Manual: MasTracker for ImageJ

Transcription

1 Manual: MasTracker for ImageJ Martin Storath 3. Juli

2 1 Introduction The following are instructions for the tracking plug-in MasTracker for ImageJ. MasTracker was implemented by Martin Storath as part of the study project Tracking of Nerve Cells in the Cerebellum of the Zebra Fish at the faculty of mathematics of the TU Munich, under supervision of Prof. Brigitte Forster-Heinlein and Dipl. Math. Stefan Held. The idea for the project came from Dr. Reinhard Köster of the Institut für Entwicklungsgenetik of the Forschungszentrum für Umwelt und Gesundheit (GSF) in Neuherberg, who also provided the test data. 2 Installation Java Runtime Environment version or higher ( and ImageJ version 1.37 or higher ( are required. Install both Java and ImageJ. Move the plug-in File MasTracker.jar and the matrix library mtj.jar into the Plugin folder of ImageJ (ImageJ/Plugins). If not bundled with MasTracker, the library mtj.jar can be downloaded from Run ImageJ and open the plug-in via Plugins MasTracker. 3 Preparation of Images MasTracker operates on image stacks and the images have to be imported first via File Import Image Sequence. Select the desired start image and adjust the increment. The increment should be the size of the recorded image stacks per timestep in the case of 3D data, otherwise set increment = 1. Example: With a confocal miscroscope, you recorded 200 images at 5 different timesteps, i.e. 40 slices per timestep. Thus choose increment = 40 and start picture between 1 and 40. Colour images (RGB) have to be converted to gray value images. In case of only one colour channel, choose Image Color RGB-Split to extract the right channel. For pictures with more than one colour channel select Image Type 8-Bit. Important: Make sure there are no imaging artifacts at the borders, i.e. that there are no uniformly coloured frames around the picture. To remove such borders mark the interior region of the image by a rectangular ROI and crop it by clicking Image Crop. To improve performance, also crop the regions which do not contain particles. 4 Start of Tracking Open MasTracker and import an imagestack (see above). Then execute the following steps: 1. At first, into the field Pixel per Micrometer enter the number of pixels per micrometer (Example: The image resolution is 500x400 pixels and the area displayed by the image is 10x8 micrometer. Thus enter 50(= 500/10 = 400/8) ). Into the field Timestep in min. enter the time (in minutes) that has elapsed between the capture of one image to the next image (Example: The microscope captured every 12 minutes an image of the experiment. Thus enter 12). 2. Choose the first picture in the stack and make a rough estimate of the outline of the cells. To do that choose the Region of interest (Roi) Tool, for example the freehand selection, and draw a rough outline around the cells. The better the estimate, the lower is the computation time. 2

3 3. Now adjust the parameters in the segmentation field (mu, nu, lambda1, lambda2 ) according to the rules in the attached list (see below). Usually, most default parameters work well, but mu is a critical parameter which has to be adjusted empirically. After this, press Start segmentation and wait until the computation has finished. After a few seconds of computation, a picture with the outlines of the found cells will appear. All cell outlines are also saved in the RoIManager. 4. If the automatic segmentation does not deliver satisfactory results, try another set of parameters and repeat the last step or correct the results via the RoIManager. Choose the second option if the cells cannot be separated correctly by the segmentation algorithm. To correct the cell outlines have a look at the RoIManager. By clicking on a RoIManager entry, the outline of the corresponding cell appears in the cell image. If an outline is not correct, delete the corresponding entry. New cells can be added manually by creating a ROI for every cell. Therefore we outline the cell as accurately as possible using the freehand ROI tool and press Add in the RoIManager. Do not close the RoIManager, it is still needed for the next step. 5. The starting point for the tracking is the RoIManager. The outlines that are saved in the RoIManager, represent the cells in the first image. Therefore make sure all cells are outlined as accurately as possible. Then adjust the parameters according to the attached list (see below) and press start tracking. 6. The computation may need a certain amount of time, depending on the number of cells and the size of the images. The current process status appears in the text field. After the computation a table with several results (see list below) and a picture with the traces of the cell will appear. To control the results, check again the RoIManager where all outlines of all cells in every image are saved. The following data is given as a result of tracking: Index: Unique index of a cell, corresponds to the number printed in the image with the traces. Track from: Image index of first appearance of the cell. Track until: Image index of last appearance of the cell. Moved Way: Total moved distance of the cell. Moved Distance (air-line) Total moved distance of cell (airline!). Average Velocity Direction of Movement (angle): Direction of the movement in 0 o 360 o angle. Average Polarisation: Average polarisation, where polarisation = height / width. Cell Division To Indices: If this entry is not empty, the cell has divided and the numbers in the field are the indices of the daughter cells. Average Directionality: The sum of all absolute values of the changes of direction (in degrees) of cell movements divided by the number of timesteps. 3

4 Cell Contacts: A printout of all cell contacts of this cell in format C[Cellindex of touched cell]: [Contact timepoint] Cont. Area [Average contact area of the 2 cells]. Example: Index 2,..., C 5: 10 min - 30 min, min, 30 um C 3: 10 min means that cell 2 had contact with cell 5 at timepoints from 10 to 30 min and 60 to 70 min and with cell 3 at timepoint 10 min. Influences of the parameters: mu Big mu: smooth curves are preferred, prefers finding bigger cells Small mu: non-smooth curves are possible, prefers finding smaller cells Default: 0.01 (most critical parameter!) nu Big nu: prefers finding a small total cell area (sum of areas of all cells) Small nu: big total cell area is possible Default: 0 (usually default value works well) lam1 Big lam1: prefers curves with smooth interior (regarding brightness) Small lam1: non-smooth interior is possible Default: 1 (usually default value works well) lam2 Big lam2: prefers smooth exterior of the curve (regarding brightness) Small lam2: non-smooth exterior possible Default: 1 (usually default value works well) gamma Big gamma: prevents curves from overlapping Small gamma: overlapping curves are possible Default: (usually default value works well) eta Big eta: forces cells to keep their initial volume Small eta: allows cells to change their initial volume Default: 10 (change of this parameter may deliver better results) Additional options: Size of Narrowband Has influence on the computation time. Should not be too far from 10 but not lower than 6 and not higher than 20. Default: 10 (usually, default value works well) Minimum Cell Size The minimum size of a cell in pixels (!). If the minimum size is unknown, set to 20 for stability reasons. Default: 10 Minimum Difference to Background The minimum brightness distance between cells and background. If the average brightness inside a cell and the average brigthness of the background differ less than this value, the cell outline is eliminated. A high value allows cells to vanish during tracking in the interior of the image. If set to zero, the cell outlines cannot vanish in the interior. Default: 20 4

5 Frequency of Searching for New Cells The frequency of a search for appearing cells. If set to zero, new cells are never searched. Example: A Value 5 means that for every fifth image a search for new cells is executed. Default: 1 Max. Iterations Upper bound for the number of computation iterations per timestep. If the computation time is too high, decrease this value, but to values lower than 150. Default: 200 stationary count The algorithm stops processing a cell in every single image if the corresponding segmentation curve has not moved for a certain number of iterations. This number is given by the parameter stationary count. Default: 3 (Usually does not need to be changed.) Fix c in c out If checked, the average colour inside and outside the objects remains fixed during a timestep. This allows to track cell displacements of up to full cell diameter, but average cell brightness must not change from image to image significantly. If unchecked, only cell displacements of half a cell diameter can be tracked correctly, but brightness inside the cells may change significantly. Default: Checked 5 Suggestions, Questions and Bug Reports Please mail suggesttions for improvement, questions or bug reports to storath(at)in.tum.de. 5