正在加载图片...
人工智能在超声医学领域中的应用 析。目前更多研究集中于读片场景的应用,而针对患者 289-301 的操作扫查研究,有学者提出可用一个计算机控制的机9]ChiJ, Walia e, Babyn F,etal. Thyroid nodule classific 械臂控制探头的位置和角度进行不同标准切面的扫查, in ultrasound images by fine-tuning deep convolutional neural network [J]. J Digit Imaging, 2017, 30: 477-486 根据高矮胖瘦不一,机械臂借助一些辅助参数和标记部[10]kesr, Cyteval c, Gallix h,ctal. Appendiciti:eal 位精确调整扫查图像的位置和范围从而得到标准切面图 ation of sensitivity, specificity, and predictive values of US 像2。但是实际工作中,一个超声诊断临床决策的实 doppler US, and laboratory findings [J]. Radiology, 2004 现往往要结合多个切面、直接与间接征象以及临床信息 230:472-478 的分析,这样的能力或许需要到强A阶段才能实现。 [11 Kim KB, Park HJ, Song DH, et al. Developing an 一种新药从研发到推广应用至少需要经过3个阶 intelligent automatic appendix extraction method from ultra- IT and image processing [J 段,表明医疗创新实例往往需要面临更高的门槛,这主 Comput Math Methods Med, 2015, 2015: 389057. 要是由于医疗行业的预防原则所致。A亦是如此,其[12] Tajbakhsh N, Shin JY, Gurudu sr,etal. Convolutional 需要经历大量的测试,测试越多,暴露的信息与问题越 neural networks for medical image analysis: full training or 多,继而越能够预测后续可能产生的风险,同时考虑到 fine tuning? [J]. IEEE Trans Med Imaging, 2016, 35 法律法规、民众接受度以及潜在的医疗事故背后相应的 1299-1312. [13]孙夏,吴蔚,吴鹏,等.基于卷积神经网络的颈动脉斑 权责等问题,在大量测试后A能否切实应用于实际工 块超声图像特征识别[J].中国医疗器械信息,2016 作尚未可知,而如果AI能够代替医生完成大部分工作, 医生就可将工作重心放在临床决策的审核及科学研究方[14] Salomon, Winer n, Bernard JP,etal. A score-based 面,这一应用将解放医生们的生产力,使医生更加具有 创新动力,为专科发展带来裨益! mester ultrasound examination [J]. Prenat Diagn, 2008 822827 n H, Ni D, Qin J, et al. Standard plane localization in 参考文献 ultrasound via domain transferred deep neural network ]. IEEE J Biomed Health Inform, 2015 [1 Turing AM. Computing machinery and intelligence [J]. [16] Abdi AH, Luong C, Tsang T, et al. Automatic quality assess- Mind,1950,59:433-460. ment of echocardiograms using convolutional neural networks Peter N. Artificial intelligence, modern asibility on the apical four-chamber view []. IEEE Tran [M]. Englewood: Prentice Hall, 1995 Med Imaging,2017,36:;122-1230 3 Sekhar L, Wechsler L, Yonas H, et al. Value of transcranial [17 Knackstedt C, Bekkers SC, Schummers G, et al. Fully au- e dagnosIs after subarachnoid hemorrhage [J]. Neurosurgery, 1988 fraction and longitudinal strain: the fast-EFs multicenter study I. J Am Coll Cardiol, 2015, 66: 1456-1466 [4] Baumgartner RW, Mattle HP, Schroth G. Assessment of 2 [18] Furiasse N, Thomas JD. Automated algorithmic software in 50% and <50% intracranial stenoses by transcranial color-co. echocardiography: artificial intelligence? [J].J Am Coll Car ded duplex sonography [J]. Stroke, 1999, 30:87-92 diol,2015,66:1467-1469 [5] Swiercz M, Swiat M, Pawlak M, et al. Narrowing of the [19 Jeganathan J, Knio Z, Amador Y, et al. Artificial intellige middle cerebral artery: Ann Card Anaesth. 2017. 