基于双能量CT定量参数的浸润性乳腺癌Ki-67高表达预测模型构建

Development of a prediction model for high Ki-67 expression in invasive breast cancer based on dual-energy CT quantitative parameters

摘要:

背景与目的 Ki-67是反映乳腺癌细胞增殖活性的重要指标，与患者预后评估及治疗方案制定密切相关。目前，Ki-67主要依赖病理免疫组织化学检测，缺乏稳定有效的无创评估方法。本研究旨在基于双能量CT（DECT）定量参数构建列线图模型，预测浸润性乳腺癌的Ki-67表达水平。方法 前瞻性纳入2023年10月—2024年6月遂宁市中心医院收治的128例经病理证实的浸润性乳腺癌患者。采用免疫组织化学法检测Ki-67表达，并以20%为界分为高表达组（85例）和低表达组（43例）。收集DECT平扫期、动脉期及静脉期定量参数，包括能谱曲线斜率（λ）、碘浓度（IC）、标准化碘浓度（NIC）、有效原子序数（Zeff）、标准化有效原子序数（NZeff）及肿瘤-动脉碘浓度增强分数（TAIF）。采用ROC曲线分析各参数的预测效能，继而通过Boruta算法筛选特征变量，并基于筛选出的变量建立Logistic回归预测模型，最终绘制列线图。采用Bootstrap重抽样进行内部验证，并通过ROC曲线、校准曲线及临床决策曲线（DCA）评价模型性能。结果 Ki-67高表达组在平扫期λ、动脉期λ、动脉期IC、动脉期NIC、动脉期Zeff、动脉期NZeff、静脉期NIC、静脉期Zeff及TAIF等参数方面均高于低表达组（均P<0.05）。ROC分析显示，各DECT参数均具有一定预测能力，其中动脉期IC预测效能最佳（AUC=0.721）。Boruta算法最终筛选出TAIF、动脉期Zeff、动脉期IC、动脉期λ及静脉期λ共五个变量构建预测模型。模型内部验证AUC为0.818（95% CI=0.728~0.900），Hosmer-Lemeshow检验显示模型拟合良好（P=0.312），DCA提示模型在较宽阈值范围内具有较好的临床净获益。结论 基于DECT定量参数构建的列线图模型可较准确地预测浸润性乳腺癌Ki-67高表达状态，为乳腺癌无创分子特征评估及临床决策提供参考依据。

Abstract:

Background and Aims Ki-67 is an important biomarker reflecting proliferative activity in breast cancer and plays a critical role in prognosis evaluation and treatment decision-making. Currently, Ki-67 assessment mainly relies on invasive immunohistochemical examination, and effective non-invasive evaluation methods remain limited. This study aimed to develop a nomogram based on dual-energy computed tomography (DECT) quantitative parameters for predicting Ki-67 expression in invasive breast cancer.Methods A total of 128 patients with pathologically confirmed invasive breast cancer who underwent DECT examination at Suining Central Hospital between October 2023 and June 2024 were prospectively enrolled. According to a Ki-67 cutoff value of 20%, patients were classified into a high-expression group (n=85) and a low-expression group (n=43). Quantitative DECT parameters from the non-contrast, arterial, and venous phases were collected, including spectral curve slope (λ), iodine concentration (IC), normalized iodine concentration (NIC), effective atomic number (Zeff), normalized effective atomic number (NZeff), and tumor-arterial iodine fraction (TAIF). Receiver operating characteristic (ROC) analysis was used to evaluate the predictive performance of each parameter. Subsequently, the Boruta algorithm was employed to select feature variables, and a logistic regression model was constructed to develop a nomogram. Internal validation was performed using Bootstrap resampling, and model performance was evaluated by ROC analysis, calibration curves, and decision curve analysis (DCA).Results Compared with the low-expression group, the high Ki-67 expression group showed significantly higher values in non-contrast λ, arterial-phase λ, arterial-phase IC, arterial-phase NIC, arterial-phase Zeff, arterial-phase NZeff, venous-phase NIC, venous-phase Zeff, and TAIF (all P<0.05). ROC analysis demonstrated that all DECT quantitative parameters had predictive value for Ki-67 expression, among which arterial-phase IC showed the best performance (AUC=0.721). Five variables, including TAIF, arterial-phase Zeff, arterial-phase IC, arterial-phase λ, and venous-phase λ, were selected to establish the prediction model. Internal validation showed that the model achieved an AUC of 0.818 (95% CI=0.728-0.900). The calibration curve demonstrated good agreement between predicted and observed probabilities (Hosmer-Lemeshow test, P=0.312), and DCA indicated favorable clinical utility across a wide range of threshold probabilities.Conclusion The nomogram based on DECT quantitative parameters can effectively predict high Ki-67 expression in invasive breast cancer and may provide a useful non-invasive tool for molecular characterization and clinical decision-making.