Sa Y, Zhang Y, Li R, Huang Y, Zhang Y, Hu X, Feng Y
Tianjin University, Tianjin Key Lab of BME Measurement Technology.
Tianjin Medical University Cancer Institute & Hospital Dept. of Radiation Oncology.
Med Phys. 2012 Jun;39(6Part27):3950. doi: 10.1118/1.4736121.
To develop a new interpolation method for accurate 3D reconstruction of cell morphology from laser scanning confocal microscope (LSCM) image data.
Current techniques are based on the assumption that pixel intensity or contour shapes of images change linearly in the interpolation direction. Gray-value and position of the pixel in interpolated image slice are obtained through weighted average calculation with gray-values and distances of corresponding pixels in two adjacent original image slices, only information from adjacent image slices is considered, often fail to meet the need of 3D reconstruction for cells because of the complex cell morphology. The new method interpolates cellular organelle contours in polar coordinate system. Coordinate system origin is chosen to be the mass center weighted by pixel intensity instead of conventional geometric center, contour points of the organelle is sampled by their angles first and fitted with uniform cubic B-spline to perform interpolation. For complex organelle structures such as branched nuclei, a special method combining morphological information and corner detection technique based on curvature scale space has been developed to solve the contour division and related problems. New method was applied to confocal images of 130 different cells acquired with an LSCM system (LSM510, Zeiss), sampling step was set as 0.5 μm in longitudinal direction, pixel size in horizontal plane was 0.07 μm and the resolution was 512×512. Marching cubes algorithm was used for 3D reconstruction.
Experiments showed that reconstructed 3D images with new method have much smoother and more valid organelle surfaces for both cytoplasm and nucleus than those from conventional methods.
The new interpolation method can significantly improve the quality of 3D reconstruction and serve as a valid and effective tool for quantitative study of 3D cell morphology in radiation biology and other areas of life science.*support by NSFC- 81171342. Supported by the National Science Foundation of China (NSFC- 81171342).
开发一种新的插值方法，用于从激光扫描共聚焦显微镜（LSCM）图像数据精确重建细胞形态的三维结构。
当前技术基于图像的像素强度或轮廓形状在插值方向上线性变化的假设。通过对两个相邻原始图像切片中对应像素的灰度值和距离进行加权平均计算，得到插值图像切片中像素的灰度值和位置，由于仅考虑相邻图像切片的信息，且细胞形态复杂，往往无法满足细胞三维重建的需求。新方法在极坐标系中对细胞器轮廓进行插值。坐标系原点选择为由像素强度加权的质心而非传统的几何中心，首先按角度对细胞器的轮廓点进行采样，并用均匀三次B样条曲线拟合以进行插值。对于如分支细胞核等复杂的细胞器结构，已开发出一种结合形态学信息和基于曲率尺度空间的角点检测技术的特殊方法来解决轮廓分割及相关问题。新方法应用于使用LSCM系统（LSM510，蔡司）采集的130个不同细胞的共聚焦图像，纵向采样步长设置为0.5μm，水平面像素大小为0.07μm，分辨率为512×512。采用移动立方体算法进行三维重建。
实验表明，与传统方法相比，新方法重建的三维图像在细胞质和细胞核的细胞器表面都更加平滑且有效。
新的插值方法可显著提高三维重建质量，成为辐射生物学和其他生命科学领域中细胞三维形态定量研究的有效工具。*由国家自然科学基金81171342资助。中国国家自然科学基金（NSFC - 81171342）资助 。
