切换至 "中华医学电子期刊资源库"

中华普外科手术学杂志(电子版) ›› 2025, Vol. 19 ›› Issue (05) : 501 -505. doi: 10.3877/cma.j.issn.1674-3946.2025.05.007

所属专题: 文献

论著

DeepSurv深度学习模型辅助胃癌术后精准化疗策略研究
杨志, 夏雪峰(), 管文贤()   
  1. 210008 南京,南京大学医学院附属鼓楼医院胃肠外科
  • 收稿日期:2025-03-19 出版日期:2025-10-26
  • 通信作者: 夏雪峰, 管文贤

Research on precision chemotherapy strategy for gastric cancer after surgery assisted by deepsurv deep learning model

Zhi Yang, Xuefeng Xia(), Wenxian Guan()   

  1. Department of General Surgery, Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School, Nanjing Jiangsu Province 210008, China
  • Received:2025-03-19 Published:2025-10-26
  • Corresponding author: Xuefeng Xia, Wenxian Guan
  • Supported by:
    National Natural Science Foundation of China(82172645); Medical Research General Project of Jiangsu Provincial Health Commission(M2022096); Nanjing Drum Tower Hospital Clinical Research Special Fund Cultivation Project(2022-YXZX-XH-03)
引用本文:

杨志, 夏雪峰, 管文贤. DeepSurv深度学习模型辅助胃癌术后精准化疗策略研究[J/OL]. 中华普外科手术学杂志(电子版), 2025, 19(05): 501-505.

Zhi Yang, Xuefeng Xia, Wenxian Guan. Research on precision chemotherapy strategy for gastric cancer after surgery assisted by deepsurv deep learning model[J/OL]. Chinese Journal of Operative Procedures of General Surgery(Electronic Edition), 2025, 19(05): 501-505.

目的

利用人工智能技术,构建个体化的化疗反应评分系统,为胃癌患者精准治疗提供决策支持。

方法

回顾性分析SEER数据库2000年至2021年间诊断为胃癌并接受根治性手术治疗的11 478例患者。利用DeepSurv神经网络模型,将患者临床病理特征进行整合,建立预后预测模型。利用生存分析评估模型的预测性能,并采用基于排列的方法量化每个输入特征对模型预测结果的重要性。

结果

模型风险评分高的患者较评分低的患者具有显著更好的预后(HR=6.19,95% CI:5.83-6.58,Log-Rank P<0.01)。遵循模型化疗建议的患者群体(n=7 367例)相比不遵循建议的患者群体(n=4 111例)具有显著更好的预后(HR=0.46,95% CI:0.44-0.49,Log-Rank P<0.01)。变量重要性分析显示,淋巴结阳性比例、T分期及年龄是影响预后和化疗推荐的三个最重要因素。模型推荐化疗的比例随肿瘤分期的进展而增加。然而,模型推荐化疗的比例随患者年龄增加而增加。

结论

基于SEER数据库构建的DeepSurv模型不仅能够准确预测胃癌患者的预后,还能为精准化疗决策提供有价值的指导。

Objective

To construct an individualized chemotherapy response scoring system using artificial intelligence technology to provide decision support for precision treatment of gastric cancer patients.

Methods

A retrospective analysis was performed on 11 478 patients diagnosed with gastric cancer and treated with radical surgery in the SEER database between 2000 and 2021. The DeepSurv neural network model was used to integrate the clinicopathological characteristics of patients and establish a prognostic prediction model. Survival analysis was applied to evaluate the predictive performance of the model, and a permutation-based method was used to quantify the importance of each input feature for the model’s predictive results.

Results

Patients with high model risk scores showed significantly better prognosis than those with low scores (HR=6.19, 95% CI: 5.83-6.58, Log-Rank P<0.01). The patient group following the model’s chemotherapy recommendations (n=7 367) had significantly better prognosis than those not following the recommendations (n=4 111) (HR=0.46, 95% CI: 0.44-0.49, Log-Rank P<0.01). Variable importance analysis showed that the proportion of positive lymph nodes, T stage, and age were the three most important factors affecting prognosis and chemotherapy recommendations. The proportion of chemotherapy recommended by the model increased with the progression of tumor stage, while it also increased with patient age.

