Development and validation of a dynamic contrast-enhanced magnetic resonance imaging-based habitat and peritumoral radiomic model to predict axillary lymph node metastasis in patients with breast cancer: a retrospective study.

Background: The heterogeneity within breast cancer and its microenvironment are associated with metastasis. Analyzing distinct tumor subregions using habitat analysis and characterizing the tumor microenvironment through radiomics may be valuable for predicting axillary lymph node metastasis (ALNM) in breast cancer. This study aimed to develop and validate a nomogram for predicting ALNM in breast cancer patients by integrating clinicopathological, intra- or peri-tumoral radiomic, and habitat signatures based on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and determine the optimal peritumoral region size for accurate prediction. Methods: Four hundred and twenty-six breast cancer patients who underwent preoperative DCE-MRI at Shenzhen People's Hospital between June 2019 and August 2021 were retrospectively enrolled (338 in training set, 88 in test set). The clinicopathological data were analyzed by univariable and multivariable analyses. Peritumoral regions were generated with thicknesses of 2, 4, 6, and 8 mm. Habitat analysis clustered three sub-regions within the tumor area. Intratumoral, peritumoral, and habitat features were extracted and selected using the least absolute shrinkage and selection operator (LASSO) regression method. The prediction models were constructed, including (I) Clinical model, (II) Intra model, (III) Peri2mm model, (IV) Peri4mm model, (V) Peri6mm model, (VI) Peri8mm model, (VII) Habitat model, and (VIII) Fusion nomogram model. Models were evaluated using the receiver operating characteristic (ROC) curve analysis, calibration curve analysis, net reclassification improvement (NRI), integrated discrimination improvement (IDI), and decision curve analysis (DCA). Results: The Clinical model showed relatively low predictive performance with an area under the curve (AUC) of 0.667 in the test set, while the Intra model demonstrated moderate predictive performance with an AUC of 0.745 in the test set. The 4 mm was identified as the optimal peritumoral region size for ALNM prediction, with AUCs of 0.871 and 0.773 in training and test sets, respectively. The Habitat model exhibited outstanding predictive performance, achieving AUCs of 0.973 and 0.854 in the training and test set, respectively. The Fusion nomogram model, incorporating clinicopathological signatures, Peri4mm radiomic signatures, and habitat signatures, achieved the highest AUCs (0.977 and 0.873 in training and test sets). This model was well-calibrated with significant clinical benefit, outperforming individual signatures according to calibration curve, NRI, IDI, and DCA. Conclusions: The optimal peritumoral region size based on DCE-MRI radiomics for predicting ALNM in breast cancer patients was 4 mm. The nomogram combining clinicopathological factors, Peri4mm radiomics, and habitat signatures derived from DCE-MRI demonstrates robust performance in predicting ALNM and may aid clinical decision-making.