人工智能辅助影像学评价局部晚期乳腺癌新辅助化疗效果

doi:10.3969/j.issn.1005-6483.20250887

摘要/Abstract

摘要： 目的探讨人工智能辅助影像学评价局部晚期乳腺癌新辅助化疗效果的应用价值。方法 2023年2月~2025年2月收治的局部晚期乳腺癌病人107例，采用表柔比星+环磷酰胺序贯多西他赛（EC-T）序贯新辅助化疗方案治疗，化疗前后均行核磁共振成像（MRI）检查，根据MillerPayne病理评估系统将病人分为有效组与无效组。比较两组病人化疗后MRI参数，采用多因素Logistic回归分析MRI各参数与新辅助化疗无效的关系。构建基于人工智能的MRI影像学评估模型，利用该模型对新辅助化疗效果进行评价。结果 107例病人中化疗有效76例（71.03%），无效31例（28.97%）；有效组肿瘤最大直径、表观扩散系数（ADC）、容量转移常数（Ktrans）、细胞外血管外间隙容积比（Ve）、速率常数（Kep）、最大强化率（Emax）、最大强化速率（Vmax）、对比增强率（CER）、达峰时间（Tmax）分别为（3.96±0.83）cm、（1.21±0.28）×10-3mm2/s、（0.41±0.07）min-1、（0.79±0.14）%、（0.52±0.15）min-1、（91.25±25.64）%、（0.55±0.12）SI/S、1.48±0.32、（301.36±40.78）s，无效组分别为（6.95±1.03）cm、（0.88±0.16）×10-3mm2/s、（0.88±0.19）min-1、（1.03±0.12）%、（0.85±0.23）min-1、（153.78±32.03）%、（0.91±0.24）SI/S、（1.64±0.35）、（169.92±34.71）s，两组比较差异有统计学意义（P＜0.05）；经多因素Logistic回归分析，肿瘤最大直径、Ktrans、Ve、Kep、Emax、Vmax、CER均是局部晚期乳腺癌新辅助化疗无效的危险因素（P＜0.05），ADC、CER是其保护因素（P＜0.05）。基于人工智能的影像学评估模型评价局部晚期乳腺癌新辅助化疗效果的灵敏度、特异度、阳性预测值、阴性预测值、准确度分别为93.55%、94.74%、87.88%、97.30%、94.39%，该模型评价结果与病理评估结果具有极强的一致性（Kappa=0.866，P＜0.05）。结论人工智能辅助影像学评价局部晚期乳腺癌新辅助化疗效果具有良好的应用价值。

关键词: 人工智能, 核磁共振成像, 乳腺癌, 新辅助化疗, 临床效果

Abstract: Objective To explore the application value of artificial intelligence assisted imaging in evaluating the effect of neoadjuvant chemotherapy for locally advanced breast cancer.Methods A total of 107 patients with locally advanced breast cancer were admitted from February 2023 to February 2025.107 patients with locally advanced breast cancer who were treated with EC-T (Epirubicin+Cyclophosphamide sequential Taxotere) sequential neoadjuvant chemotherapy were examined by magnetic resonance imaging (MRI) before and after chemotherapy.According to the Miller-Payne pathological evaluation system,patients were divided into an effective group and an ineffective group.The MRI parameters between the two groups after chemotherapy was compared,and the relationship between MRI parameters and ineffective neoadjuvant chemotherapy was investigated using multiple Logistic regression analysis.The MRI imaging evaluation model based on artificial intelligence was constructed,and this model was used to evaluate the effectiveness of neoadjuvant chemotherapy.Results 76 patients (71.03%) were effective in chemotherapy and 31 (28.97%) were ineffective among the 107 patients.The maximum tumor diameter,apparent diffusion coefficient (ADC),volume transfer constant (Ktrans),extracellular space volume percentage (Ve),rate constant (Kep),maximum enhancement ratio (Emax),maximum enhancement rate (Vmax),contrast enhancement rate (CER) and peak time (Tmax) in the effective group were (3.96±0.83)cm,(1.21±0.28)×10-3mm2/s,(0.41±0.07)min-1,(0.79±0.14)%,(0.52±0.15)min-1,(91.25±25.64)%,(0.55±0.12)SI/S,(1.48±0.32) and (301.36±40.78)s respectively，which in the ineffective group were (6.95±1.03)cm,(0.88±0.16)×10-3mm2/s,(0.88±0.19)min-1,(1.03±0.12)%,(0.85±0.23)min-1,(153.78±32.03)%,(0.91±0.24)SI/S,(1.64±0.35) and (169.92±34.71)s respectively,and the differences between the two groups were statistically significant (P<0.05).Multivariate logistic regression analysis showed that the maximum tumor diameter,Ktrans,Ve,Kep,Emax,Vmax and CER were risk factors for the ineffectiveness of neoadjuvant chemotherapy for locally advanced breast cancer (P<0.05),and ADC and CER were protective factors (P<0.05).The sensitivity,specificity,positive predictive value,negative predictive value and accuracy of the imaging evaluation model based on artificial intelligence to evaluate the effect of neoadjuvant chemotherapy for locally advanced breast cancer were 93.55%,94.74%,87.88%,97.30% and 94.39% respectively,and the evaluation result of this model was extremely strong consistency with pathological evaluation (Kappa=0.866,P<0.05).Conclusion Artificial intelligence assisted imaging has good application value in evaluating the effect of neoadjuvant chemotherapy for locally advanced breast cancer.

