邢益民, 张天飞, 邱清勇, 戴慧勇. 基于深度学习的心脏超声血流成像对肥厚型心肌病合并心律失常的预测价值[J]. 心脏杂志, 2024, 36(2): 176-181. DOI: 10.12125/j.chj.202304109
    引用本文: 邢益民, 张天飞, 邱清勇, 戴慧勇. 基于深度学习的心脏超声血流成像对肥厚型心肌病合并心律失常的预测价值[J]. 心脏杂志, 2024, 36(2): 176-181. DOI: 10.12125/j.chj.202304109
    XING Yi-min, ZHANG Tian-fei, QIU Qing-yong, DAI Hui-yong. Predictive value of deep learning-based cardiac ultrasound flow imaging for hypertrophic cardiomyopathy complicated with arrhythmias[J]. Chinese Heart Journal, 2024, 36(2): 176-181. DOI: 10.12125/j.chj.202304109
    Citation: XING Yi-min, ZHANG Tian-fei, QIU Qing-yong, DAI Hui-yong. Predictive value of deep learning-based cardiac ultrasound flow imaging for hypertrophic cardiomyopathy complicated with arrhythmias[J]. Chinese Heart Journal, 2024, 36(2): 176-181. DOI: 10.12125/j.chj.202304109

    基于深度学习的心脏超声血流成像对肥厚型心肌病合并心律失常的预测价值

    Predictive value of deep learning-based cardiac ultrasound flow imaging for hypertrophic cardiomyopathy complicated with arrhythmias

    • 摘要:
      目的  探讨基于深度学习的心脏超声血流成像对肥厚型心肌病(HCM)合并心律失常的预测价值。
      方法  对2020年3月~2023年3月收治的90例HCM患者的临床资料进行回顾性分析,按4:1:1的比例分为训练组(n=60)、验证组(n=15)和试验组(n=15)。训练组患者根据是否合并心律失常分为并发组(n=20)和非并发组(n=40)。收集患者的一般数据,建立心脏超声血流成像的深度学习模型,提取图像数据:左心室射血分数(LVEF)、左心房容积指数(LAVI)、E/e’、涡流面积变化率、循环强度变化率、平均血流速度和平均能量损耗(EL)值。
      结果 3组患者一般资料不存在统计学差异;训练组中并发组与非并发组患者的一般资料也不存在显著性差异,但是并发组的LAVI(P<0.01)、E/e’(P<0.01)、涡流面积变化率(P<0.01)、循环强度变化率(P<0.01)、平均血流速度(P<0.05)、平均EL(P<0.01)显著高于非并发组,LVEF则显著低于非并发组(P<0.01);多因素Logistic回归分显示:涡流面积变化率、循环强度变化率、平均血流速度、平均EL与LAVI、E/e’均为HCM伴发心律失常的危险因素,LVEF是HCM伴发心律失常的保护因素;训练组模型ROC曲线AUC值为0.985,验证组ROC曲线AUC值为0.989,试验组ROC曲线AUC值为0.980。
      结论 基于深度学习的心脏超声血流成像能更准确地识别心脏超声图像,对HCM并发心律失常有较高的预测价值,涡流面积变化率、循环强度变化率、平均血流速度、平均EL、LAVI和E/e’均是HCM并发心律失常的危险因素。LVEF是HCM伴发心律失常的保护因素。

       

      Abstract:
      AIM To explore the predictive value of deep learning-based cardiac ultrasound flow imaging for hypertrophic cardiomyopathy (HCM) complicated with arrhythmias.
      METHODS The clinical data of 90 patients with hypertrophic cardiomyopathy from March 2020 to March 2023 were retrospectively collected. Based on the ratio of 4:1:1, they were divided into training group (60 cases), validation group (15 cases) and test group (15 cases) , and based on HCM complicated with arrhythmia or without arrhythmia, the cases in training group were divided into concurrent group and non-concurrent group. A deep learning model for cardiac ultrasound flow imaging was established, and image data, LVEF, LAVI, E/e’, vortex area change rate, circulation intensity change rate, mean blood flow velocity and mean EL value were extracted.
      RESULTS The differences in general data between the three groups were not statistically significant and the differences in general data between patients in concurrent group and non-concurrent group in the training group were also not statistically significant But the LAVI, E/e’, vortex area change rate, circulation intensity change rate, mean blood flow velocity and average EL of concurrent group were significantly higher than those of non-concurrent group, while LVEF was significantly lower than that of non-concurrent group (P<0.05 or P<0.01). Multivariate logistic regression scores showed that vortex area change rate, circulation intensity change rate, mean blood flow velocity, mean EL, LAVI, and E/e’ were all risk factors for arrhythmia in HCM, while LVEF was a protective factor for arrhythmia in HCM. The AUC values of ROC curve in training group, validation group and test group were 0.985, 0.989 and 0.980, respectively.
      CONCLUSION Deep learning-based cardiac ultrasound flow imaging can more accurately identify cardiac ultrasound images and has a high predictive value for hypertrophic cardiomyopathy complicated with arrhythmias. Vortex area change rate, circulation intensity change rate, mean flow velocity, mean EL, LAVI, and E/e’ are all risk factors for hypertrophic cardiomyopathy complicated with arrhythmias. LVEF is a protective factor for HCM accompanied by arrhythmia.

       

    /

    返回文章
    返回