摘要
目的探讨基于双模态超声图像的影像组学模型预测乳腺癌肿瘤浸润淋巴细胞(TIL)水平的应用价值。方法回顾性收集2018年1月至2024年6月丽水市人民医院经术后病理检查确诊的211例乳腺癌患者,以7∶3比例随机分为训练集148例和验证集63例。所有患者均在术前接受灰阶超声(US)及应变式弹性成像(SE)检查,并根据病理检查结果分为低TIL水平(<10%)组和高TIL水平(≥10%)组。收集患者的临床病理资料,并分析超声图像特征。筛选基于US和SE的影像组学特征,并构建3种影像组学模型,包括US、SE和US+SE模型。采用多因素logistic回归筛选独立影响因素并结合最优影像组学评分建立联合模型,绘制列线图。采用ROC曲线、决策曲线和校准曲线评价不同模型的诊断效能、临床净获益和一致性。结果在训练集和验证集中,US+SE模型的AUC均为最高,选择其作为最优影像组学模型。基于形态、后方回声及US+SE影像组学评分建立联合模型,绘制列线图。ROC曲线分析结果显示,联合模型的预测效能较好,在训练集和验证集中的AUC分别为0.934、0.898。决策曲线分析表明,当训练集和验证集的风险阈值分别在0.49~0.78、0.33~0.74时,联合模型具有更好的临床净获益。校准曲线分析显示,联合模型的预测概率与实际概率具有良好的一致性。结论基于双模态超声图像影像组学的联合模型可在术前较好地预测乳腺癌TIL水平,为临床提供准确、全面、高效的评估手段。
Objective To evaluate the value of a radiomics model based on dual-modality ultrasound(US)images for preoperative prediction of tumor infiltrating lymphocyte(TIL)levels in breast cancer.Methods This retrospective study enrolled 211 patients with pathologically confirmed breast cancer in Lishui People's Hospital from January 2018 to June 2024.Patients were randomly divided into a training set(n=148)and a validation set(n=63)in a 7∶3 ratio.All patients underwent preoperative gray-scale US and strain elastography(SE).Based on postoperative pathological assessment,patients were categorized into low-TIL(<10%)and high-TIL(≥10%)groups.Clinical-pathological data and conventional US features were analyzed.Radiomics features were extracted from US and SE images.Three radiomics models were constructed using features from US alone,SE alone,and combined US+SE.Multivariate logistic regression identified independent predictors,which were then integrated with the optimal radiomics score to build a combined nomogram model.The diagnostic performance,net clinical benefit,and calibration of the models were evaluated using ROC curves,decision curve analysis(DCA),and calibration curves,respectively.Results The US+SE radiomics model demonstrated the highest AUC in both the training and validation sets and was therefore selected as the optimal radiomics model.A nomogram was subsequently developed by incorporating morphological features,posterior echo pattern,and the US+SE radiomics score.The combined nomogram showed good predictive performance,with AUC of 0.934 in the training set and 0.898 in the validation set.DCA indicated that the nomogram provided superior net clinical benefit within threshold probability ranges of 0.49-0.78 for the training set and 0.33-0.74 for the validation set.Calibration curves demonstrated good agreement between the nomogram's predicted probabilities and actual outcomes.Conclusion The radiomics nomogram based on dual-modality US images can effectively predict TIL levels in breast cancer preoperatively,offering an accurate,comprehensive,and efficient tool for clinical assessment.
作者
毛雅
蔡云奇
练宏欢
李莹
MAO Ya;CAI Yunqi;LIAN Honghuan;LI Ying(Department of Ultrasonography,Lishui People's Hospital,Lishui 323000,China;不详)
出处
《浙江医学》
2025年第23期2521-2526,2531,I0002,共8页
Zhejiang Medical Journal
