2025 , Vol. 22 >Issue 02: 97 - 105
DOI: https://doi.org/10.3877/cma.j.issn.1672-6448.2025.02.002
基于超声的深度学习列线图预测乳腺癌新辅助化疗后腋窝淋巴结状态的研究
Copy editor: 汪荣
收稿日期: 2024-12-07
网络出版日期: 2025-04-01
基金资助
南京市博士后科研资助计划(291937)南京医科大学第一附属医院国自然科学基金青年基金培育计划(PY2021043)江苏省科学技术厅基础研究计划自然科学基金——青年基金项目(BK20241122)南京邮电大学-江苏省人民医院联合开放课题资助和江苏省人民医院医工交叉转化基金资助(RG202412)
版权
Ultrasound-based deep learning nomogram for predicting axillary lymph node status after neoadjuvant chemotherapy for breast cancer
Received date: 2024-12-07
Online published: 2025-04-01
Copyright
目的
探讨基于超声的深度学习列线图预测乳腺癌新辅助化疗(NAC)后腋窝淋巴结(ALN)状态的价值。
方法
回顾性选取2020 年3 月至2023 年6 月在南京医科大学第一附属医院(训练集,n=257)和解放军东部战区总医院(外部测试集,n=157)接受NAC 的414 例ALN 转移的乳腺癌患者,并根据NAC 后ALN 手术病理结果分为病理完全缓解(pCR)组和非病理完全缓解(npCR)组。使用NAC 前的乳腺肿瘤二维超声图像训练并构建基于ResNet50 架构的深度学习模型;通过单因素和多因素Logistic 回归分析临床病理特征,筛选出与NAC 后ALN 病理状态有关的独立危险因素,构建临床模型;联合独立危险因素与深度学习预测概率构建深度学习列线图。利用受试者工作特征(ROC)曲线、校准曲线、决策曲线分析、临床影响曲线评估模型性能。由2 位不同年资超声医师根据超声图像及NAC 前免疫组化结果对外部测试集进行独立预测,并在深度学习列线图的辅助下进行第二次预测,比较分析两次预测结果。
结果
临床病理资料中,雌激素受体(ER)和人表皮生长因子受体2(Her-2)是预测NAC 后ALN 病理状态的独立危险因素。临床模型、深度学习模型和深度学习列线图的曲线下面积(AUC)在训练集中分别为0.724、0.872、0.878,在外部测试集中分别为0.698、0.831、0.859。深度学习列线图的预测效能优于临床模型(训练集、外部测试集中P 值均<0.001),且在外部测试集中其优于深度学习模型(P=0.024)。医师1(低年资)和医师2(高年资)独立判断的AUC 值分别为0.570、0.606,均低于深度学习模型和深度学习列线图(P 均<0.001)。在深度学习列线图的帮助下,医师1 和医师2 的诊断能力AUC 分别提升至0.796 和0.807,与独立判断比较,差异均有统计学意义(P 均<0.001)。
结论
基于NAC 前超声图像的深度学习列线图可以在治疗前有效预测乳腺癌NAC 后ALN 病理状态,为个性化治疗方案的制定提供更多依据。
孙舒涵 , 陈雅静 , 宗晴晴 , 栗翠英 , 缪殊妹 , 杨斌 , 俞飞虹 . 基于超声的深度学习列线图预测乳腺癌新辅助化疗后腋窝淋巴结状态的研究[J]. 中华医学超声杂志(电子版), 2025 , 22(02) : 97 -105 . DOI: 10.3877/cma.j.issn.1672-6448.2025.02.002
Objective
To assess the value of a deep learning-based nomogram in predicting axillary lymph node (ALN) status following neoadjuvant chemotherapy (NAC) in breast cancer patients.
Methods
Four hundred and fourteen ALN-positive breast cancer patients who received NAC between March 2020 and June 2023 were enrolled in this retrospective study and divided into a training set and an external test set.The training set consisted of 257 patients from the First Affiliated Hospital of Nanjing Medical University, while the external test set included 157 patients from the General Hospital of Eastern Theater Command.All patients were divided into pathologically complete response (pCR) and non-pCR(npCR) groups based on the pathology results of ALN surgery post-NAC.A deep learning model based on the ResNet50 architecture was trained and established using pre-NAC ultrasound images of breast tumors.Univariate and multivariate logistic regression analyses were performed on the training set to identify independent risk factors for post-NAC ALN status.These independent risk factors were then used to construct a clinical model.A deep learning-based nomogram was constructed by combining independent risk factors and deep learning predictive probabilities.The performance of the models was evaluated using receiver operating characteristic (ROC) curve, calibration curve, decision curve analysis, and clinical impact curve.Two radiologists with different experience levels independently predicted ALN status in the external test set based on ultrasound images and pre-NAC immunohistochemical results, and performed a second prediction with the assistance of the deep learning-based nomogram.The two prediction results were compared.