20 parison of transcranial color coded duplex sonography with con- 129134 ventional TCD [J]. Ultrasound Med Biol, 2010, 36: 17-28. [20 Kumar S, Nilsen WJ, Abernethy A, et al. Mobile health [6] Zhang L, Li QY, Duan YY, et al. Artificial neural network technology evaluation: the health evidence workshop [J] aided non-invasive grading evaluation of hepatic fibrosis by Am J Prev Med,2013,45:228-236. duplex ultrasonography[J]. BMC Med Inform Decis Mak,[2l]王弈,李传富.人工智能方法在医学图像处理中的研究新进 展[J].中国医学物理学杂志,2013,3:4138-4143 7] Zhu LC, Ye YL, Luo WH, et al. A model to discriminate [22] Priester AM, Natarajan S, Culjat MO. Robotic ultrasound alignant from benign thyroid nodules using artificial neural systems in medicine [J]. IEEE Trans Ultrason Ferroelectr etwork [J]. PLos One, 2013: e82211 Freq Control, 2013, 60: 507-523 [8 Acharya UR, Swapna G, Sree SV, et al. A review on ultra- und-based thyroid cancer tissue characterization and auto- 收稿日期:2017-09-06) mated classification [J]. Technol Cancer Res Treat, 2014 vo.9No.5457人工智能在超声医学领域中的应用 Vol􀆰 9 No􀆰 5 457 析ꎮ 目前更多研究集中于读片场景的应用ꎬ 而针对患者 的操作扫查研究ꎬ 有学者提出可用一个计算机控制的机 械臂控制探头的位置和角度进行不同标准切面的扫查ꎬ 根据高矮胖瘦不一ꎬ 机械臂借助一些辅助参数和标记部 位精确调整扫查图像的位置和范围从而得到标准切面图 像[22] ꎮ 但是实际工作中ꎬ 一个超声诊断临床决策的实 现往往要结合多个切面、 直接与间接征象以及临床信息 的分析ꎬ 这样的能力或许需要到强 AI 阶段才能实现ꎮ 一种新药从研发到推广应用至少需要经过 3 个阶 段ꎬ 表明医疗创新实例往往需要面临更高的门槛ꎬ 这主 要是由于医疗行业的预防原则所致ꎮ AI 亦是如此ꎬ 其 需要经历大量的测试ꎬ 测试越多ꎬ 暴露的信息与问题越 多ꎬ 继而越能够预测后续可能产生的风险ꎬ 同时考虑到 法律法规、 民众接受度以及潜在的医疗事故背后相应的 权责等问题ꎬ 在大量测试后 AI 能否切实应用于实际工 作尚未可知ꎬ 而如果 AI 能够代替医生完成大部分工作ꎬ 医生就可将工作重心放在临床决策的审核及科学研究方 面ꎬ 这一应用将解放医生们的生产力ꎬ 使医生更加具有 创新动力ꎬ 为专科发展带来裨益! 参 考 文 献 [1] Turing AM. Computing machinery and intelligence [ J ]. Mindꎬ 1950ꎬ 59: 433 ̄ 460. [2] Stuart JRꎬ Peter N. Artificial intelligence: modern approach [M]. Englewood: Prentice Hallꎬ 1995. [3] Sekhar Lꎬ Wechsler Lꎬ Yonas Hꎬ et al. Value of transcranial doppler examination in the diagnosis of cerebral vasospasm after subarachnoid hemorrhage [ J ]. Neurosurgeryꎬ 1988ꎬ 22: 813 ̄ 821. [4] Baumgartner RWꎬ Mattle HPꎬ Schroth G. Assessment of ≥ 50% and <50% intracranial stenoses by transcranial color ̄co ̄ ded duplex sonography [J]. Strokeꎬ 1999ꎬ 30: 87 ̄ 92. [5] Swiercz Mꎬ Swiat Mꎬ Pawlak Mꎬ et al. Narrowing of the middle cerebral artery: artificial intelligence methods and com ̄ parison of transcranial color coded duplex sonography with con ̄ ventional TCD [J]. Ultrasound Med Biolꎬ 2010ꎬ 36: 17 ̄ 28. [6] Zhang Lꎬ Li QYꎬ Duan YYꎬ et al. Artificial neural network aided non ̄invasive grading evaluation of hepatic fibrosis by duplex ultrasonography [ J]. BMC Med Inform Decis Makꎬ 2012ꎬ 12: 55. [7] Zhu LCꎬ Ye YLꎬ Luo WHꎬ et al. A model to discriminate malignant from benign thyroid nodules using artificial neural