瘤内联合瘤周影像组学预测乳腺癌新辅助治疗疗效的研究

A Study on Intratumoral and Peritumoral Radiomics for Predicting the Response to Neoadjuvant Therapy in Breast Cancer

摘要: 研究目的：本研究旨在通过融合动态对比增强磁共振成像(dynamic contrast-enhanced magnetic resonance imaging, DCE-MRI)的瘤内与瘤周影像组学特征，开发并验证一种无创预测乳腺癌患者新辅助治疗(neoadjuvant systemic therapy, NST)后病理完全缓解(pathological complete response, pCR)状态的工具。研究方法：回顾性分析534名乳腺癌患者的病理及影像资料。融合瘤内与瘤周3 mm特征构建联合预测模型，与单纯瘤内模型进行比较。最后在训练集与验证集中计算曲线下面积(area under the curve, AUC)量化区分度；绘制校准曲线评价模型预测概率与实际发生概率的一致性；最后进行决策曲线分析(decision curve analysis, DCA)，在不同阈值概率范围内计算净获益，以评估模型的潜在临床实用性。主要结果：研究发现联合模型预测pCR的效能优于单纯瘤内模型。在验证集中，其曲线下面积AUC达到0.851 (95% CI: 0.792~0.910)，优于单一瘤内模型0.840 (95% CI: 0.779~0.901)。此外，在校准性能方面，联合模型在训练集和验证集中均表现出良好的拟合度；DCA证实，该模型在临床决策分析中显示出较高的净获益。结论：本研究证实结合瘤内、瘤周信息特征的联合模型能精准预测乳腺癌NST的疗效。该模型在判别能力、校准性能和临床实用性方面均表现优越，可作为治疗前个体化疗效预测的候选工具。但本研究为回顾性单中心设计，研究结果尚需前瞻性多中心外部验证。

Abstract: Objective: This study aimed to develop and validate a noninvasive tool for predicting pathological complete response (pCR) after neoadjuvant systemic therapy (NST) in breast cancer patients, by integrating intratumoral and peritumoral radiomic features derived from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). Methods: A total of 534 breast cancer patients were retrospectively enrolled, and their clinical pathological and imaging data were analyzed. A combined prediction model was constructed by integrating intratumoral features and 3-mm peritumoral radiomic features, which was further compared with the single intratumoral model. The area under the curve (AUC) was calculated in the training and validation cohorts to quantify the discriminative ability of models. Calibration curves were plotted to assess the consistency between predicted probability and actual probability. Decision curve analysis (DCA) was performed to calculate the net clinical benefit across different threshold probabilities, so as to evaluate the clinical application value of the model. Results: The combined model yielded better predictive performance for pCR than the single intratumoral model. In the validation cohort, the combined model achieved an AUC of 0.851 (95% CI: 0.792~0.910), superior to 0.840 (95% CI: 0.779~0.901) of the single intratumoral model. Moreover, the combined model presented favorable calibration goodness-of-fit in both training and validation sets. DCA demonstrated that the combined model provided higher net clinical benefit for clinical decision-making. Conclusion: The combined model integrating intratumoral and peritumoral radiomic features can accurately predict NST response in breast cancer. It exhibits excellent performance in discrimination, calibration and clinical practicability, which can serve as a promising candidate tool for individualized pretreatment efficacy prediction. Nevertheless, this study is a retrospective single-center research, and further prospective multicenter external validation is required to confirm the generalizability of the findings.