目的通过研究和搭建人工智能深度学习网络,实现多模态心脏磁共振(cardiac magnetic resonance,CMR)图像分割,并提升Dice系数。材料与方法回顾性分析来自2019年多序列CMR分割挑战赛的公开数据集,它包含了45例患者平衡稳态自由进动(balanc...目的通过研究和搭建人工智能深度学习网络,实现多模态心脏磁共振(cardiac magnetic resonance,CMR)图像分割,并提升Dice系数。材料与方法回顾性分析来自2019年多序列CMR分割挑战赛的公开数据集,它包含了45例患者平衡稳态自由进动(balanced-steady state free precession,bSSFP)模态,晚期钆增强(late gadolinium enhancement,LGE)模态与T2WI模态的CMR图像数据。本文构建了一种新的双流U型网络框架,实现bSSFP与LGE两种模态以及bSSFP与T2WI两种模态的CMR图像分割。在编码阶段,未配准各模态图像被交替地送入各自分支进行特征学习,所获取的特征图接着都流入共享层,实现多模态信息的交互补充,最终共享特征分开流出到各自分支进行解码输出。通过在45例患者的CMR图像数据集上进行五折交叉验证实验,分别对bSSFP与LGE模态、bSSFP与T2WI模态进行了分割,以Dice系数对提出的模型进行性能评估,Wilcoxon符号秩检验被用来检验模型差异性。结果在bSSFP与LGE模态的分割实验中,本文方法在bSSFP模态的平均Dice系数相较于传统UNet模型和最新的Swin-Unet模型都有显著提升(P<0.001);在LGE模态的平均Dice系数较传统UNet模型(P<0.001)、Swin-Unet模型(P=0.001)、双流UNet(P=0.021)均有显著提升。在bSSFP与T2WI模态的分割实验中,本文方法在bSSFP模态的平均Dice系数较UNet模型、Swin-Unet模型与双流UNet均有显著提升(P<0.001);在T2WI模态的平均Dice系数较UNet模型有显著提升(P<0.001),较Swin-Unet模型有提升(P=0.025)。结论本研究提出的双流U型网络框架为CMR图像多模态分割提供有效方法,且该网络提高了CMR图像bSSFP模态与LGE模态及bSSFP模态与T2WI模态的Dice系数,很好地解决了多模态CMR图像个体解剖学差异大和图像间存在灰度不一致问题,提升了模型的泛化能力。展开更多
In low permeability porous media which permeability anisotropy is ubiquitous,the percolation of fluid no longer follows linear Darcy’s Law.Oil-water two phase flow equation of low permeability reservoir with permeabi...In low permeability porous media which permeability anisotropy is ubiquitous,the percolation of fluid no longer follows linear Darcy’s Law.Oil-water two phase flow equation of low permeability reservoir with permeability anisotropy is established based on generalized Darcy’s law and starting pressure gradient,corresponding finite element program is developed and simulated based on the Finite Element Program Generator system (FEPG).The results show that energy-gathering exists in the flow event of flowing area front in low permeability reservoir.In the process of energy-gathering,the flow velocity changed little but increased rapidly as soon as the pressure gradient exceeded the starting pressure gradient of the reservoir,then gradually stabilized.The greater the starting pressure gradient is,the greater the near wellbore pressure drop is,the smaller the area influenced by the reservoir pressure changes caused by water injection and oil recovery.The greater the starting pressure gradient is,the lower the water saturation in same point of the reservoir is,the smaller the water flood swept area is.There will be more difficulties in water injection to the same extent.展开更多
文摘目的通过研究和搭建人工智能深度学习网络,实现多模态心脏磁共振(cardiac magnetic resonance,CMR)图像分割,并提升Dice系数。材料与方法回顾性分析来自2019年多序列CMR分割挑战赛的公开数据集,它包含了45例患者平衡稳态自由进动(balanced-steady state free precession,bSSFP)模态,晚期钆增强(late gadolinium enhancement,LGE)模态与T2WI模态的CMR图像数据。本文构建了一种新的双流U型网络框架,实现bSSFP与LGE两种模态以及bSSFP与T2WI两种模态的CMR图像分割。在编码阶段,未配准各模态图像被交替地送入各自分支进行特征学习,所获取的特征图接着都流入共享层,实现多模态信息的交互补充,最终共享特征分开流出到各自分支进行解码输出。通过在45例患者的CMR图像数据集上进行五折交叉验证实验,分别对bSSFP与LGE模态、bSSFP与T2WI模态进行了分割,以Dice系数对提出的模型进行性能评估,Wilcoxon符号秩检验被用来检验模型差异性。结果在bSSFP与LGE模态的分割实验中,本文方法在bSSFP模态的平均Dice系数相较于传统UNet模型和最新的Swin-Unet模型都有显著提升(P<0.001);在LGE模态的平均Dice系数较传统UNet模型(P<0.001)、Swin-Unet模型(P=0.001)、双流UNet(P=0.021)均有显著提升。在bSSFP与T2WI模态的分割实验中,本文方法在bSSFP模态的平均Dice系数较UNet模型、Swin-Unet模型与双流UNet均有显著提升(P<0.001);在T2WI模态的平均Dice系数较UNet模型有显著提升(P<0.001),较Swin-Unet模型有提升(P=0.025)。结论本研究提出的双流U型网络框架为CMR图像多模态分割提供有效方法,且该网络提高了CMR图像bSSFP模态与LGE模态及bSSFP模态与T2WI模态的Dice系数,很好地解决了多模态CMR图像个体解剖学差异大和图像间存在灰度不一致问题,提升了模型的泛化能力。
文摘In low permeability porous media which permeability anisotropy is ubiquitous,the percolation of fluid no longer follows linear Darcy’s Law.Oil-water two phase flow equation of low permeability reservoir with permeability anisotropy is established based on generalized Darcy’s law and starting pressure gradient,corresponding finite element program is developed and simulated based on the Finite Element Program Generator system (FEPG).The results show that energy-gathering exists in the flow event of flowing area front in low permeability reservoir.In the process of energy-gathering,the flow velocity changed little but increased rapidly as soon as the pressure gradient exceeded the starting pressure gradient of the reservoir,then gradually stabilized.The greater the starting pressure gradient is,the greater the near wellbore pressure drop is,the smaller the area influenced by the reservoir pressure changes caused by water injection and oil recovery.The greater the starting pressure gradient is,the lower the water saturation in same point of the reservoir is,the smaller the water flood swept area is.There will be more difficulties in water injection to the same extent.
