Circulation:深度神经网络自动判读超声心动图心室壁运动异常

2020-10-24 星云 MedSci原创

深度神经网络自动判读超声心动图检查可以支持临床报告和提高效率。虽然既往研究已经使用静止图像评估了心脏结构的空间关系,该研究旨在通过结合空间和时间信息来开发一个用于视频分析的深度神经网络,并进行检测,以

深度神经网络自动判读超声心动图检查可以支持临床报告和提高效率。虽然既往研究已经使用静止图像评估了心脏结构的空间关系,该研究旨在通过结合空间和时间信息来开发一个用于视频分析的深度神经网络,并进行检测,以自动化识别左心室区域壁运动异常。

研究人员收集了2017年7月-2018年4月期间在特定医院进行的10 638次经胸超声心动图检查。通过生理学家和心脏病专家共同判定心室壁运动异常。首先开发了一个三维卷积神经网络模型,用于视图选择,确保严格的图像质量控制。其次,U-net 模型分割图像,以注释每个左心室壁的位置。第三,三维卷积神经网络模型评估分段前后4个标准视图的超声心动图视频,计算每段心室壁运动异常置信度(0~1)。为了评估模型稳定性,研究人员还进行了 5 倍交叉验证和外部验证。

研究流程

在10 638次超声心动图检查中,该视图选择模型识别出6454 次(61%)检查图像质量符合分析标准。在开发过程中,对2740帧图像进行注释,以开发分段模型(Dice相似系数0.756)。在独立医院的1756次检查中进行了外部验证。在开发和外部验证队列中,分别观察到8.9%和4.9%的区域心室壁运动异常。在交叉验证和外部验证队列中,终极版模型识别区域壁运动异常的曲线下的区域分别为0.912和0.891。在外部验证队列中,灵敏度为81.8%,特异性为81.6%。

终极版模型的输出图示

在图像质量符合要求的超声心动图检查中,深度神经网络利用动态图像的时空信息自动识别心室壁运动异常是可行的。当然,还需要进一步研究,以优化模型性能并评估临床应用。

原始出处:

Mu-Shiang Huang, et al. Automated Recognition of Regional Wall Motion Abnormalities Through Deep Neural Network Interpretation of Transthoracic Echocardiography. Circulation. 2020;142:1510–1520

