Image Filters - Fixed and Live Data

Introduction

This technical note will show you how to analyse the area or a wound/scratch from a wound/scratch assay. The first part will take you through analysis of a fixed, high quality image. The second will take you through the analysis of a time-lapse data set. 

Time-lapse capture and analysis of wound assays can provide much more information than a simple end point readout. It can provide information about the dynamics of the wound closure over time. Due to the compromises that have to be made when capturing live cells (low light intensities, short exposures etc.) the quality of the images for analysis are usually never as good as a fixed sample. Because of this several extra filtering steps need to be incorporated in the analysis to extract the area of interest for analysis.

This technical note is a useful demonstration in the use of image filters (background subtraction, binary open etc.) in making an image amiable to analysis

Fixed Sample Analysis – Fluorescence

Selecting the Scratch

1. Open T0.tif from the Demo Images\Widefield\Wounding\Fixed folder 

2. To be able to measure the wound it needs to be thresholded. Go to Image🡪Adjust🡪Threshold 

3. Configure the Threshold dialog box as follows 

4. The result should be the area of the scratch selected in red 

Measuring the Scratch and Logging the Data

1. To measure the area of the scratch the right measurements first need to be set. Go to Analyse 🡪 Set Measurements… 

2. Select Area Fraction, Limit to Threshold and Display Label as shown below and press OK 

3. Go to Analyse 🡪 Measure or press Ctrl + M 

4. The percentage area of the wound will displayed in the Result log window 

NOTE: The results in the result window can be saved to an Excel or TXT file

5. Close the open image and repeat the previous steps with T24.tif from the Demo Images\Widefield\Wounding\Fixed folder

NOTE: The threshold required to select the wound may be slightly different 

Removing Unwanted Area from the Analysis

Close all previously opened images. Open and threshold T0.tif again. Notice that there are areas that are not part of the scratch being selected as well. This may be real gaps that should be analysed, but if they are not they can be removed in two different ways. 

Method 1 – Size exclusion

By using a variation of the Measure option from above, called Analyse Particles, objects can be excluded from the analysis based on their size or shape.

1. Go to Analyse 🡪 Set Measurements… deselect Area Fraction and select Area. Press OK 

2. Go to Analyse 🡪 Analyze Particles 

3. Configure the Analyse Particles dialog as below and press OK 

4. You will get a new image that will show the outlines (and associated object number) of each object counted. The data will also be in the results table. 

5. If you scroll through the results list you will notice that all the smaller chunks have an area of less than 500 pixels. Close the outlines image and go to Analyse Particles again. This time change the Size (pixel^2): value to 500-Infinity. Also tick the Summarise box. Press OK 

The result this time is only 3 parts of the image measured that are greater than 500 pixels in area. Also the summary results list the percentage area covered by the selected bits.

Method 2 – Using Image Filters

Instead of excluding objects from the analysis based on their size. They can be removed prior to analysis with morphology filters.

1. Open and threshold T0.tif again 

2. Go to Process 🡪 Binary 🡪 Make Binary 

3. The result will be a black mask on a white background of the thresholded area (this may be reversed on some versions of Fiji). 

4. If you want to make the mask white for the positive areas, which most people prefer. Press the LUT button and select Invert LUT 

5. The best filter for removing small round blobs (like the falsely selected regions in this image) is an Close filter. Go to Process 🡪 Binary 🡪 Options 

6. In the Binary Options dialog, set it as follows and press OK. Play around with the settings to see what different filters and setting do to the image. 

7. The resulting image will be a crude mask that represents the scratch with the smaller erroneous parts removed 

8. This image can now be thresholded and measured as before.

Binary Reconstruction of the Mask

1. Keep the crude mask open. Reopen T0.tif and recreate the initial mask with all the erroneous fragments 

2. Go to Plugins 🡪 Morphology 🡪 Binary Reconstruct 

3. Set the Mask to the original mask with the fragments and the Seed as the open filtered mask. Configure the other options as below. 

4. The result will be a clean mask that accurately represents the scratch that can be measured as before 

Create an Outline of the Result

Sometimes it is useful to have an outline of the segmented result placed on the original image so you can show what was actually measured from the original.