Conclusion

The DeepSurv model constructed based on the SEER database can not only accurately predict the prognosis of gastric cancer patients but also provide valuable guidance for precision chemotherapy decision-making.

图1 Deepsurv模型及治疗反应评分网络结构
表1 患者基线信息与临床征分析
临床特征 总例数(n=11 478) 接受化疗(n=6 443) 未接受化疗(n=5 035) 统计量 P
年龄(岁,±s 65.4± 13.4 68.5±13.04 61.4±12.8 t=29.4 <0.01
性别[例(%)]       χ2=1.3 0.25
6 686(58.3) 3 723(57.8) 2 963(58.8)    
4 792(41.7) 2 720(42.2) 2 072(41.2)    
种族[例(%)]       χ2=30.4 <0.01
白人 7 060(61.5) 4 039(62.7) 3 021(60.0)    
亚太人 2 911(25.4) 1 657(25.7) 1 254(24.9)    
黑人 1 507(13.1) 747(11.6) 760(15.1)    
形态学[例(%)]       χ2=74.7 <0.01
印戒细胞癌 2 465(21.5) 1 195(18.5) 1 270(25.2)    
非印戒细胞癌 9 013(78.5) 5 248(81.5) 3 765(74.8)    
T分期[例(%)]       χ2=1 659.0 <0.01
T1 3 194(27.8) 2 748(42.7) 446(8.9)    
T2 2 383(20.8) 1 209(18.8) 1 174(23.3)    
T3 3 168(27.6) 1 317(20.4) 1 851(36.8)    
T4 2 733(23.8) 1 169(18.1) 1 564(31.1)    
N分期[例(%)]       χ2=1 897.6 <0.01
N0 4 653(40.5) 3 746(58.1) 907(18.0)    
N1 2 539(22.1) 1 061(16.5) 1 478(29.4)    
N2 1 964(17.1) 732(11.4) 1 232(24.5)    
N3 2 322(20.2) 904(14.0) 1 418(28.2)    
M分期[例(%)]       χ2=46.3 <0.01
M0 10 297(89.7) 5 890(91.4) 4 407(87.5)    
M1 1 181(10.3) 553(8.6) 628(12.5)    
肿瘤分期[例(%)]       χ2=2 144.2 <0.01
3 787(33.0) 3 275(50.8) 512(10.2)    
2 408(21.0) 1 023(15.9) 1 385(27.5)    
3 753(32.7) 1 437(22.4) 2 316(46.0)    
1 530(13.3) 708(11.0) 822(16.3)    
阳性淋巴结数目MQ1Q3 1(0,7) 0(0,4) 4(1,10) Z=-Inf <0.01
阳性淋巴结比例MQ1Q3 0.10(0.00,0.47) 0.00(0.00,0.30) 0.25(0.06,0.57) Z=-35.4 <0.01
图2 Deepsurv模型的预后预测能力评估及变量重要性分析注:A为生存曲线图,按最佳切点将患者分为高危组及低危组,对两组生存预后进行比较。B为风险分数散点图,显示所有患者的风险分数分布;C为生存时间和状态散点图。D为条形图,表明模型预测结果中各种临床特征的重要性。
图3 治疗反应评分与实际治疗方案对患者预后影响的比较注:A为遵循与不遵循推荐治疗方案的患者预后比较;B为不推荐化疗患者中接受与未接受治疗的预后比较;C为推荐化疗患者中接受与未接受治疗的预后比较,D-F为I-IV期胃癌患者中,遵循模型推荐治疗方案的预后均优于不遵循推荐方案的预后
图4 治疗反应评分模型的可解释性分析注:A为条形图,展示各临床特征对治疗反应评分的影响程度;(B-E)为堆叠柱状图,展示在各个肿瘤分期(B)、T分期(C)、不同淋巴结阳性比例
[1]
Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries[J]. CA Cancer J Clin, 2021, 71(3): 209-249.
[2]
Smyth EC, Nilsson M, Grabsch HI, et al. Gastric cancer[J]. Lancet, 2020, 396(10251): 635-648.
[3]
