[Construction of prediction model combined dual-energy CT quantitative parameters and conventional CT features for assessing the Ki-67 expression levels in invasive breast cancer].

To develop a model combined with dual-energy CT quantitative parameters and conventional CT features for evaluating the expression level of Ki-67 in invasive breast cancer. A total of 191 patients with histologically confirmed invasive breast cancer in Lishui Central Hospital from March 2019 to December 2020, were retrospectively enrolled, all of them were females, aged from 25 to 77 (53.2±11.3) years. All patients underwent preoperative non-contrast chest and contrast-enhanced Dual energy CT scans, and the normalized iodine concentration (NIC) of lesions on arterial and venous phase, spectral curve slope (λ), and normalized effective atomic number (nZ) were measured and calculated, and their conventional CT characteristics were assessed. According to the results of immunohistochemistry (IHC), the patients were divided into Ki-67 high expression group (=129 patients) and low expression group (=62 patients) level. The differences in clinical data, conventional CT characteristics and dual-energy CT quantitative parameters between the two groups were analyzed. The receiver operating characteristic curve (ROC) curve was conducted to assess the efficacy of each individual model and joint model in evaluating Ki-67 expression levels, and the area under the curve (AUC), sensitivity, specificity, and accuracy were calculated, respectively. In the analysis of CT features, the longest diameter, shape and enhancement pattern of the tumor were significantly difference between the two groups (all <0.05). The NIC, nZ on the arterial phase and NIC, nZ and λ [(,)] on the venous phase were higher in the high Ki-67 expression group compared to the low expression group [0.13 (0.12, 0.16) vs 0.11 (0.08, 0.14), 0.71 (0.70, 0.75) vs 0.70 (0.67, 0.72), 0.40 (0.32, 0.48) vs 0.23 (0.17, 0.32), 3.10 (2.58, 3.63) vs 2.86 (2.19, 3.48), 0.88 (0.85, 0.92) vs 0.85 (0.84, 0.86), all <0.05]. The logistic regression model, which integrated significant conventional CT features and dual-energy CT quantitative parameters, demonstrated the highest diagnostic performance for assessing Ki-67 expression levels, with an AUC of 0.924, sensitivity of 88.37%, specificity of 83.87%, and accuracy of 86.91%; the AUC of the dual-energy CT parameter model was 0.908, sensitivity of 82.17%, specificity of 88.71%, and accuracy of 84.29%. Though the diagnostic efficacy was no significant difference (=0.238), both models showed superior to the conventional CT feature model (all <0.001). A dual-energy CT quantitative parameter combined with a conventional CT feature model was successfully constructed, which has a good evaluation performance on the expression level of Ki-67 in invasive breast cancer.