Results
Estrogen receptor (ER) and human epidermal growth factor receptor 2 (Her-2) were identifled as independent risk factors for predicting post-NAC ALN status.The area under the curve (AUC) values of the clinical model, deep learning model, and deep learning-based nomogram were 0.724, 0.872, and 0.878 in the training set, and 0.698, 0.831, and 0.859 in the external test set, respectively.The deep learning-based nomogram outperformed the clinical model (both P<0.001 in training and external test sets) and showed superior performance to the deep learning model in the external test set (P=0.024).The AUC values of radiologist 1 (low-experience) and radiologist 2 (high-experience) for independent judgment were 0.570 and 0.606, respectively, both signiflcantly lower than those of the deep learning model and the deep learningbased nomogram (all P<0.001).With the assistance of the deep learning-based nomogram, the AUC values of radiologist 1 and radiologist 2 improved to 0.796 and 0.807, respectively, showing statistically signiflcant differences compared to independent judgment (both P<0.001).
Conclusion
The deep learning-based nomogram based on pre-NAC ultrasound images can effectively predict the pathological status of ALN in breast cancer patients after NAC treatment, providing more evidence for the development of personalized treatment plans.
表1 训练集与外部测试集临床病理资料比较[例(%)] |
资料 | 训练集(n=257) | 外部测试集(n=157) | χ 2 值 | P 值 |
---|---|---|---|---|
年龄 | 0.140 | 0.708 | ||
< 50 岁 | 126(49.0) | 74(47.1) | ||
≥50 岁 | 131(51.0) | 83(52.9) | ||
ER | 1.896 | 0.169 | ||
阴性 | 105(40.9) | 75(47.8) | ||
阳性 | 152(59.1) | 82(52.2) | ||
PR | 3.298 | 0.069 | ||
阴性 | 137(53.3) | 98(62.4) | ||
阳性 | 120(46.7) | 59(37.6) | ||
Her-2 | 0.826 | 0.363 | ||
阴性 | 146(56.8) | 82(52.2) | ||
阳性 | 111(43.2) | 75(47.8) | ||
Ki-67 | 0.037 | 0.847 | ||
低表达 | 21(8.2) | 12(7.6) | ||
高表达 | 236(91.8) | 145(92.4) | ||
分子分型 | 5.597 | 0.133 | ||
Luminal A 型 | 14(5.4) | 5(3.2) | ||
Luminal B 型 | 148(57.6) | 78(49.7) | ||
Her-2 + 型 | 47(18.3) | 42(26.8) | ||
三阴性型 | 48(18.7) | 32(20.3) |
注:ER 为雌激素受体;PR 为孕激素受体;Her-2 为人表皮生长因子受体2;Luminal 型为管腔型 |
表2 训练集中ALN pCR 的单因素Logistic 回归分析结果(n=257) |
因素 | β 值 | SE | Wald χ2 值 | P 值 | OR 值 | 95%CI |
---|---|---|---|---|---|---|
年龄 | 0.183 | 0.251 | 0.527 | 0.468 | 1.200 | 0.733 ~ 1.965 |
ER | -1.089 | 0.263 | 17.132 | < 0.001 | 0.337 | 0.201 ~ 0.564 |
PR | -0.721 | 0.257 | 7.895 | 0.005 | 0.486 | 0.294 ~ 0.804 |
Her-2 | 1.376 | 0.267 | 26.578 | < 0.001 | 3.958 | 2.346 ~ 6.677 |
Ki-67 | 0.281 | 0.468 | 0.362 | 0.548 | 1.325 | 0.529 ~ 3.316 |
分子分型 | 0.254 | 0.148 | 2.959 | 0.085 | 1.289 | 0.965 ~ 1.721 |
注:ER 为雌激素受体;PR 为孕激素受体;Her-2 为人表皮生长因子受体2;CI 为置信区间;ALN pCR 为腋窝淋巴结病理完全缓解 |
表3 训练集中ALN pCR 的多因素Logistic 回归分析结果(n=257) |
因素 | β 值 | SE | Wald χ2 值 | P 值 | OR 值 | 95%CI |
