欢迎来到《四川大学学报(医学版)》

基于未增强T1 mapping的深度学习模型早期评价肥厚型心肌病心肌纤维化的初步研究

师轲 李颖 张天静 李真林 黎海霞 彭婉琳 夏春潮

师轲, 李颖, 张天静, 等. 基于未增强T1 mapping的深度学习模型早期评价肥厚型心肌病心肌纤维化的初步研究[J]. 四川大学学报(医学版), 2021, 52(5): 819-824. doi: 10.12182/20210960506
引用本文: 师轲, 李颖, 张天静, 等. 基于未增强T1 mapping的深度学习模型早期评价肥厚型心肌病心肌纤维化的初步研究[J]. 四川大学学报(医学版), 2021, 52(5): 819-824. doi: 10.12182/20210960506
SHI Ke, LI Ying, ZHANG Tian-jing, et al. Early Assessment of Myocardial Fibrosis of Hypertrophic Cardiomyopathy with Native-T1-Mapping-Based Deep Learning: A Preliminary Study[J]. JOURNAL OF SICHUAN UNIVERSITY (MEDICAL SCIENCE EDITION), 2021, 52(5): 819-824. doi: 10.12182/20210960506
Citation: SHI Ke, LI Ying, ZHANG Tian-jing, et al. Early Assessment of Myocardial Fibrosis of Hypertrophic Cardiomyopathy with Native-T1-Mapping-Based Deep Learning: A Preliminary Study[J]. JOURNAL OF SICHUAN UNIVERSITY (MEDICAL SCIENCE EDITION), 2021, 52(5): 819-824. doi: 10.12182/20210960506

基于未增强T1 mapping的深度学习模型早期评价肥厚型心肌病心肌纤维化的初步研究

doi: 10.12182/20210960506
基金项目: 四川大学华西医院学科卓越发展1·3·5工程项目(No. ZYGD18019)资助
详细信息
    通讯作者:

    E-mail:xiachunchao@wchscu.cn

Early Assessment of Myocardial Fibrosis of Hypertrophic Cardiomyopathy with Native-T1-Mapping-Based Deep Learning: A Preliminary Study

More Information
  • 摘要:   目的   通过深度学习(deep learning, DL)模型分析心脏磁共振检查中不伴有心肌延迟强化(late gadolinium enhancement, LGE)的肥厚型心肌病(hypertrophic cardiomyopathy, HCM)未增强T1 mapping图像,探讨其早期识别心肌间质性纤维化的能力。   方法   纳入接受心脏磁共振检查的60例HCM患者及44例正常对照者,以有无LGE判断并标记对应的未增强T1 mapping图像,将其与正常对照者的未增强T1 mapping图像经过预处理后作为矩阵形式输入SE-ResNext-50模型进行训练、验证及测试。   结果   总共241幅未增强T1 mapping图像输入SE-ResNext-50模型,该模型识别测试集中LGE(−)未增强T1 mapping图像的特异性0.87,敏感性0.79,曲线下面积0.83(P<0.05)。   结论   基于SE-ResNext-50的DL模型可较准确地识别LGE(−)未增强T1 mapping图像,可在不依赖对比剂的情况下早期发现HCM心肌纤维化。
  • 图  1  未增强T1 mapping图像分割示意图

    Figure  1.  Segmentation of native T1 maps

    图  2  SE-ResNext-50模型工作流程图

    Figure  2.  The workflow of SE-ResNext-50

    X: Matrix; H: Height; W: Width; C: Number of channels; FC: Fully convolutional layers; ReLU: Rectified linear unit.

    图  3  SE-ResNext-50网络训练的loss曲线(A)和acc曲线(B)

    Figure  3.  The loss curve (A) and acc curve (B) of SE-ResNext-50

    图  4  SE-ResNext-50在训练集、内部验证集和测试集中的诊断效能

    Figure  4.  Diagnostic performance of SE-ResNext-50 in the training, validation and testing data sets

