature. The combined model exhibited the highest diagnostic efficacy, achieving the AUC of 0.893, sensitivity of 0.875, and specificity of 0.835 in the training set, and 0.835, 0.800, and 0.789 in the testing set. Conclusion: The combined model, integrating CUS, pathology information, and radiomics features derived from CUS, exhibited a notable capacity for preoperatively distinguish IDC-IC. This capability holds promise for guiding personalized surgical strategies for BC individuals undergoing breast-conserving surgery (BCS).