---|---|---|---|---|---|---|
ER | -1.082 | 0.342 | 9.996 | 0.002 | 0.339 | 0.173 ~ 0.663 |
PR | 0.084 | 0.342 | 0.061 | 0.805 | 1.088 | 0.557 ~ 2.125 |
Her-2 | 1.342 | 0.278 | 23.282 | < 0.001 | 3.828 | 2.219 ~ 6.603 |
注:ER 为雌激素受体;PR 为孕激素受体;Her-2 为人表皮生长因子受体2;CI 为置信区间;ALN pCR 为腋窝淋巴结病理完全缓解 |
表4 三种模型预测NAC 后ALN pCR 的效能分析 |
模型 | AUC(95%CI) | 敏感度(%) | 特异度(%) | 准确性(%) | 阳性预测值(%) | 阴性预测值(%) |
---|---|---|---|---|---|---|
训练集(n=257) | ||||||
临床模型 | 0.724(0.664 ~ 0.784) | 86.0 | 54.5 | 68.5 | 60.1 | 83.0 |
深度学习模型 | 0.872(0.830 ~ 0.914) | 83.3 | 74.1 | 78.2 | 72.0 | 84.8 |
深度学习列线图 | 0.878(0.838 ~ 0.918) | 86.8 | 74.1 | 79.8 | 72.8 | 87.6 |
外部测试集(n=157) | ||||||
临床模型 | 0.698(0.618 ~ 0.777) | 85.5 | 44.3 | 62.4 | 54.6 | 79.6 |
深度学习模型 | 0.831(0.769 ~ 0.893) | 79.7 | 71.6 | 75.2 | 68.8 | 81.8 |
深度学习列线图 | 0.859(0.801 ~ 0.916) | 84.1 | 73.9 | 78.3 | 71.6 | 85.5 |
注:NAC 为新辅助化疗;ALN 为腋窝淋巴结;pCR 为病理完全缓解;AUC 为曲线下面积;CI 为置信区间 |
表5 不同分子分型组深度学习列线图预测NAC 后ALN pCR 的效能分析 |
组别 | 例数 | AUC(95%CI) | 敏感度(%) | 特异度(%) | 准确性(%) | 阳性预测值(%) | 阴性预测值(%) |
---|---|---|---|---|---|---|---|
Luminal 型组 | |||||||
训练集 | 162 | 0.914(0.873 ~ 0.956) | 90.3 | 79.0 | 83.3 | 72.7 | 92.9 |
外部测试集 | 83 | 0.816(0.718 ~ 0.915) | 83.3 | 71.2 | 74.7 | 54.1 | 91.3 |
非Luminal 型组 | |||||||
训练集 | 95 | 0.854(0.779 ~ 0.929) | 84.6 | 72.1 | 78.9 | 78.6 | 79.5 |
外部测试集 | 74 | 0.848(0.758 ~ 0.938) | 82.2 | 75.9 | 79.7 | 84.1 | 73.3 |
注:NAC 为新辅助化疗;ALN 为腋窝淋巴结;pCR 为病理完全缓解;Luminal 型为管腔型;AUC 为曲线下面积;CI 为置信区间 |
1 |
Gradishar WJ, Anderson BO, Abraham J, et al.Breast cancer, version 3.2020, NCCN clinical practice guidelines in oncology[J].J Natl Compr Canc Netw, 2020, 18(4):452-478.
|
2 |
Mamounas EP, Anderson SJ, Dignam JJ, et al.Predictors of locoregional recurrence after neoadjuvant chemotherapy:results from combined analysis of national surgical adjuvant breast and bowel project B-18 and B-27[J].J Clin Oncol, 2012, 30(32):3960-3966.
|
3 |
Pilewskie M, Morrow M.Axillary nodal management following noadjuvant chemotherapy:a review [J].JAMA Oncol, 2017, 3(4):549-555.
|
4 |
Boughey JC, McCall LM, Ballman KV, et al.Tumor biology correlates with rates of breast-conserving surgery and pathologic complete response after neoadjuvant chemotherapy for breast cancer:flndings from the ACOSOG Z1071 (Alliance) Prospective Multicenter Clinical Trial[J].Ann Surg, 2014, 260(4):608-614.
|
5 |
Wang W, Wang X, Liu J, et al.Nomogram for predicting axillary lymph node pathological response in node-positive breast cancer patients after neoadjuvant chemotherapy[J].Chin Med J, 2022, 135(3):333-340.
|
6 |
Chang JM, Kim R, Lee HB, et al.Predicting axillary response to neoadjuvant chemotherapy:breast MRI and US in patients with nodepositive breast cancer[J].Radiology, 2019, 293(1):49-57.
|
7 |
LeCun Y, Bengio Y, Hinton G.Deep learning[J].Nature, 2015,521(7553):436-444.