    表  1  SE-ResNext-50网络结构参数

    Table  1.   The structural parameters of SE-ResNext-50 network

    Output sizeSE-ResNext-50
    112×112 conv, 7×7,64, stride 2
    56×56 max pool, 3×3, stride 2
    $ \left[\begin{array}{c}\text{c}\text{o}\text{n}\text{v}\text{,}\text{1} \times \text{1,128}\\ \text{c}\text{o}\text{n}\text{v}\text{,}\text{3} \times \text{3,128}\\ \text{c}\text{o}\text{n}\text{v}\text{,}\text{1} \times \text{1,256}\\ {f}{c}\text{,}\text{[}\text{16,256}\text{]}\end{array}\;\;\;{C}\text{= 32}\right] $×3
    28×28 $ \left[\begin{array}{c}\text{c}\text{o}\text{n}\text{v}\text{,}\text{1} \times \text{1,256}\\ \text{c}\text{o}\text{n}\text{v}\text{,}\text{3} \times \text{3,256}\\ \text{c}\text{o}\text{n}\text{v}\text{,}\text{1} \times \text{1,512}\\ {f}{c}\text{,}\text{[}\text{32,512}\text{]}\end{array}\;\;\;{C}\text{= 32}\right] $×4
    14×14 $ \left[\begin{array}{c}\text{c}\text{o}\text{n}\text{v}\text{,}\text{1} \times \text{1,512}\\ \text{c}\text{o}\text{n}\text{v}\text{,}\text{3} \times \text{3,512}\\ \text{c}\text{o}\text{n}\text{v}\text{,}\text{1} \times \text{1,1\;024}\\ {f}{c}\text{,}\text{[}\text{64,1\;024}\text{]}\end{array}\;\;\;{C}\text{= 32}\right] $×6
    7×7 $ \left[\begin{array}{c}\text{c}\text{o}\text{n}\text{v}\text{,}\text{1} \times \text{1,1\;024}\\ \text{c}\text{o}\text{n}\text{v}\text{,}\text{3} \times \text{3,1\;024}\\ \text{c}\text{o}\text{n}\text{v}\text{,}\text{1} \times \text{1,2\;048}\\ {f}{c}\text{,}\text{[}\text{128,2\;048}\text{]}\end{array}\;\;\;{C}\text{= 32}\right] $×3
    1×1 Global average pool,100-d, $ {fc}{,} $ softmax
    下载: 导出CSV

