目的探究深度学习技术在中国汉族人群CT三维重建图像自动性别识别中的可靠性和准确率。方法收集20~85岁汉族人群骨盆CT影像学数据700例(男性350例,女性350例),将其重建为三维虚拟骨骼模型,并截取坐骨耻骨支内侧缘(medial aspect of the ...目的探究深度学习技术在中国汉族人群CT三维重建图像自动性别识别中的可靠性和准确率。方法收集20~85岁汉族人群骨盆CT影像学数据700例(男性350例,女性350例),将其重建为三维虚拟骨骼模型,并截取坐骨耻骨支内侧缘(medial aspect of the ischiopubic ramus,MIPR)特征区域图像。采用Inception v4作为图像识别模型,以初始化学习和迁移学习两种方式进行训练。随机选取80%的图像作为训练验证集,20%的图像作为测试集。将左右两侧MIPR图像进行单独以及合并训练。之后使用总准确率、女性准确率、男性准确率等指标进行模型的性能评价。结果将左右两侧MIPR图像单独进行初始化学习训练,右侧MIPR模型的总准确率为95.7%,其中女性准确率为95.7%、男性准确率为95.7%;左侧MIPR模型的总准确率为92.1%,其中女性准确率为88.6%、男性准确率为95.7%。将左右两侧MIPR图像合并以初始化学习进行训练,模型的总准确率为94.6%,其中女性准确率为92.1%、男性准确率为97.1%。将左右两侧MIPR图像合并以迁移学习进行训练,模型的总准确率为95.7%,其中女性准确率为95.7%,男性准确率为95.7%。结论利用Inception v4深度学习模型和迁移学习算法对中国汉族人群骨盆MIPR图像构建性别推断模型,可对成人骨骼遗骸开展有效性别鉴定,具有较高的准确率及泛化能力。展开更多
Objective： To develop a novel scaffolding method for the copolymers poly lactide-co-glycolide acid （PLGA） to construct a three-dimensional （3-D） scaffold and explore its biocompatibility through culturing Schwann...Objective： To develop a novel scaffolding method for the copolymers poly lactide-co-glycolide acid （PLGA） to construct a three-dimensional （3-D） scaffold and explore its biocompatibility through culturing Schwann cells （SCs） on it. Methods： The 3-D scaffolds were made by means of melt spinning, extension and weaving. The queueing discipline of the micro-channels were observed under a scanning electronic microscope （SEM）.The sizes of the micropores and the factors of porosity were also measured. Sciatic nerves were harvested from 3-day-old Sprague Dawley （SD） rats for culture of SCs. SCs were separated, purified, and then implanted on PLGA scaffolds, gelatin sponge and poly-L-lysine （PLL）-coated tissue culture poly-styrene （TCPS） were used as biomaterial and cell-supportive controls, respectively. The effect of PLGA on the adherence, proliferation and apoptosis of SCs were examined in vitro in comparison with gelatin sponge and TCPS. Results： The micro-channels arrayed in parallel manners, and the pore sizes of the channels were uniform. No significant difference was found in the activity of Schwann cells cultured on PLGA and those on TCPS （P〉0.05）, and the DNA of PLGA scaffolds was not damaged. Conclusion： The 3-D scaffolds developed in this study have excellent structure and biocompatibility, which may be taken as a novel scaffold candidate for nerve-tissue engineering.展开更多
文摘目的探究深度学习技术在中国汉族人群CT三维重建图像自动性别识别中的可靠性和准确率。方法收集20~85岁汉族人群骨盆CT影像学数据700例(男性350例,女性350例),将其重建为三维虚拟骨骼模型,并截取坐骨耻骨支内侧缘(medial aspect of the ischiopubic ramus,MIPR)特征区域图像。采用Inception v4作为图像识别模型,以初始化学习和迁移学习两种方式进行训练。随机选取80%的图像作为训练验证集,20%的图像作为测试集。将左右两侧MIPR图像进行单独以及合并训练。之后使用总准确率、女性准确率、男性准确率等指标进行模型的性能评价。结果将左右两侧MIPR图像单独进行初始化学习训练,右侧MIPR模型的总准确率为95.7%,其中女性准确率为95.7%、男性准确率为95.7%;左侧MIPR模型的总准确率为92.1%,其中女性准确率为88.6%、男性准确率为95.7%。将左右两侧MIPR图像合并以初始化学习进行训练,模型的总准确率为94.6%,其中女性准确率为92.1%、男性准确率为97.1%。将左右两侧MIPR图像合并以迁移学习进行训练,模型的总准确率为95.7%,其中女性准确率为95.7%,男性准确率为95.7%。结论利用Inception v4深度学习模型和迁移学习算法对中国汉族人群骨盆MIPR图像构建性别推断模型,可对成人骨骼遗骸开展有效性别鉴定,具有较高的准确率及泛化能力。
文摘Objective： To develop a novel scaffolding method for the copolymers poly lactide-co-glycolide acid （PLGA） to construct a three-dimensional （3-D） scaffold and explore its biocompatibility through culturing Schwann cells （SCs） on it. Methods： The 3-D scaffolds were made by means of melt spinning, extension and weaving. The queueing discipline of the micro-channels were observed under a scanning electronic microscope （SEM）.The sizes of the micropores and the factors of porosity were also measured. Sciatic nerves were harvested from 3-day-old Sprague Dawley （SD） rats for culture of SCs. SCs were separated, purified, and then implanted on PLGA scaffolds, gelatin sponge and poly-L-lysine （PLL）-coated tissue culture poly-styrene （TCPS） were used as biomaterial and cell-supportive controls, respectively. The effect of PLGA on the adherence, proliferation and apoptosis of SCs were examined in vitro in comparison with gelatin sponge and TCPS. Results： The micro-channels arrayed in parallel manners, and the pore sizes of the channels were uniform. No significant difference was found in the activity of Schwann cells cultured on PLGA and those on TCPS （P〉0.05）, and the DNA of PLGA scaffolds was not damaged. Conclusion： The 3-D scaffolds developed in this study have excellent structure and biocompatibility, which may be taken as a novel scaffold candidate for nerve-tissue engineering.