network [J]. PLoS Oneꎬ 2013: e82211. [8] Acharya URꎬ Swapna Gꎬ Sree SVꎬ et al. A review on ultra ̄ sound ̄based thyroid cancer tissue characterization and auto ̄ mated classification [J]. Technol Cancer Res Treatꎬ 2014: 289 ̄ 301. [9] Chi Jꎬ Walia Eꎬ Babyn Pꎬ et al. Thyroid nodule classification in ultrasound images by fine ̄tuning deep convolutional neural network [J]. J Digit Imagingꎬ 2017ꎬ 30: 477 ̄ 486. [10] Kessler Nꎬ Cyteval Cꎬ Gallix Bꎬ et al. Appendicitis: evalu ̄ ation of sensitivityꎬ specificityꎬ and predictive values of USꎬ doppler USꎬ and laboratory findings [ J]. Radiologyꎬ 2004ꎬ 230: 472 ̄ 478. [11] Kim KBꎬ Park HJꎬ Song DHꎬ et al. Developing an intelligent automatic appendix extraction method from ultra ̄ sonography based on fuzzy ART and image processing [ J]. Comput Math Methods Medꎬ 2015ꎬ 2015: 389057. [12] Tajbakhsh Nꎬ Shin JYꎬ Gurudu SRꎬ et al. Convolutional neural networks for medical image analysis: full training or fine tuning? [ J ]. IEEE Trans Med Imagingꎬ 2016ꎬ 35: 1299 ̄ 1312. [13] 孙夏ꎬ 吴蔚ꎬ 吴鹏ꎬ 等. 基于卷积神经网络的颈动脉斑 块超声图像特征识别 [ J]. 中国医疗器械信息ꎬ 2016ꎬ 9: 4 ̄ 8. [14] Salomon LJꎬ Winer Nꎬ Bernard JPꎬ et al. A score ̄based method for quality control of fetal images at routine second ̄tri ̄ mester ultrasound examination [ J ]. Prenat Diagnꎬ 2008ꎬ 28: 822 ̄ 827. [15] Chen Hꎬ Ni Dꎬ Qin Jꎬ et al. Standard plane localization in fetal ultrasound via domain transferred deep neural networks [J]. IEEE J Biomed Health Informꎬ 2015ꎬ 19: 1627 ̄ 1636. [16] Abdi AHꎬ Luong Cꎬ Tsang Tꎬ et al. Automatic quality assess ̄ ment of echocardiograms using convolutional neural networks: feasibility on the apical four ̄chamber view [J]. IEEE Trans Med Imagingꎬ 2017ꎬ 36: 1221 ̄ 1230. [17] Knackstedt Cꎬ Bekkers SCꎬ Schummers Gꎬ et al. Fully au ̄ tomated versus standard tracking of left ventricular ejection fraction and longitudinal strain: the fast ̄EFs multicenter study [J]. J Am Coll Cardiolꎬ 2015ꎬ 66: 1456 ̄ 1466. [18] Furiasse Nꎬ Thomas JD. Automated algorithmic software in echocardiography: artificial intelligence? [J]. J Am Coll Car ̄ diolꎬ 2015ꎬ 66: 1467 ̄ 1469. [19] Jeganathan Jꎬ Knio Zꎬ Amador Yꎬ et al. Artificial intelligence in mitral valve analysis [J]. Ann Card Anaesthꎬ 2017ꎬ 20: 129 ̄ 134. [20] Kumar Sꎬ Nilsen WJꎬ Abernethy Aꎬ et al. Mobile health technology evaluation: the health evidence workshop [ J ]. Am J Prev Medꎬ 2013ꎬ 45: 228 ̄ 236. [21] 王弈ꎬ 李传富. 人工智能方法在医学图像处理中的研究新进 展 [J]. 中国医学物理学杂志ꎬ 2013ꎬ 3: 4138 ̄ 4143. [22] Priester AMꎬ Natarajan Sꎬ Culjat MO. Robotic ultrasound systems in medicine [ J]. IEEE Trans Ultrason Ferroelectr Freq Controlꎬ 2013ꎬ 60: 507 ̄ 523. (收稿日期: 2017 ̄ 09 ̄ 06)
<<向上翻页
©2008-现在 cucdc.com 高等教育资讯网 版权所有