    表  2  基本资料

    Table  2.   Baseline data of the subjects of the study

    VariablesHCM (n=60)Normal controls (n=44)
    Age/yr. 52.9±13.3 52.1±8.1
    Male/case (%) 36 (60.0) 22 (50.0)
    BMI/(kg/m2) 24.6±2.9 22.7±2.7
    Hypertension/case (%) 27 (45.0) 0
    Diabetes mellitus/case (%) 4 (6.7) 0
    Smoking/case (%) 20 (33.3) 7 (15.9)
    Drinking/case (%) 14 (23.3) 6 (28.6)
    NYHA functional class/case (%)
     Ⅰ 31 (51.7) 44 (100)
     Ⅱ 22 (36.7) 0
     Ⅲ 7 (11.7) 0
    LV end-diastolic volume/mL 151.7±33.9* 133.2±24.1
    LV end-systolic volume/mL 54.9±16.8 53.2±14.8
    LV ejection fraction/% 64.0±7.1 60.6±5.9
    LV mass/g, median (P25, P75) 120.9 (96.2,162.7)* 77.6 (66.1, 93.5)
    LV maximum WT/mm 17.4±3.8* 7.9±1.4
    Global T1/ms 1282±47* 1254±39
    Basal T1/ms 1283±51* 1242±32
    Mid T1/ms 1280±52* 1249±42
    Apical T1/ms 1281±51 1281±71
    Septal T1/ms 1302±50* 1270±40
     HCM: Hypertrophic cardiomyopathy patients; BMI: Body mass index; NYHA: New York Heart Association; LV: Left ventricle; WT: Wall thickness. *P<0.05, vs. normal control.
    下载: 导出CSV
  • [1] SEMSARIAN C, INGLES J, MARON M S, et al. New perspectives on the prevalence of hypertrophic cardiomyopathy. J Am Coll Cardiol,2015,65(12): 1249–1254. doi: 10.1016/j.jacc.2015.01.019
    [2] HO C Y, DAY S M, ASHLEY E A, et al. Genotype and lifetime burden of disease in hypertrophic cardiomyopathy: Insights from the sarcomeric human cardiomyopathy registry (SHaRe). Circulation,2018,138(14): 1387–1398. doi: 10.1161/CIRCULATIONAHA.117.033200
    [3] HASHIMURA H, KIMURA F, ISHIBASHI-UEDA H, et al. Radiologic-pathologic correlation of primary and secondary cardiomyopathies: MR imaging and histopathologic findings in hearts from autopsy and transplantation. Radiographics,2017,37(3): 719–736. doi: 10.1148/rg.2017160082
    [4] WENG Z, YAO J, CHAN R H, et al. Prognostic value of LGE-CMR in HCM: A meta-analysis. JACC Cardiovasc Imaging,2016,9(12): 1392–1402. doi: 10.1016/j.jcmg.2016.02.031
    [5] GALATI G, LEONE O, PASQUALE F, et al. Histological and histometric characterization of myocardial fibrosis in end-stage hypertrophic cardiomyopathy: A clinical-pathological study of 30 explanted hearts. Circ Heart Fail,2016,9(9): e003090[2021-08-11]. https://doi.org/10.1161/CIRCHEARTFAILURE.116.003090. doi: 10.1161/CIRCHEARTFAILURE.116.003090
    [6] TODIERE G, AQUARO G D, PIAGGI P, et al. Progression of myocardial fibrosis assessed with cardiac magnetic resonance in hypertrophic cardiomyopathy. J Am Coll Cardiol,2012,60(10): 922–929. doi: 10.1016/j.jacc.2012.03.076
    [7] AXELSSON RAJA A, FARHAD H, VALENTE A M, et al. Prevalence and progression of late gadolinium enhancement in children and adolescents with hypertrophic cardiomyopathy. Circulation,2018,138(8): 782–792. doi: 10.1161/CIRCULATIONAHA.117.032966
    [8] NEUBAUER S, KOLM P, HO C Y, et al. Distinct subgroups in hypertrophic cardiomyopathy in the NHLBI HCM registry. J Am Coll Cardiol,2019,12; 74(19): 2333–2345. doi: 10.1016/j.jacc.2019.08.1057
    [9] ROBINSON A A, CHOW K, SALERNO M. Myocardial T1 and ECV measurement: Underlying concepts and technical considerations. JACC Cardiovasc Imaging,2019,12(11 Pt 2): 2332–2344. doi: 10.1016/j.jcmg.2019.06.031
    [10] HINOJAR R, VARMA N, CHILD N, et al. T1 mapping in discrimination of hypertrophic phenotypes: Hypertensive heart disease and hypertrophic cardiomyopathy: Findings from the international T1 multicenter cardiovascular magnetic resonance study. Circ Cardiovasc Imaging,2015,8(12): e003285[2021-08-11]. https://doi.org/10.1161/CIRCIMAGING.115.003285. doi: 10.1161/CIRCIMAGING.115.003285
    [11] SWOBODA P P, MCDIARMID A K, ERHAYIEM B, et al. Effect of cellular and extracellular pathology assessed by T1 mapping on regional contractile function in hypertrophic cardiomyopathy. J Cardiovasc Magn Reson,2017,19(1): 16. doi: 10.1186/s12968-017-0334-x
    [12] KARUR G R, ROBISON S, IWANOCHKO R M, et al. Use of myocardial T1 mapping at 3.0 T to differentiate Anderson-Fabry disease from hypertrophic cardiomyopathy. Radiology,2018,288(2): 398–406. doi: 10.1148/radiol.2018172613
    [13] 李湘, 师轲, 杨志刚. 心脏磁共振多模态成像在肥厚性心肌病中的临床应用. 西部医学,2020,32(3): 460–464. doi: 10.3969/j.issn.1672-3511.2020.03.035
    [14] GOTTBRECHT M, KRAMER C M, SALERNO M. Native T1 and extracellular volume measurements by cardiac MRI in healthy adults: A meta-analysis. Radiology,2019,290(2): 317–326. doi: 10.1148/radiol.2018180226
    [15] LEINER T, RUECKERT D, SUINESIAPUTRA A, et al. Machine learning in cardiovascular magnetic resonance: Basic concepts and applications. J Cardiovasc Magn Reson,2019,21(1): 61. doi: 10.1186/s12968-019-0575-y
    [16] CARDIM N, GALDERISI M, EDVARDSEN T, et al. Role of multimodality cardiac imaging in the management of patients with hypertrophic cardiomyopathy: An expert consensus of the European Association of Cardiovascular Imaging Endorsed by the Saudi Heart Association. Eur Heart J Cardiovasc Imaging,2015,16(3): 280. doi: 10.1093/ehjci/jeu291
    [17] OLIVOTTO I, CECCHI F, POGGESI C, et al. Patterns of disease progression in hypertrophic cardiomyopathy: An individualized approach to clinical staging. Circ Heart Fail,2012,5(4): 535–546. doi: 10.1161/circheartfailure.112.967026
    [18] RADENKOVIC D, WEINGÄRTNER S, RICKETTS L, et al. T1 mapping in cardiac MRI. Heart Fail Rev,2017,22(4): 415–430. doi: 10.1007/s10741-017-9627-2
    [19] XU J, ZHUANG B, SIRAJUDDIN A, et al. MRI T1 mapping in hypertrophic cardiomyopathy: Evaluation in patients without late gadolinium enhancement and hemodynamic obstruction. Radiology,2020,294(2): 275–286. doi: 10.1148/radiol.2019190651
    [20] NEISIUS U, EL-REWAIDY H, NAKAMORI S, et al. Radiomic analysis of myocardial native T1 imaging discriminates between hypertensive heart disease and hypertrophic cardiomyopathy. JACC Cardiovasc Imaging,2019,12(10): 1946–1954. doi: 10.1016/j.jcmg.2018.11.024
    [21] NEISIUS U, El-REWAIDY H, KUCUKSEYMEN S, et al. Texture signatures of native myocardial T1 as novel imaging markers for identification of hypertrophic cardiomyopathy patients without scar. J Magn Reson Imaging,2020,52(3): 906–919. doi: 10.1002/jmri.27048
    [22] 陈梓娴, 林晨, 倪金荣, 等. 2018年北美放射学年会心脏MRI人工智能研究进展. 中国医学影像技术,2019,35(7): 1111–1114. doi: 10.13929/j.1003-3289.201901062
  • 加载中
图(4) / 表(2)
计量
  • 文章访问数:  63
  • HTML全文浏览量:  25
  • PDF下载量:  9
  • 被引次数: 0
出版历程
  • 收稿日期:  2020-10-13
  • 修回日期:  2021-08-12
  • 网络出版日期:  2021-12-06
  • 刊出日期:  2021-09-20

目录

    /

    返回文章
    返回