版权声明:
本网站所有内容来源注明为“梅斯医学”或“MedSci原创”的文字、图片和音视频资料,版权均属于梅斯医学所有。非经授权,任何媒体、网站或个人不得转载,授权转载时须注明来源为“梅斯医学”。其它来源的文章系转载文章,或“梅斯号”自媒体发布的文章,仅系出于传递更多信息之目的,本站仅负责审核内容合规,其内容不代表本站立场,本站不负责内容的准确性和版权。如果存在侵权、或不希望被转载的媒体或个人可与我们联系,我们将立即进行删除处理。
在此留言
评论区 (4)
#插入话题
  1. [GetPortalCommentsPageByObjectIdResponse(id=1651074, encodeId=ea3f16510e438, content=<a href='/topic/show?id=2aa35090181' target=_blank style='color:#2F92EE;'>#心动图#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=40, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=50901, encryptionId=2aa35090181, topicName=心动图)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=8a7624121693, createdName=zhu_jun9845, createdTime=Fri Jun 25 09:55:50 CST 2021, time=2021-06-25, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1813910, encodeId=1086181391086, content=<a href='/topic/show?id=101f9416466' target=_blank style='color:#2F92EE;'>#运动异常#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=0, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=94164, encryptionId=101f9416466, topicName=运动异常)], attachment=null, authenticateStatus=null, createdAvatar=, createdBy=7a9987, createdName=gous, createdTime=Sat Jan 02 10:55:50 CST 2021, time=2021-01-02, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=894750, encodeId=e365894e506e, content=在图像质量符合要求的超声心动图检查中,深度神经网络利用动态图像的时空信息自动识别心室壁运动异常是可行的。, beContent=null, objectType=article, channel=null, level=null, likeNumber=60, replyNumber=0, topicName=null, topicId=null, topicList=[], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=ae205385871, createdName=Lexi, createdTime=Tue Oct 27 10:46:13 CST 2020, time=2020-10-27, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1630363, encodeId=48c61630363cf, content=<a href='/topic/show?id=44e0e449217' target=_blank style='color:#2F92EE;'>#神经网络#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=32, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=74492, encryptionId=44e0e449217, topicName=神经网络)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=e6ff21497204, createdName=by2016, createdTime=Mon Oct 26 11:55:50 CST 2020, time=2020-10-26, status=1, ipAttribution=)]
  2. [GetPortalCommentsPageByObjectIdResponse(id=1651074, encodeId=ea3f16510e438, content=<a href='/topic/show?id=2aa35090181' target=_blank style='color:#2F92EE;'>#心动图#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=40, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=50901, encryptionId=2aa35090181, topicName=心动图)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=8a7624121693, createdName=zhu_jun9845, createdTime=Fri Jun 25 09:55:50 CST 2021, time=2021-06-25, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1813910, encodeId=1086181391086, content=<a href='/topic/show?id=101f9416466' target=_blank style='color:#2F92EE;'>#运动异常#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=0, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=94164, encryptionId=101f9416466, topicName=运动异常)], attachment=null, authenticateStatus=null, createdAvatar=, createdBy=7a9987, createdName=gous, createdTime=Sat Jan 02 10:55:50 CST 2021, time=2021-01-02, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=894750, encodeId=e365894e506e, content=在图像质量符合要求的超声心动图检查中,深度神经网络利用动态图像的时空信息自动识别心室壁运动异常是可行的。, beContent=null, objectType=article, channel=null, level=null, likeNumber=60, replyNumber=0, topicName=null, topicId=null, topicList=[], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=ae205385871, createdName=Lexi, createdTime=Tue Oct 27 10:46:13 CST 2020, time=2020-10-27, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1630363, encodeId=48c61630363cf, content=<a href='/topic/show?id=44e0e449217' target=_blank style='color:#2F92EE;'>#神经网络#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=32, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=74492, encryptionId=44e0e449217, topicName=神经网络)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=e6ff21497204, createdName=by2016, createdTime=Mon Oct 26 11:55:50 CST 2020, time=2020-10-26, status=1, ipAttribution=)]
    2021-01-02 gous
  3. [GetPortalCommentsPageByObjectIdResponse(id=1651074, encodeId=ea3f16510e438, content=<a href='/topic/show?id=2aa35090181' target=_blank style='color:#2F92EE;'>#心动图#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=40, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=50901, encryptionId=2aa35090181, topicName=心动图)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=8a7624121693, createdName=zhu_jun9845, createdTime=Fri Jun 25 09:55:50 CST 2021, time=2021-06-25, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1813910, encodeId=1086181391086, content=<a href='/topic/show?id=101f9416466' target=_blank style='color:#2F92EE;'>#运动异常#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=0, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=94164, encryptionId=101f9416466, topicName=运动异常)], attachment=null, authenticateStatus=null, createdAvatar=, createdBy=7a9987, createdName=gous, createdTime=Sat Jan 02 10:55:50 CST 2021, time=2021-01-02, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=894750, encodeId=e365894e506e, content=在图像质量符合要求的超声心动图检查中,深度神经网络利用动态图像的时空信息自动识别心室壁运动异常是可行的。, beContent=null, objectType=article, channel=null, level=null, likeNumber=60, replyNumber=0, topicName=null, topicId=null, topicList=[], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=ae205385871, createdName=Lexi, createdTime=Tue Oct 27 10:46:13 CST 2020, time=2020-10-27, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1630363, encodeId=48c61630363cf, content=<a href='/topic/show?id=44e0e449217' target=_blank style='color:#2F92EE;'>#神经网络#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=32, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=74492, encryptionId=44e0e449217, topicName=神经网络)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=e6ff21497204, createdName=by2016, createdTime=Mon Oct 26 11:55:50 CST 2020, time=2020-10-26, status=1, ipAttribution=)]
    2020-10-27 Lexi

    在图像质量符合要求的超声心动图检查中,深度神经网络利用动态图像的时空信息自动识别心室壁运动异常是可行的。

    0

  4. [GetPortalCommentsPageByObjectIdResponse(id=1651074, encodeId=ea3f16510e438, content=<a href='/topic/show?id=2aa35090181' target=_blank style='color:#2F92EE;'>#心动图#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=40, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=50901, encryptionId=2aa35090181, topicName=心动图)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=8a7624121693, createdName=zhu_jun9845, createdTime=Fri Jun 25 09:55:50 CST 2021, time=2021-06-25, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1813910, encodeId=1086181391086, content=<a href='/topic/show?id=101f9416466' target=_blank style='color:#2F92EE;'>#运动异常#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=0, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=94164, encryptionId=101f9416466, topicName=运动异常)], attachment=null, authenticateStatus=null, createdAvatar=, createdBy=7a9987, createdName=gous, createdTime=Sat Jan 02 10:55:50 CST 2021, time=2021-01-02, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=894750, encodeId=e365894e506e, content=在图像质量符合要求的超声心动图检查中,深度神经网络利用动态图像的时空信息自动识别心室壁运动异常是可行的。, beContent=null, objectType=article, channel=null, level=null, likeNumber=60, replyNumber=0, topicName=null, topicId=null, topicList=[], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=ae205385871, createdName=Lexi, createdTime=Tue Oct 27 10:46:13 CST 2020, time=2020-10-27, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1630363, encodeId=48c61630363cf, content=<a href='/topic/show?id=44e0e449217' target=_blank style='color:#2F92EE;'>#神经网络#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=32, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=74492, encryptionId=44e0e449217, topicName=神经网络)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=e6ff21497204, createdName=by2016, createdTime=Mon Oct 26 11:55:50 CST 2020, time=2020-10-26, status=1, ipAttribution=)]