Value of dynamic contrast-enhanced magnetic resonance imaging combined with diffusion weighted imaging radiomics in diagnosis of small breast cancer lesions

Objective To investigate the value of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) combined with diffusion weighted imaging (DWI) in diagnosis of small breast cancer lesions (maximum diameter < 2 cm).

Methods A total of 110 patients with 119 breast lesions (maximum diameter < 2 cm) confirmed by surgical pathology were selected. According to different inclusion time, the subjects were assigned to a training set (62 patients with 67 lesions) and a test set (48 patients with 52 lesions). The imaging parameters of benign and malignant lesions were comparatively analyzed. A radiomics model was established by gradient boosting decision tree (GBDT) based on preoperative DCE-MRI and DWI sequences to predict benign and malignant lesions in the test set. The receiver operating characteristic (ROC) curve was drawn to analyze and compare the diagnostic efficacy of GBDT model and three methods of radiologists.

Results Flux rate constant (K^ep), transfer constant (K^trans), minimum apparent diffusion coefficient (ADC_min), mean apparent diffusion coefficient (ADC_mean) of benign lesions were higher than those of malignant lesions, while extravascular extracellular volume fraction (Ve) was lower than that of malignant lesions (P < 0.05). The maximum diameter of lesions, detected rates of malignant and benign lesions in the training set were similar to those in the test set (P > 0.05). ROC curve showed that the area under the ROC curve (AUC) of GBDT model in the diagnosis of breast cancer smaller than 2 cm was 0.945, which showed no significant difference compared with 0.923 by combined diagnosis of DCE-MRI and DWI (P > 0.05), and was larger than 0.845, 0.851 by DCE-MRI and DWI alone (P < 0.05). The sensitivity, specificity and accuracy of the best cut-off value of GBDT model were 0.91, 0.94 and 0.93, respectively, which showed no significant difference compared with 0.94, 0.81 and 0.86 of DCE-MRI combined with DWI radiomics by radiologists (P > 0.05).

Conclusion DCE-MRI combined with DWI radiomics has higher value in the diagnosis of breast cancer lesions smaller than 2 cm when combining with GBDT model, and the diagnostic results are consistent with those from experienced radiologists based on DCE-MRI combined with DWI.