Modeling and Visualization Techniques for Analysis of 3D/4D Cell Images
Description
The research aims to develop modeling and visualization techniques for three- and four-dimensional living cell tomography, and to develop a novel cell morphological analysis technique for studies in Life Science and Biomedicine. Through the use of image segmentation and modeling techniques, we provide a method for measuring properties of the complex cell including physical measure, shape, and cellular interaction. Furthermore, we investigate effective visualization for living cell data based on characteristics of the 3D cell image.
Figure 1. Optical diffraction tomography of a living cell.
Figure 2. The appearance of Huh-7 cell and gradient magnitude of the original image.
We focused on building the shape database for various types of cells. For the template model construction, we develop an image registration technique for 3D cell image, applied to the mean shape by preserving the morphology of structures of the cell. The subject-specific deformable modeling technique will be adapted for image segmentation as well as a modeling technique for some specific cell and cellular organelles. From the preliminary study on the shape of the cell, we figured out that the general appearance of the cell cannot be normalized in terms of morphology due to its fluidness and inhomogeneous figures.
To cover a variety of cellular characteristics, we first utilized both the original image and the gradient image when visualizing the 3D shape of the cell. The color mapping process in the volume rendering technique which is 2D transfer function can be managed with an intuitive rectangular drawing interface so that the user can come up with diverse visual outcomes of cell morphology conveniently. Furthermore, we proposed a model approximation method that reconstructs the quantitative 3D boundary of the cellular organelle from the visible features of the volume rendering result.
Figure 3. Volume rendering result of B-cell and 2D transfer function canvas.
Figure 4. Various visualization results of 3D cells.
Our cell modeling and visualization methods will encourage the optical diffraction tomography of cell images in the clinical area. It will extend the morphological analysis of cells from the contour of the cell to detailed changes of cell structures. Using those results, we can attain new clinical measures regarding cell diseases and impairments. Also, by constructing a database for labeled cell images and models, we can provide a base technique to large-scale analysis of the cells. Moreover, our cell visualization framework will contribute to the biomedical industry in the form of labor-intensive tasks of manual disease and impairment analysis.
Figure 5. 3D visualization based cellular boundary reconstruction of a red blood cell.
Figure 6. Clustering-based 2D Transfer Function for Volume Visualization of 3D Cell Image Data.
Figure 7. Lipid Droplet Tracking Method using Optical Flow in 3D Cell Image Data.
Contact
Taeho Kim (kdhtheo at kaist.ac.kr)
Publications
Seoyoung Kim, Taeho Kim, Jinah Park, "Modeling 3D Cell Nucleus by Template-based Deformable Model with Confined-region Determined by Cellular Characteristics," International Forum on Medical Imaging in Asia (IFMIA) 2017, pp. 17-20, 2017.
Jihoon Cho, Taeho Kim, and Jinah Park, "2D transfer function design for live-cell image data using ISODATA clustering method," BIEN 2017, P-IT-033, August 2017.
Taeho Kim and Jinah Park, "Analyzing 3D Cell Data of Optical Diffraction Tomography through Volume Rendering," International Workshop on Advanced Image Technology (IWAIT) 2018, January 2018.
Taeho Kim and Jinah Park, "Effective Volume Rendering and Virtual Staining Framework for Visualizing 3D Cell Image Data," Journal of the Korea Computer Graphics Society, Vol. 24, No. 1, pp. 9-17, March 2018.