徐瑜杰,赵国栋.晚期胃癌治疗方法的研究进展和挑战[J/CD].中华普外科手术学杂志(电子版), 2023,17(04): 451-455.
[4]
Guan WL, He Y, Xu RH. Gastric cancer treatment: recent progress and future perspectives [J]. J Hematol Oncol, 2023, 16(1): 57.
[5]
Ychou M, Boige V, Pignon JP, et al. Perioperative chemotherapy compared with surgery alone for resectable gastroesophageal adenocarcinoma: an FNCLCC and FFCD multicenter phase III trial [J]. J Clin Oncol, 2011, 29(13): 1715-1721.
[6]
Macdonald JS, Smalley SR, Benedetti J, et al.Chemoradiotherapy after surgery compared with surgery alone for adenocarcinoma of the stomach or gastroesophageal junction [J]. N Engl J Med, 2001, 345(10): 725-730.
[7]
李磊,费建东,宋利琴, 等 新辅助化疗结合腹腔镜胃癌根治术治疗进展期胃癌的生存质量及生存率分析[J/CD]. 中华普外科手术学杂志(电子版), 2020, 14(01): 66-69.
[8]
Brianna, Lee SH. Chemotherapy: how to reduce its adverse effects while maintaining the potency? [J]. Med Oncol, 2023, 40(3): 88.
[9]
黄雨琳,王浩云,李燕梅, 等. 胃癌患者化疗期间症状群的范围综述[J]. 中国全科医学: 2024,1-7.
[10]
Amin B, Greene FL, Edge SB, et al. The Eighth Edition AJCC Cancer Staging Manual: Continuing to build a bridge from a population-based to a more "personalized" approach to cancer staging [J]. CA Cancer J Clin, 2017, 67(2): 93-99.
[11]
Tseng PT, Cheng YS, Chen YW, et al. Peripheral iron levels in children with autism spectrum disorders vs controls: a systematic review and meta-analysis [J]. Nutr Res, 2018, 50: 44-52.
[12]
Zeng J, Li K, Cao F, et al. Development and validation of survival prediction model for gastric adenocarcinoma patients using deep learning: A SEER-based study[J]. Front Oncol, 2023, 13: 1131859.
[13]
Jin P, Ji X, Kang W, et al. Artificial intelligence in gastric cancer: a systematic review[J]. J Cancer Res Clin Oncol, 2020, 146(9): 2339-2350.
[14]
She Y, Jin Z, Wu J, et al. Development and Validation of a Deep Learning Model for Non-Small Cell Lung Cancer Survival[J]. JAMA Netw Open, 2020, 3(6): e205842.
[15]
Katzman JL, Shaham U, Cloninger A, et al. DeepSurv: personalized treatment recommender system using a Cox proportional hazards deep neural network[J]. BMC Med Res Methodol, 2018, 18(1): 24.
[16]
Fisher A, Rudin C, Dominici F. All Models are Wrong, but Many are Useful: Learning a Variable’s Importance by Studying an Entire Class of Prediction Models Simultaneously[J]. J Mach Learn Res, 2019, 20.
