银杏胚的有无及发育状况直接影响播种后的发芽率,而这些无法人工检测。该研究基于低场核磁共振成像(Low-field Magnetic Resonance Imaging,LF-MRI)技术,结合深度学习,探索研究银杏种子内部缺陷检测方法。设计了一种基于改进的VGG-16迁...银杏胚的有无及发育状况直接影响播种后的发芽率,而这些无法人工检测。该研究基于低场核磁共振成像(Low-field Magnetic Resonance Imaging,LF-MRI)技术,结合深度学习,探索研究银杏种子内部缺陷检测方法。设计了一种基于改进的VGG-16迁移学习图像分类方法,在利用VGG-16卷积层对数据集进行迁移学习时,对模型的学习率、模型结构和归一化3个方面进行改进,比较研究调整初始学习率与更新周期、将全连接层替换为全局卷积层或全局平均池化层并扩展8种微调模型、添加3种归一化层对模型性能的影响。结果表明,全局平均池化层调用方便,能够替代全连接层融合深度特征,提升模型性能;分类输出层置于全局平均池化层后的分类效果较好;组归一化能提升模型性能,并使验证曲线收敛更稳定;学习率对迁移学习模型性能影响极大,使用分段常数衰减策略针对具体应用可有效提高模型性能。相比较VGG-16,该研究构建的基于VGG-Net的银杏种子MR图像识别模型(Ginkgo seed low-field MR images recognition model based on VGG-Net,GV)大小降低89%,参数量降低89%,训练时间减少64%,训练集与验证集损失分别降低64%与45%、准确率分别提高2.4与2.5个百分点,对正常、孔洞、有胚和无胚4个类别的银杏种子平均分类准确率达到97.40%。研究结果为无损监测银杏种子萌发过程、合理预测种子播种后的发芽率提供了思路。展开更多
The aluminum shielded room has been an important part of ultra-low-field magnetic resonance imaging (ULF MRI) based on the superconducting quantum interference device (SQUID). The shielded room is effective to att...The aluminum shielded room has been an important part of ultra-low-field magnetic resonance imaging (ULF MRI) based on the superconducting quantum interference device (SQUID). The shielded room is effective to attenuate the external radio-frequency field and keep the extremely sensitive detector, SQUID, working properly. A high-performance shielded room can increase the signal-to-noise ratio (SNR) and improve image quality. In this study, a circular coil with a diameter of 50 cm and a square coil with a side length of 2.0 m was used to simulate the magnetic fields from the nearby electric apparatuses and the distant environmental noise sources. The shielding effectivenesses (SE) of the shielded room with different thicknesses of aluminum sheets were calculated and simulated. A room using 6-mm-thick aluminum plates with a dimension of 1.5 m x 1.5 m x 2.0 m was then constructed. The SE was experimentally measured by using three-axis SQUID magnetometers, with tranisent magnetic field induced in the aluminum plates by the strong pre-polarization pulses. The results of the measured SE agreed with that from the simulation. In addition, the introduction of a 0.5-mm gap caused the obvious reduction of SE indicating the importance of door design. The nuclear magnetic resonance (NMR) signals of water at 5.9 kHz were measured in free space and in a shielded room, and the SNR was improved from 3 to 15. The simulation and experimental results will help us design an aluminum shielded room which satisfies the requirements for future ULF human brain imaging. Finally, the cancellation technique of the transient eddy current was tried, the simulation of the cancellation technique will lead us to finding an appropriate way to suppress the eddy current fields.展开更多
文摘银杏胚的有无及发育状况直接影响播种后的发芽率,而这些无法人工检测。该研究基于低场核磁共振成像(Low-field Magnetic Resonance Imaging,LF-MRI)技术,结合深度学习,探索研究银杏种子内部缺陷检测方法。设计了一种基于改进的VGG-16迁移学习图像分类方法,在利用VGG-16卷积层对数据集进行迁移学习时,对模型的学习率、模型结构和归一化3个方面进行改进,比较研究调整初始学习率与更新周期、将全连接层替换为全局卷积层或全局平均池化层并扩展8种微调模型、添加3种归一化层对模型性能的影响。结果表明,全局平均池化层调用方便,能够替代全连接层融合深度特征,提升模型性能;分类输出层置于全局平均池化层后的分类效果较好;组归一化能提升模型性能,并使验证曲线收敛更稳定;学习率对迁移学习模型性能影响极大,使用分段常数衰减策略针对具体应用可有效提高模型性能。相比较VGG-16,该研究构建的基于VGG-Net的银杏种子MR图像识别模型(Ginkgo seed low-field MR images recognition model based on VGG-Net,GV)大小降低89%,参数量降低89%,训练时间减少64%,训练集与验证集损失分别降低64%与45%、准确率分别提高2.4与2.5个百分点,对正常、孔洞、有胚和无胚4个类别的银杏种子平均分类准确率达到97.40%。研究结果为无损监测银杏种子萌发过程、合理预测种子播种后的发芽率提供了思路。
基金Project supported in part by the Strategic Priority Research Program(B)of the Chinese Academy of Sciences(Grant No.XDB04020200)in part by the National Natural Science Foundation of China(Grant No.11204339)
文摘The aluminum shielded room has been an important part of ultra-low-field magnetic resonance imaging (ULF MRI) based on the superconducting quantum interference device (SQUID). The shielded room is effective to attenuate the external radio-frequency field and keep the extremely sensitive detector, SQUID, working properly. A high-performance shielded room can increase the signal-to-noise ratio (SNR) and improve image quality. In this study, a circular coil with a diameter of 50 cm and a square coil with a side length of 2.0 m was used to simulate the magnetic fields from the nearby electric apparatuses and the distant environmental noise sources. The shielding effectivenesses (SE) of the shielded room with different thicknesses of aluminum sheets were calculated and simulated. A room using 6-mm-thick aluminum plates with a dimension of 1.5 m x 1.5 m x 2.0 m was then constructed. The SE was experimentally measured by using three-axis SQUID magnetometers, with tranisent magnetic field induced in the aluminum plates by the strong pre-polarization pulses. The results of the measured SE agreed with that from the simulation. In addition, the introduction of a 0.5-mm gap caused the obvious reduction of SE indicating the importance of door design. The nuclear magnetic resonance (NMR) signals of water at 5.9 kHz were measured in free space and in a shielded room, and the SNR was improved from 3 to 15. The simulation and experimental results will help us design an aluminum shielded room which satisfies the requirements for future ULF human brain imaging. Finally, the cancellation technique of the transient eddy current was tried, the simulation of the cancellation technique will lead us to finding an appropriate way to suppress the eddy current fields.