Skip to content

Latest commit

 

History

History
29 lines (14 loc) · 3.97 KB

README.md

File metadata and controls

29 lines (14 loc) · 3.97 KB

Estimation of nuclear surfaces from confocal images

This code estimates the nuclear surfaces of individual nuclei from confocal image stacks containing multiple nuclei, as follows. First, the location of each nucleus in a stack is obtained by projecting the 3D stack into a 2D image and running a 2D active contour algorithm. Next, we removed any incorrectly segmented nuclei manually using a GUI-based script. Finally, we use the location of individual nuclei to segment them in 3D using a 3D active contour algorithm. We utilize the following toolboxes from the Mathworks website

  • Active Contour algorithm by D. Kroon, University of Twente

  • Geometry methods by D. Legland, INRA

Instructions for execution

  1. Create a readme.txt file in each folder containing your image stacks. This file contains information on the filename pattern for each stack of confocal images in a folder. Figure. 1 explains a sample readme.txt file

  1. Open the proj_run2D_manyfolders.m file and enter the path to the folders containing your image data. Change the pxl variable to the size of the image voxel in μm of your images. This information is typically available in the metadata of the image stacks. Then run the proj_run2D_manyfolders.m file, which will read each stack, project the 3D stack into a 2D image file and obtain the 2D boundaries of each nuclei using a 2D active contours algorithm. A MATLAB binary file (i.e., .mat format) is created for each stack, and the corresponding 2D nuclei boundaries are stored. Note: If your images are in RGB format, the program assumes that the nucleus is in Blue. If not, then the line number 25 – zimg = squeeze(tmp(:,:,3)); in read_stack_proj_run2D.m should be updated.

  2. A few nuclei might not be appropriately segmented—these need to be manually removed using the remove_nuclei.fig GUI program. Point to the folder containing your image data and type the filename of your image stack without the extension (For example, Series007). The projected image and the 2D nuclear boundaries will be loaded in the GUI. Left click on any nuclei that hasn’t been appropriately segmented to remove them from the analysis. A red cross on the nuclei shows that it will be removed from the analysis after you have clicked on it. Manually check all the image stacks for nuclei that haven’t been appropriately segmented and remove them. See figure below

  1. Open the volume_manyfolders.m file and enter the folder names containing your image data. Change the pxl variable to the size of the image voxel in μm of your images. Run volume_manyfolders.m. It will run the 3D active contours algorithm on individual nuclei and obtain the nuclear surface as a triangular mesh.

  2. Open the save_folder_areas_vol.m file, edit the fldr variable to the path of your data folder and run the script. This script will calculate the surface mesh's projected area, surface area, and volume of nuclei.

  3. To check if the 3D active contours algorithm has converged correctly to the nuclear surface, use the check_3D_convergence.m script. Edit the fldr variable to the path to the folder containing the image stacks, stck_num to the serial number of the stack and nuclei_num to the serial number of the nuclei. The script will show the orthogonal sections of the image stack superimposed with the nuclear surface mesh and a 3D plot of the nuclear surface mesh.

  4. A corresponding .mat file is created for each image stack. The shape data is stored in a struct array variable named dat. Each element of this array corresponds to individual nuclei. dat.surf is the surface mesh of the nuclear surface, dat.PA is the projected area, dat.SA is the surface area, and dat.V is the volume of the nucleus.