[17]
Athauda A, Nankivell M, Langley RE, et al. Impact of sex and age on chemotherapy efficacy, toxicity and survival in localised oesophagogastric cancer: A pooled analysis of 3265 individual patient data from four large randomised trials (OE02, OE05, MAGIC and ST03) [J]. Eur J Cancer, 2020,137: 45-56.
[18]
Sylvie L, Silvia S, Salah-Eddin AB, et al. Impact of age on the feasibility and efficacy of neoadjuvant chemotherapy in patients with locally advanced oesophagogastric cancer [J]. Eur J Cancer, 2015, 51(14): 1918-26.
[1] 中华医学会外科学分会腹腔镜与内镜外科学组, 中国抗癌协会胃癌专业委员会, 中国抗癌协会腔镜与机器人外科分会, 《中华消化外科杂志》编辑委员会. 精准外科胃癌转化治疗专家共识(2025 版)[J/OL]. 中华普通外科学文献(电子版), 2025, 19(03): 145-156.
[2] 王思卓, 段晓鑫, 陈隆, 董胜利. 肠道微生物群、血液代谢物和胃癌的因果关系:东亚人群中介孟德尔随机化研究[J/OL]. 中华普通外科学文献(电子版), 2025, 19(03): 163-168.
[3] 陈隆, 段晓鑫, 王思卓, 董胜利. 胃癌免疫治疗的现状[J/OL]. 中华普通外科学文献(电子版), 2025, 19(03): 177-182.
[4] 陈使功, 王文静, 黎秀兰, 龙勃, 焦作义. 精氨酸和谷氨酰胺在胃癌治疗中的应用及研究进展[J/OL]. 中华普通外科学文献(电子版), 2025, 19(03): 183-187.
[5] 王小军, 蔡瑜, 安艳新, 刘斌, 冯永安. 完全腹腔镜远端胃癌根治术治疗局部进展期胃癌的临床研究[J/OL]. 中华普外科手术学杂志(电子版), 2025, 19(05): 509-512.
[6] 王浩瑜, 杨拴元, 任彦顺, 阴志强. 两种改良食管空肠吻合术的腹腔镜全胃切除术中临床对比研究[J/OL]. 中华普外科手术学杂志(电子版), 2025, 19(04): 425-428.
[7] 林炳涛, 陈君填. 不同病理类型进展期胃癌患者临床特征及腹腔镜辅助根治术后短期预后的影响因素[J/OL]. 中华普外科手术学杂志(电子版), 2025, 19(04): 429-433.
[8] 杨维军, 张文文, 付凯. 单切口腹腔镜下胃癌根治术治疗早中期胃癌的疗效观察[J/OL]. 中华普外科手术学杂志(电子版), 2025, 19(04): 434-437.
[9] 李华志, 孙海涛, 曹广, 张雅静. 基于膜解剖的完整系膜切除+D2根治术在进展期胃癌治疗中的应用[J/OL]. 中华普外科手术学杂志(电子版), 2025, 19(03): 262-265.
[10] 汪志翔, 何战洋. 不同淋巴结清扫术在No.16淋巴结局限性转移的局部进展期胃癌中的临床研究[J/OL]. 中华普外科手术学杂志(电子版), 2025, 19(03): 266-269.
[11] 申昌军, 刘文奇. 不同食管空肠重建方法在胃癌患者TLTG术食管空肠吻合术中的应用对比[J/OL]. 中华普外科手术学杂志(电子版), 2025, 19(03): 270-273.
[12] 李鹏, 刘光世, 李涛. 基于黑色素瘤相关抗原A6在胃癌转移与预后的作用机制研究[J/OL]. 中华普外科手术学杂志(电子版), 2025, 19(03): 282-284.
[13] 李玉庆, 王刚, 李强, 李云川, 杜明新. 不同吻合术对腹腔镜远端胃癌根治术患者的影响[J/OL]. 中华普外科手术学杂志(电子版), 2025, 19(03): 286-289.
[14] 艾世超, 孙艺文, 宋鹏, 沈晓菲, 刘颂, 孙锋, 陆晓峰, 王萌, 管文贤. 静脉注射吲哚菁绿导航胃癌根治术的单臂开放前瞻性研究[J/OL]. 中华普外科手术学杂志(电子版), 2025, 19(02): 166-169.
[15] 廖俊豪, 周理林, 曾健文. 广东省医学会泌尿外科疑难病例多学科会诊(第25期)——膀胱癌根治术后盆腔复发[J/OL]. 中华腔镜泌尿外科杂志(电子版), 2025, 19(03): 390-394.
阅读次数
全文


摘要


AI


AI小编
你好!我是《中华医学电子期刊资源库》AI小编,有什么可以帮您的吗?