Habitat imaging based on dual-energy computed tomography for predicting axillary lymph node metastasis in breast cancer.

Abstract

BackgroundQuantitative analysis with habitat clustering represents an innovative, non-invasive approach to quantify tumor heterogeneity.PurposeTo characterize intratumoral spatial heterogeneity using dual-energy computed tomography (DECT) in breast cancer patients and investigate the performance of habitat imaging in predicting axillary lymph node (ALN) metastasis compared with radiomics.Material and MethodsA total of 135 patients were randomly assigned to a training group (n = 95) and a testing group (n = 40). An additional 50 patients served as the validation group. Four intratumoral subregions with different wash-in and wash-out enhancement modes were identified through cluster analysis of arterial and venous phase iodine concentration maps. The percentage of each subregion was quantified to construct habitat imaging. Radiomics features were extracted from iodine concentration maps, and Boruta was used for feature selection. Habitat imaging and radiomics model performance was compared by net reclassification improvement (NRI) and integrated discrimination improvement (IDI).ResultsHabitat imaging demonstrated areas under the receiver operating characteristic curve (AUCs) of 0.82, 0.80, and 0.78 in the training, testing, and validation groups, respectively. In addition, the AUCs of the radiomics models were 0.78, 0.70, and 0.65 in the training, testing, and validation groups, respectively. NRI and IDI demonstrated that habitat imaging was statistically superior to the radiomics model (P < 0.05).ConclusionsHabitat imaging based on intratumoral spatial heterogeneity can predict ALN metastasis in breast cancer and was superior to radiomics.