Key words: artificial intelligence, magnetic resonance imaging, breast cancer, neoadjuvant chemotherapy, clinical efficacy

张楠张亚芳周琪. 人工智能辅助影像学评价局部晚期乳腺癌新辅助化疗效果[J]. 临床外科杂志, 2026, 34(8): 876-879.

ZHANG Nan, ZHANG Yafang, ZHOU Qi. Evaluation of effect ofneoadjuvant chemotherapy for locally advanced breast cancer with artificial intelligence assisted imaging[J]. JOURNAL OF CLINICAL SURGERY, 2026, 34(8): 876-879.

参考文献

相关文章 15

[1]	郭世伟. 胰腺癌精准治疗现状和展望：从A到AI[J]. 临床外科杂志, 2026, 34(7): 746-751.
[2]	郭延洪，唐璟，卢世霞，包煜忠，赵雪梅，阿良德，刘震. 超声引导下胸神经阻滞与胸椎旁神经阻滞对乳腺癌根治术后麻醉效果的比较[J]. 临床外科杂志, 2026, 34(6): 645-648.
[3]	刘梓涛陈梦荷彭湃沈浩元胡超华. 趋化因子受体4及基质细胞衍生因子1在三阴性乳腺癌组织中的表达及意义[J]. 临床外科杂志, 2026, 34(5): 506-511.
[4]	韩静张祎潇杜战威马志军. 三阴性乳腺癌术后继发甲状腺髓样癌一例[J]. 临床外科杂志, 2026, 34(5): 592-593.
[5]	张倩徐心. 奥美定水凝胶注射隆乳术后并发HER-2阳性乳腺癌一例[J]. 临床外科杂志, 2026, 34(4): 400-401.
[6]	郭笑扬阮远李金岩周洪勇陈利胡子好柯炜. 基于磁共振-经直肠超声图像人工智能融合功能辅助前列腺穿刺活检在前列腺癌筛查中的应用[J]. 临床外科杂志, 2026, 34(4): 454-457.
[7]	李文张舟径孙大为. 破局甲状腺癌诊疗：人工智能的临床应用新范式[J]. 临床外科杂志, 2026, 34(3): 249-252.
[8]	张帆殷素鹏杨泽宇. 人工智能在甲状腺及颈部淋巴结超声诊断中的应用：现状、挑战与展望[J]. 临床外科杂志, 2026, 34(3): 253-257.
[9]	高徵泓徐波蔡文松. 超声弹性成像及人工智能辅助诊断在低危甲状腺乳头状癌主动监测中的应用现状[J]. 临床外科杂志, 2026, 34(3): 258-261.
[10]	夏海水房晓芳王伟马上. 血清单酰基甘油酯酶、甘油三磷酸脱氢酶1和人循环前胃泌素表达水平与乳腺癌病人分子分型及预后的关系[J]. 临床外科杂志, 2026, 34(3): 281-285.
[11]	马曼淇张乐鸿陈家乐凌子君高震坚黄倩瑜张梓程李慕鹏朱艺琳王赛凤刘民锋. 脂联素在三阴性乳腺癌发生发展中的作用及机制研究进展[J]. 临床外科杂志, 2026, 34(3): 351-354.
[12]	吴梓政魏元君郑杰韩猛. 趋化因子配体9 mRNA在三阴性乳腺癌中的表达及临床意义[J]. 临床外科杂志, 2026, 34(2): 150-155.
[13]	曹卓涵, 高冠英, 朱一川, 周湘, 徐雁. 股骨髋臼撞击综合征髋关节镜翻修手术的临床效果及影响因素分析[J]. 临床外科杂志, 2026, 34(1): 78-81.
[14]	黄瑛, 吴小蔚, 邓海峰, 靳茜雅, 何梦君, 贺胜, 吴海东, 章强. 改良缝纫式埋线法重睑成形术的临床应用[J]. 临床外科杂志, 2026, 34(1): 103-106.
[15]	王心怡韩丁培李鹤成. 机器人辅助支气管镜在肺结节诊治中的应用进展[J]. 临床外科杂志, 2025, 33(9): 922-927.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed