切换至 "中华医学电子期刊资源库"

中华腔镜泌尿外科杂志(电子版) ›› 2024, Vol. 18 ›› Issue (06) : 597 -605. doi: 10.3877/cma.j.issn.1674-3253.2024.06.011

临床研究

基于盆底彩超的人工智能模型在女性压力性尿失禁分度诊断中的应用
黄俊龙1, 李文双1, 李晓阳1, 刘柏隆1, 陈逸龙2, 丘惠平2, 周祥福1,()   
  1. 1.510630 广州,中山大学附属第三医院泌尿外科
    2.510000 广州,广州天鹏计算机科技有限公司
  • 收稿日期:2024-10-18 出版日期:2024-12-01
  • 通信作者: 周祥福
  • 基金资助:
    广州市科技计划项目(202103000035)广东省泌尿系统疾病临床医学中心项目(2020B1111170006)中山大学附属第三医院临床研究专项基金远航计划项目(YHJH202205)广州地区临床高新、重大和特色技术项目(2023P-TS34)

Application of artificial intelligence model based on pelvic floor ultrasound for the diagnosis of female stress urinary incontinence

Junlong Huang1, Wenshuang Li1, Xiaoyang Li1, bolong Liu1, Yilong Chen2, Huiping Qiu2, Xiangfu Zhou1,()   

  1. 1.Department of Urology,the Third Affliated Hospital of Sun Yat-sen University,Guangzhou 510630,China
    2.Guangzhou Tianpeng Computer Technology Co.,Ltd 510000,China
  • Received:2024-10-18 Published:2024-12-01
  • Corresponding author: Xiangfu Zhou
引用本文:

黄俊龙, 李文双, 李晓阳, 刘柏隆, 陈逸龙, 丘惠平, 周祥福. 基于盆底彩超的人工智能模型在女性压力性尿失禁分度诊断中的应用[J]. 中华腔镜泌尿外科杂志(电子版), 2024, 18(06): 597-605.

Junlong Huang, Wenshuang Li, Xiaoyang Li, bolong Liu, Yilong Chen, Huiping Qiu, Xiangfu Zhou. Application of artificial intelligence model based on pelvic floor ultrasound for the diagnosis of female stress urinary incontinence[J]. Chinese Journal of Endourology(Electronic Edition), 2024, 18(06): 597-605.

目的

基于Valsalva状态下的二维经会阴盆底超声图像,开发深度学习人工智能模型,以期在临床实践中辅助女性压力性尿失禁(SUI)的分度诊断。

方法

所有患者在就诊后1月内完成经会阴盆底彩超检查。根据国际尿失禁咨询委员会简表问卷(ICI-Q-SF),SUI患者被分为轻度(S1)、中度(S2)、重度(S3)三组。收集患者基础信息和超声图像,并选择Valsalva状态下的盆底正中矢状切面图,图像经预处理后用计算机算法半自动分割感兴趣区域(ROI)。建模采用Imagenet预训练的ResNet50模型进行迁移学习,数据按8∶2随机分为训练集和测试集,训练集进行五折交叉验证。模型训练完成后,通过混淆矩阵和受试者工作特征曲线(ROC)下面积(AUC)评估性能。最后,应用梯度加权类激活映射(Grad-CAM)生成热图,直观展示模型关注的特征区域。

结果

本研究最终纳入282例患者,其中SUI患者167例,非SUI患者115例(S0)。模型在训练集和测试集上的准确率分别为84.1%和83.9%,训练集AUC值分别为无SUI 0.89(95%CI:0.78~0.97)、轻度SUI 0.96(95%CI:0.90~1.00)、中度SUI 0.92(95%CI:0.81~1.00)和重度SUI 0.93(95%CI:0.84~0.98)。测试集AUC值分别为无SUI 0.88(95%CI:0.77~0.98)、轻度SUI 0.91(95%CI:0.80~0.99)、中度SUI 0.92(95%CI:0.80~1.00)、重度SUI 0.88(95%CI:0.75~0.98)。

结论

基于Valsalva状态下的二维经会阴盆底超声图像构建的SUI分度诊断人工智能模型有望成为SUI患者分度诊断的客观依据。

Objective

Based on two-dimensional transperineal pelvic floor ultrasound images obtained during the Valsalva maneuver,this study aims to develop a deep learning artificial intelligence model to assist in the clinical grading diagnosis of female stress urinary incontinence (SUI).

Methods

All patients underwent transperineal pelvic floor ultrasound examination within one month of their initial visit. According to the International Consultation on Incontinence Questionnaire-Short Form (ICI-QSF),the SUI patients were divided into three groups: mild (S1),moderate (S2),and severe (S3). Baseline information and ultrasound images of the patients were collected,and midsagittal plane images of the pelvic floor during the Valsalva maneuver were selected. After preprocessing,the regions of interest(ROI) were semi-automatically segmented using computer algorithms. A transfer learning approach was applied using a ResNet50 model pretrained on ImageNet. The data were randomly divided into a training set and a test set in an 8∶2 ratio,with five-fold cross-validation performed on the training set. After model training,performance was evaluated using a confusion matrix and the area under the receiver operating characteristic curve (AUC). Finally,gradient-weighted class activation mapping (Grad-CAM) was used to generate heatmaps to visually demonstrate the areas of interest identified by the model.

Results

A total of 282 patients were included in the study,comprising 167 patients with SUI and 115 non-SUI patients(S0). The model achieved an accuracy of 84.1% in the training set and 83.9% in the test set. The AUC values in the training set were as follows: non-SUI 0.89 (95%CI: 0.78-0.97),mild SUI 0.96 (95%CI: 0.90-1.00),moderate SUI 0.92 (95%CI: 0.81-1.00),and severe SUI 0.93 (95%CI: 0.84-0.98). In the test set,the AUC values were non-SUI 0.88 (95%CI: 0.77-0.98),mild SUI 0.91 (995%CI: .80-0.99),moderate SUI 0.92 (95%CI: 0.80-1.00),and severe SUI 0.88 (95%CI: 0.75-0.98).

Conclusion

The grading diagnostic artificial intelligence model based on two-dimensional transperineal pelvic floor ultrasound images during the Valsalva maneuver shows potential to become an objective basis for the grading diagnosis of SUI patients.

图1 患者入选流程图
图2 基于盆底超声的SUI分度诊断ResNet50模型构建流程简图及Grad-CAM热图可视化 注:a为图像预处理及感兴趣区域的半自动分割,图中数字为图像的像素尺寸;b为ResNet50模型示意图,图中的数字为图像像素尺寸×特征图的数量;c为Grad-CAM热图,分别展示了模型在不同卷积层(layer1、layer2、layer3、layer4)上的激活情况
表1 行盆底彩超检查的训练集和测试集患者特征比较[例(%)]
组别 例数 年龄(岁) BMI(kg/m2 产次(次) 阴道分娩
≤35 36~55 ≥56 <24 ≥24 0 1 2 ≥3
训练集(80%) 总数 225 45(20.00) 94(41.78) 86(38.22) 156(69.33) 69(30.67) 21(9.33) 76(33.78) 83(36.89) 45(20.00) 155(68.89) 70(31.11)
S0 96 20(20.84) 38(39.58) 38(39.58) 67(69.79) 29(30.21) 8(8.33) 33(34.38) 37(38.54) 18(18.75) 65(67.71) 31(32.29)
S1 38 7(18.42) 16(42.11) 15(39.47) 26(68.42) 12(31.58) 4(10.53) 13(34.21) 13(34.21) 8(21.05) 26(68.42) 12(31.58)
S2 49 10(20.41) 20(40.82) 19(38.77) 34(69.39) 15(30.61) 5(10.20) 16(32.65) 18(32.74) 10(20.41) 33(67.35) 16(32.65)
S3 42 8(19.05) 20(47.62) 14(33.33) 29(69.05) 13(30.95) 4(9.52) 14(33.33) 15(35.72) 9(21.43) 31(73.81) 11(26.19)
测试集(20%) 总数 57 12(21.05) 30(52.63) 15(26.32) 44(77.19) 13(22.81) 2(3.51) 21(36.84) 23(40.35) 11(19.30) 39(68.42) 18(31.58)
S0 19 4(21.05) 9(47.37) 6(31.58) 15(78.95) 4(21.05) 1(5.26) 7(36.84) 7(36.84) 4(21.06) 13(68.42) 6(31.58)
S1 10 2(20.00) 5(50.00) 3(30.00) 7(70.00) 3(30.00) 1(10.00) 3(30.00) 4(40.00) 2(20.00) 7(70.00) 3(30.00)
S2 13 3(23.08) 7(53.86) 3(23.08) 10(76.92) 3(23.08) 0(0) 5(38.46) 6(46.15) 2(15.39) 8(61.54) 5(38.46)
S3 15 3(20.00) 9(60.00) 3(20.00) 12(80.00) 3(20.00) 0(0) 6(40.00) 6(40.00) 3(20.00) 11(73.33) 4(26.67)
χ2 3.043 1.362 2.172 0.005
P 0.218 0.243 0.537 0.946
图3 基于盆底超声的SUI分度诊断ResNet50模型五折交叉验证的ROC曲线图(a)及混淆矩阵(b~f)
图4 基于盆底超声的压力性尿失禁(SUI)分度诊断ResNet50模型在训练集(a)和测试集(b)的ROC曲线
表2 基于盆底超声的SUI分度诊断ResNet50模型在训练集和测试集中的评估性能(%)
[1]
Rogers RG,Pauls RN,Thakar R,et al. An international Urogynecological association (IUGA)/international continence society (ICS) joint report on the terminology for the assessment of sexual health of women with pelvic floor dysfunction[J]. Int Urogynecol J,2018,29(5): 647-666.
[2]
沈丹萍. 基于盆底超声参数的Nomogram模型对产后压力性尿失Shen DP. The predictive value of a Nomogram model based onpelvic floor ultrasound parameters for postpartum stressurinary incontinence[D].Jiangsu :Suzhou University,2020.
[3]
NICE Guidance - Urinary incontinence and pelvic organ prolapse in women: management: © NICE (2019) Urinary incontinence and pelvic organ prolapse in women: management [J]. BJU Int,2019,123(5): 777-803.
[4]
AIUM/IUGA practice parameter for the performance of Urogynecological ultrasound examinations: developed in collaboration with the ACR,the AUGS,the AUA,and the SRU[J].Int Urogynecol J,2019,30(9): 1389-1400.
[5]
Turkoglu A,Coskun A,Arinkan SA,et al. The role of transperineal ultrasound in the evaluation of stress urinary incontinence cases[J].Int Braz J Urol,2022,48(1): 70-77.
[6]
Guo L,Hao X,Chen L,et al. Early warning of hepatocellular carcinoma in cirrhotic patients by three-phase CT-based deep learning radiomics model: a retrospective,multicentre,cohort study[J].EClinicalMedicine,2024,74: 102718.
[7]
Choudhary T,Mishra V,Goswami A,et al. A transfer learning with structured filter pruning approach for improved breast cancer classification on point-of-care devices[J]. Comput Biol Med,2021,134: 104432.
[8]
Wang X,He D,Feng F,et al. Multi-label classification of pelvic organ prolapse using stress magnetic resonance imaging with deep learning[J]. Int Urogynecol J,2022,33(10): 2869-2877.
[9]
黄俊龙,刘柏隆,罗瑞翔,等. 联合盆底彩超数据和临床资料探讨压力性尿失禁的危险因素[J]. 中华腔镜泌尿外科杂志(电子版),2024,18(4): 323-330.Huang JL,Liu BL,Luo RX,et al. A combined analysis of clinical data and pelvic floor ultrasound data for exploring the risk factors of stress urinary incontinence[J]. Chin J Endourol Electron Ed,2024,18(4): 323-330.
[10]
Xiao T,Cao Y,Zhen C,et al. Nomogram analysis based on clinical and sonographic characteristics for the assessment of postpartum stress urinary incontinence[J]. J Ultrasound Med,2023,42(11): 2591-2601.
[11]
Bu L,Yang D,Nie F,et al. Correlation of the type and degree of cystocele with stress urinary incontinence by transperineal ultrasound[J]. J Med Ultrason,2020,47(1): 123-130.
[12]
Lukacz ES,Santiago-Lastra Y,Albo ME,et al. Urinary incontinence in women: a review[J]. JAMA,2017,318(16): 1592-1604.
[13]
Avery K,Donovan J,Peters TJ,et al. ICIQ: a brief and robust measure for evaluating the symptoms and impact of urinary incontinence[J]. Neurourol Urodyn,2004,23(4): 322-330.
[14]
Herrmann V,Di Sessa RG,Ricceto CL,et al. Correlation of the International Consultation on Incontinence Questionaire: urinary Incontinence/Short Form to Urodynamic diagnosis in women with urinary incontinence[J]. Rev Bras Ginecol Obstet,2013,35(1): 16-20.
[15]
胡可,鲁蓉. 基于多参数超声特征的中老年女性压力性尿失禁诊断模型研究[J]. 中华医学超声杂志(电子版),2024,21(5): 477-483.Hu K,Lu R. Multi-parameter ultrasound-based diagnostic model for stress urinary incontinence in middle-aged and elderly women[J].Chin J Med Ultrasound Electron Ed,2024,21(5): 477-483.
[16]
黄健. 中国泌尿外科和男科疾病诊断治疗指南(2019版)[M].北京: 科学出版社,2019,345-396.Huang J. Chinese Guidelines for diagnosis and treatment of Urology and Andrology diseases (2019 Edition) [M]. Beijing: Science Press,2019,345-396.
[17]
Brooks K,Varette K,Harvey M A,et al. A model identifying characteristics predictive of successful pelvic floor muscle training outcomes among women with stress urinary incontinence[J]. Int Urogynecol J,2021,32(3): 719-728.
[18]
Skorupska K,Grzybowska ME,Kubik-Komar A,et al. Identification of the Urogenital Distress Inventory-6 and the Incontinence Impact Questionnaire-7 cutoff scores in urinary incontinent women[J].Health Qual Life Outcomes,2021,19(1): 87. [PubMed]
[19]
Price DM,Noblett K. Comparison of the cough stress test and 24 h pad test in the assessment of stress urinary incontinence[J]. Int Urogynecol J,2012,23(4): 429-433.
[20]
Chen L,Luo D,Yu X,et al. Predicting stress urinary incontinence during pregnancy: combination of pelvic floor ultrasound parameters and clinical factors[J]. Acta Obstet Gynecol Scand,2018,97(8): 966-975.
[21]
Zhuo Z,Ye Z,Zhang J,et al. Correlation between three-dimensional transperineal ultrasound and pelvic floor electromyography in women with stress urinary incontinence[J]. Ginekol Pol,2023,94(1): 25-32.
[22]
Liu W,Qian L. Establishment and validation of a risk prediction model for postpartum stress urinary incontinence based on pelvic floor ultrasound and clinical data[J]. Int Urogynecol J,2022,33(12):3491-3497.
[23]
Song C,Wen W,Pan L,et al. Analysis of the anatomical and biomechanical characteristics of the pelvic floor in cystocele[J]. Acta Obstet Gynecol Scand,2023,102(12): 1661-1673.
[24]
李志毅,朱兰. 女性压力性尿失禁流行病学现状[J]. 实用妇产科杂志,2018,34(3): 161-162.Li ZY,Zhu L. Epidemiological status of female stress urinary incontinence[J]. J Pract Obstet Gynecol,2018,34(3): 161-162.
[25]
Yin W,Ma Q,Xie W,et al. Three-dimensional ultrasound assessment of risk factors for cystocele and Green classification in primipara[J].Front Med,2022,9: 979989.
[26]
Zhang M,Lin X,Zheng Z,et al. Artificial intelligence models derived from 2D transperineal ultrasound images in the clinical diagnosis of stress urinary incontinence[J]. Int Urogynecol J,2022,33(5): 1179-1185.
[27]
Naranjo-Ortiz C,Shek K L,Martin A J,et al. What is normal bladder neck anatomy?[J]. Int Urogynecology J,2016,27(6): 945-950.
[28]
Bader-El-Den M,Teitei E,Perry T. Biased random forest for dealing with the class imbalance problem[J]. IEEE Trans Neural Netw Learn Syst,2019,30(7): 2163-2172.
[29]
Jamard E,Blouet M,Thubert T,et al. Utility of 2D-ultrasound in pelvic floor muscle contraction and bladder neck mobility assessment in women with urinary incontinence[J]. J Gynecol Obstet Hum Reprod,2020,49(1): 101629.
[30]
Horosz E,Pomian A,Zwierzchowska A,et al. Epidemiological features of the bladder neck rest position and mobility[J]. J Clin Med,2020,9(8): 2413.
[1] 李洋, 蔡金玉, 党晓智, 常婉英, 巨艳, 高毅, 宋宏萍. 基于深度学习的乳腺超声应变弹性图像生成模型的应用研究[J]. 中华医学超声杂志(电子版), 2024, 21(06): 563-570.
[2] 胡可, 鲁蓉. 基于多参数超声特征的中老年女性压力性尿失禁诊断模型研究[J]. 中华医学超声杂志(电子版), 2024, 21(05): 477-483.
[3] 杨敬武, 周美君, 陈雨凡, 李素淑, 何燕妮, 崔楠, 刘红梅. 人工智能超声结合品管圈活动对低年资超声医师甲状腺结节风险评估能力的作用[J]. 中华医学超声杂志(电子版), 2024, 21(05): 522-526.
[4] 罗刚, 泮思林, 孙玲玉, 李志新, 陈涛涛, 乔思波, 庞善臣. 一种新型语义网络分析模型对室间隔完整型肺动脉闭锁和危重肺动脉瓣狭窄胎儿右心发育不良程度的评价作用[J]. 中华医学超声杂志(电子版), 2024, 21(04): 377-383.
[5] 明昊, 肖迎聪, 巨艳, 宋宏萍. 乳腺癌风险预测模型的研究现状[J]. 中华乳腺病杂志(电子版), 2024, 18(05): 287-291.
[6] 张嘉炜, 王瑞, 张克诚, 易磊, 周增丁. 烧烫伤创面深度智能检测模型P-YOLO的建立及测试效果[J]. 中华损伤与修复杂志(电子版), 2024, 19(05): 379-385.
[7] 熊鹰, 林敬莱, 白奇, 郭剑明, 王烁. 肾癌自动化病理诊断:AI离临床还有多远?[J]. 中华腔镜泌尿外科杂志(电子版), 2024, 18(06): 535-540.
[8] 李伟, 宋子健, 赖衍成, 周睿, 吴涵, 邓龙昕, 陈锐. 人工智能应用于前列腺癌患者预后预测的研究现状及展望[J]. 中华腔镜泌尿外科杂志(电子版), 2024, 18(06): 541-546.
[9] 屈勤芳, 束方莲. 盆腔器官脱垂患者盆底重建手术后压力性尿失禁发生的影响因素及列线图预测模型构建[J]. 中华腔镜泌尿外科杂志(电子版), 2024, 18(06): 606-612.
[10] 莫淇舟, 苏劲, 黄健, 李健维, 李思宁, 柳建军. 智能控压输尿管软镜碎石吸引取石术在直径10~25 mm上尿路结石中的应用[J]. 中华腔镜泌尿外科杂志(电子版), 2024, 18(05): 497-502.
[11] 苏博兴, 肖博, 李建兴. 2024年美国泌尿外科学会年会结石领域手术治疗相关热点研究及解读[J]. 中华腔镜泌尿外科杂志(电子版), 2024, 18(04): 303-308.
[12] 莫林键, 杨舒博, 农卫赟, 程继文. 人工智能虚拟数字医师在钬激光前列腺剜除日间手术患教管理中的应用[J]. 中华腔镜泌尿外科杂志(电子版), 2024, 18(04): 318-322.
[13] 黄俊龙, 刘柏隆, 罗瑞翔, 李晓阳, 李文双, 柳政, 陈嘉良, 周祥福. 联合盆底彩超数据和临床资料探讨压力性尿失禁的危险因素[J]. 中华腔镜泌尿外科杂志(电子版), 2024, 18(04): 323-330.
[14] 阮星星, 黄智渊, 刘芙香, 狄金明. 从临床医师诊治患者的思路出发撰写临床研究论文[J]. 中华腔镜泌尿外科杂志(电子版), 2024, 18(04): 397-401.
[15] 李义亮, 苏拉依曼·牙库甫, 麦麦提艾力·麦麦提明, 克力木·阿不都热依木. 机器人与腹腔镜食管裂孔疝修补术联合Nissen 胃底折叠术短期疗效分析[J]. 中华疝和腹壁外科杂志(电子版), 2024, 18(05): 512-517.
阅读次数
全文


摘要