The mechanical properties of red sandstone subjected to cyclic point loading were investigated. Tests were conducted using MTS servohydraulic landmark test system, under cyclic loadings with constant amplitudes and in...The mechanical properties of red sandstone subjected to cyclic point loading were investigated. Tests were conducted using MTS servohydraulic landmark test system, under cyclic loadings with constant amplitudes and increasing multi-level amplitudes. The frequencies range from 0.1 to 5 Hz and lower limit load ratios range from 0 to 0.60. Laboratory investigations were performed to find the effect of the frequency and the lower limit load ratio on the fatigue life and hysteresis properties of sandstone. The results show that the fatigue life of sandstone decreases first and then increases with the increase of frequency and lower limit load ratio. Under the same cycle number, the spacing between hysteresis loops increases with rising frequency and decreasing lower limit load ratio. The existence of “training” and “memory” effects in red sandstone under cyclic point loading was proved.展开更多
目的:采用深度学习方法,通过人在回路的方式进行迭代式标注-训练,建立垂体分割模型,实现垂体体积人工智能(AI)测量。方法:将1285例颅脑3D T 1WI图像按5~15岁、16~25岁、26~50岁、51~70岁年龄段分组,每个年龄组随机选择80例,分成4批次进...目的:采用深度学习方法,通过人在回路的方式进行迭代式标注-训练,建立垂体分割模型,实现垂体体积人工智能(AI)测量。方法:将1285例颅脑3D T 1WI图像按5~15岁、16~25岁、26~50岁、51~70岁年龄段分组,每个年龄组随机选择80例,分成4批次进行试验。初始每组选择3例图像进行人工预标注神经垂体和腺垂体,输入计算机进行学习,获取初始模型。应用模型对一批数据进行分割,获得分割后的神经垂体、腺垂体与垂体总体积数据,将分割结果进行人工校准,获得校准后相对应的体积数据作为金标准。用前一组校准后的分割图像进行计算机迭代式学习优化模型,再用优化后模型对新一组图像分割与校准,重复上述过程,最终将校准前后差异没有统计学意义的数据认定深度学习建模成功。数据采用配对t检验、Dice和Spearman相关性分析进行统计。结果:从第2批次开始,除5~15岁年龄段外,其它年龄段神经垂体体积在校准前后的差异没有统计学意义,腺垂体与垂体总体积的差异有统计学意义(P<0.05)。第4批次,各年龄段神经垂体、腺垂体与垂体总体积在校准前后的差异均无统计学意义(P=0.137~0.928),Dice值大于0.941,Spearman相关系数大于0.969。结论:通过迭代式训练进行深度学习建模可构建垂体分割模型,实现垂体体积AI自动测量。展开更多
基金Projects(51322403,51274254)supported by the National Natural Science Foundation of ChinaProject(2015CB060200)supported by the National Basic Research Program of China
文摘The mechanical properties of red sandstone subjected to cyclic point loading were investigated. Tests were conducted using MTS servohydraulic landmark test system, under cyclic loadings with constant amplitudes and increasing multi-level amplitudes. The frequencies range from 0.1 to 5 Hz and lower limit load ratios range from 0 to 0.60. Laboratory investigations were performed to find the effect of the frequency and the lower limit load ratio on the fatigue life and hysteresis properties of sandstone. The results show that the fatigue life of sandstone decreases first and then increases with the increase of frequency and lower limit load ratio. Under the same cycle number, the spacing between hysteresis loops increases with rising frequency and decreasing lower limit load ratio. The existence of “training” and “memory” effects in red sandstone under cyclic point loading was proved.
文摘目的:采用深度学习方法,通过人在回路的方式进行迭代式标注-训练,建立垂体分割模型,实现垂体体积人工智能(AI)测量。方法:将1285例颅脑3D T 1WI图像按5~15岁、16~25岁、26~50岁、51~70岁年龄段分组,每个年龄组随机选择80例,分成4批次进行试验。初始每组选择3例图像进行人工预标注神经垂体和腺垂体,输入计算机进行学习,获取初始模型。应用模型对一批数据进行分割,获得分割后的神经垂体、腺垂体与垂体总体积数据,将分割结果进行人工校准,获得校准后相对应的体积数据作为金标准。用前一组校准后的分割图像进行计算机迭代式学习优化模型,再用优化后模型对新一组图像分割与校准,重复上述过程,最终将校准前后差异没有统计学意义的数据认定深度学习建模成功。数据采用配对t检验、Dice和Spearman相关性分析进行统计。结果:从第2批次开始,除5~15岁年龄段外,其它年龄段神经垂体体积在校准前后的差异没有统计学意义,腺垂体与垂体总体积的差异有统计学意义(P<0.05)。第4批次,各年龄段神经垂体、腺垂体与垂体总体积在校准前后的差异均无统计学意义(P=0.137~0.928),Dice值大于0.941,Spearman相关系数大于0.969。结论:通过迭代式训练进行深度学习建模可构建垂体分割模型,实现垂体体积AI自动测量。
