Fast auto-acquisition tomography tilt series by using HD video camera in ultra-high voltage electron microscope.
Fast auto-acquisition tomography tilt series by using HD video camera in ultra-high voltage electron microscope.
复制标题
DOI：
10.1093/jmicro/dfu066
复制
发表时间：
2014-11
期刊：
影响因子：
1.8
通讯作者：
R. Nishi;M. Cao;Atsuko Kanaji;T. Nishida;K. Yoshida;S. Isakozawa
中科院分区：
文献类型：
--
作者：
R. Nishi;M. Cao;Atsuko Kanaji;T. Nishida;K. Yoshida;S. Isakozawa
超高压电子显微镜（UHVEM） H-3000在世界最高加速电压为3 MV的情况下，可以观察到微米厚样品[1]的显著三维微观结构。获得倾斜系列的电子断层扫描是一项费力的工作，因此高度需要一种自动技术。我们提出了使用图像清晰度（AFS）的自动对焦系统[2,3]来获取UHVEM断层扫描的倾斜序列。该方法首先获取5幅不同离焦值的图像，计算图像的锐度。然后将清晰度拟合到准高斯函数中以确定最佳焦距值[3]。慢扫描CCD （SS-CCD）相机（日立F486BK）采集的散焦图像质量高，但采集五张散焦图像需要一分钟。在这项研究中，我们介绍了一种高清摄像机（HD video camera; Hamamatsu Photonics K. K. C9721S），用于快速采集图像[4]。它是一台模拟摄像机，但摄像机图像由PC机捕获，有效图像分辨率为1280×1023像素。该分辨率低于SS-CCD相机的4096×4096像素。然而，高清摄像机拍摄一张图像的时间只有1/30秒。为了换取更快的采集，图像的信噪比很低。为了提高信噪比，对采集的22帧图像进行整合，使每幅图像的清晰度都足够，拟合误差更小。为了解决低分辨率问题，我们选择了一个大的离焦步长，通常是手动离焦步长的5倍，来区分不同的离焦图像。通过使用高清摄像机进行自动对焦，将每次自动对焦的时间减少到6秒左右。图像位置的校正时间为1秒，总校正时间为7秒，比SS-CCD相机的校正时间缩短了一个数量级。当我们使用SS-CCD相机进行最后的图像采集时，记录一张倾斜图像需要30秒。我们可以在30分钟内获得61张图像的倾斜序列。准确性和可重复性足够好，可以实际使用（图1）。我们成功地将断层扫描倾斜序列的总采集时间减少了一半。jmicro；63/suppl_1/i25/ dfu066f1f1dfu066f1图。1.物镜电流随断层扫描序列倾斜角度的变化（样本：大鼠肝细胞，厚度：2 m，放大倍率：4k， acc）。电压：2mv)。倾斜角度范围为±60度，步进角2度。在同一地区获得了两个系列。两个数据几乎相同，偏差小于手动的最小步长，因此自动对焦效果良好。我们还开发了用于电子断层扫描的计算机辅助三维（3D）可视化和分析软件“HawkC”，该软件可以半自动分割三维数据[5,6]。如果将该自动采集系统与IMOD重建软件[7]和HawkC软件配合使用，就可以实现UHVEM的在线断层扫描。该系统对今后的病理检查有一定的帮助。本研究得到日本文部科学省（MEXT）科学研究资助项目（批准号：23560024、23560786）和日本科学技术振兴机构SENTAN的支持。
The ultra-high voltage electron microscope (UHVEM) H-3000 with the world highest acceleration voltage of 3 MV can observe remarkable three dimensional microstructures of microns-thick samples[1]. Acquiring a tilt series of electron tomography is laborious work and thus an automatic technique is highly desired. We proposed the Auto-Focus system using image Sharpness (AFS)[2,3] for UHVEM tomography tilt series acquisition. In the method, five images with different defocus values are firstly acquired and the image sharpness are calculated. The sharpness are then fitted to a quasi-Gaussian function to decide the best focus value[3]. Defocused images acquired by the slow scan CCD (SS-CCD) camera (Hitachi F486BK) are of high quality but one minute is taken for acquisition of five defocused images.In this study, we introduce a high-definition video camera (HD video camera; Hamamatsu Photonics K. K. C9721S) for fast acquisition of images[4]. It is an analog camera but the camera image is captured by a PC and the effective image resolution is 1280×1023 pixels. This resolution is lower than that of the SS-CCD camera of 4096×4096 pixels. However, the HD video camera captures one image for only 1/30 second. In exchange for the faster acquisition the S/N of images are low. To improve the S/N, 22 captured frames are integrated so that each image sharpness is enough to become lower fitting error. As countermeasure against low resolution, we selected a large defocus step, which is typically five times of the manual defocus step, to discriminate different defocused images.By using HD video camera for autofocus process, the time consumption for each autofocus procedure was reduced to about six seconds. It took one second for correction of an image position and the total correction time was seven seconds, which was shorter by one order than that using SS-CCD camera. When we used SS-CCD camera for final image capture, it took 30 seconds to record one tilt image. We can obtain a tilt series of 61 images within 30 minutes. Accuracy and repeatability were good enough to practical use (Figure 1). We successfully reduced the total acquisition time of a tomography tilt series in half than before.jmicro;63/suppl_1/i25/DFU066F1F1DFU066F1Fig. 1.Objective lens current change with a tilt angle during acquisition of tomography series (Sample: a rat hepatocyte, thickness: 2 m, magnification: 4k, acc. voltage: 2 MV). Tilt angle range is ±60 degree with 2 degree step angle. Two series were acquired in the same area. Both data were almost same and the deviation was smaller than the minimum step by manual, so auto-focus worked well. We also developed a computer-aided three dimensional (3D) visualization and analysis software for electron tomography "HawkC" which can sectionalize the 3D data semi-automatically[5,6]. If this auto-acquisition system is used with IMOD reconstruction software[7] and HawkC software, we will be able to do on-line UHVEM tomography. The system would help pathology examination in the future.This work was supported by the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan, under a Grant-in-Aid for Scientific Research (Grant No. 23560024, 23560786), and SENTAN, Japan Science and Technology Agency, Japan.