|
8 |
Cui Y, Zhang J, Li Z, et al.A CT-based deep learning radiomics nomogram for predicting the response to neoadjuvant chemotherapy in patients with locally advanced gastric cancer:A multicenter cohort study[J].EClinicalMedicine, 2022, 46:101348.
|
9 |
Shao Y, Dang Y, Cheng Y, et al.Predicting the efficacy of neoadjuvant cemotherapy for pancreatic cancer using deep learning of contrastenhanced ultrasound videos[J].Diagnostics, 2023, 13(13):2183.
|
10 |
Leng X, Amidi E, Kou S, et al.Rectal cancer treatment management:deep-learning neural network bsed on photoacoustic microscopy image outperforms histogram-feature-based classification[J].Front Oncol,2021, 11:715332.
|
11 |
Zhang B, Yu Y, Mao Y, et al.Development of MRI-based deep learning signature for prediction of axillary response after NAC in breast cancer[J].Acad Radiol, 2024, 31(3):800-811.
|
12 |
Li Z, Gao J, Zhou H, et al.Multiregional dynamic contrast-enhanced MRI-based integrated system for predicting pathological complete response of axillary lymph node to neoadjuvant chemotherapy in breast cancer:multicentre study[J].eBioMedicine, 2024, 107:105311.
|
13 |
免疫组织化学在乳腺病理中的应用共识(2022 版)[J].中华病理学杂志, 2022, 51(9):803-811.
|
14 |
Corben AD, Abi-Raad R, Popa I, et al.Pathologic response and longterm follow-up in breast cancer patients treated with neoadjuvant chemotherapy:a comparison between classiflcations and their practical application[J].Arch Pathol Lab Med, 2013, 137(8):1074-1082.
|
15 |
于丹阳, 吴桐, 荆慧, 等.乳腺癌多模态超声评估新辅助化疗后腋窝淋巴结病理状态的研究[J].中华超声影像学杂志, 2022, 31(8):685-690.
|
16 |
李雨漫, 梁星宇, 吴桐, 等.双模态影像联合免疫组化构建模型预测cN_1 期乳腺癌新辅助化疗后腋窝淋巴结状态[J].中华超声影像学杂志, 2023, 32(8):699-706.
|
17 |
Gu J, Tong T, Xu D, et al.Deep learning radiomics of ultrasonography for comprehensively predicting tumor and axillary lymph node status after neoadjuvant chemotherapy in breast cancer patients:A multicenter study[J].Cancer, 2023, 129(3):356-366.
|
18 |
洪玮, 叶细容, 刘枝红, 等.超声影像组学联合临床病理特征预测乳腺癌新辅助化疗完全病理缓解的价值[J/OL].中华医学超声杂志(电子版), 2024, 21(6):571-579.
|
19 |
Barron AU, Hoskin TL, Day CN, et al.Association of low nodal positivity rate among patients with ERBB2-positive or triple-negative breast cancer and breast pathologic complete response to neoadjuvant chemotherapy[J].JAMA Surg, 2018, 153(12):1120-1126.
|
20 |
Slanetz PJ, Moy L, Baron P, et al.ACR appropriateness criteria®monitoring response to neoadjuvant systemic therapy for breast cancer[J].J Am Coll Radiol, 2017, 14(11):S462-S475.
|
21 |
Samiei S, de Mooij CM, Lobbes MBI, et al.Diagnostic performance of noninvasive imaging for assessment of axillary response after neoadjuvant systemic therapy in clinically node-positive breast cancer:a systematic review and meta-analysis[J].Ann Surg, 2021, 273(4):694.
|
22 |
Zheng Y, Qiu B, Liu S, et al.A transformer-based deep learning model for early prediction of lymph node metastasis in locally advanced gastric cancer after neoadjuvant chemotherapy using pretreatment CT images[J].EClinicalMedicine, 2024, 75:102805.
|
23 |
Jiang M, Li CL, Luo XM, et al.Ultrasound-based deep learning radiomics in the assessment of pathological complete response to neoadjuvant chemotherapy in locally advanced breast cancer[J].Eur J Cancer, 2021, 147:95-105.
|
24 |
Mamtani A, Barrio AV, King TA, et al.How often does neoadjuvant chemotherapy avoid axillary dissection in patients with histologically confirmed nodal metastases:results of a prospective study[J].Ann Surg Oncol, 2016, 23(11):3467-3474.
|
25 |
Yu FH, Miao SM, Li CY, et al.Pretreatment ultrasound-based deep learning radiomics model for the early prediction of pathologic response to neoadjuvant chemotherapy in breast cancer[J].Eur Radiol,2023, 33(8):5634-5644.
|
/
〈 |
|
〉 |