1. Open T0.tif again and go to Image 🡪 Duplicate 

2. Press OK in the resulting Duplicate… dialog 

3. The result should be two copies of the original image 

4. Select one of the image (in this case T0-1.tif) and threshold it as before. Use the analyse particles command again with the size exclusion filter on. But this time change the Show value to Masks. Press OK. 

5. The result should be a binary mask of the analysed area. It may be inverted depending on your version of Fiji. If it is not inverted (Black wound, white cells) invert it now. 

6. Select the mask and go to Process 🡪 Binary 🡪 Outline 

7. Invert the outline image to get a white outline representing the edges of the wound. 

8. The next steps will involve merging the outline and original images together. For this to work the two images need to have the same bit depth. As this will be just an overview image to show a result, dropping the original image down to 8bit is the best option. 

Select the T0.tif image and go to Image 🡪 Type 🡪 8 bit 

9. Go to Image 🡪 Colour 🡪 Merge Channels… 

10. In the Merge Channels dialog, configure it as below and press OK. This will give you a composite image with a green outline around the measured wound. If you want a different coloured outline just put the mask image into a different colour channel. 

11. The resulting image is a composite (2 channels in a stack) image. To make it compatible with PowerPoint etc. it needs to be converted to RGB. Select the merged image and go to Image 🡪 Type 🡪 RGB Colour 

Fixed Sample Analysis – Brightfield

Sometimes it is not possible to fluorescently label a sample, in these cases it is still possible to analyse the scratch. To do this in must first be turned into a pseudo fluorescent image.

1. Open Phase Wound.tif from the Demo Images\Widefield\Fixed folder 

2. The first step is to define the edges of the cells. Go to Process 🡪 Find Edges 

3. To sharpen the image up a bit go to Process 🡪 Sharpen 

4. To remove a lot of the roughness in the image and make it easier to work with go to Process 🡪 Filters 🡪 Gaussian Blur… 

5. In the Gaussian Blur… dialog tick the Preview box and enter a value into the Sigma (Radius) box. Try different values to see what happens, for this image a value of 5 works quite well. Press OK 

6. The resulting image, while looking blurry, can be now used like the previous fluorescent images to select the wound and measure it. 

7. You can also generate and outline and place it back on the original image to show the validity of the analysis 

Live Sample Analysis

Measuring the wound closure rate in live imaged sample can provide a lot of power to any analysis but the images used are not always the best due to compromises that must be made to achieve a live image in the first place.

In this example we will measure the wound closure rate over 24 hours, but the image quality is not great. The image has an uneven background and bleaches over time.

Selecting the Scratch

1. Open Control.tif from the Demo Images\Widefield\Wounding\Live folder

If you play through the stack you will see the cells migrate but you will also notice that the image dims over time.

2. Threshold the first plane. You will notice that it is fairly good at picking up the scratch of the first plane (values used in this example are 270 to 509).

3. Select plane 17 and try to adjust the threshold to select the wound. Notice that the threshold fades as you change planes. This is because the image gets dimmer over time due to photobleaching during capture.

You will also notice that it is not possible to evenly select the scratch due to uneven illumination.

4. Reset the threshold and go to Process 🡪 Subtract Background… 

5. In the Subtract Background… dialog box tick the Preview box and try different Rolling Ball Radius sizes. The larger the rolling ball the larger the objects that will be left behind but the less dramatic the background subtraction. For this image a value of 300 works well. 

6. Press OK and press yes on the next window that opens asking if you want to process all planes. 

7. If you now go to plane 17 and try the threshold again you will see that while it can now be evenly applied it selects a lot of space between all the cells. This is because this information was removed by the background subtraction step. 

8. Reset the threshold and got to Process 🡪 Filters 🡪 Gaussian Blur… 

9. Set the filter size to 10 and press OK. Once again say yes when asked to process all planes. 

10. Now you should be able to evenly threshold the wound on plane 17. 

11. Now navigate to the last plane and set the threshold to select the wound. Make sure the Stack Histogram box is ticked otherwise things can go funny. A range of 16 to 55 works well. 

12. Play through the stack and notice that the threshold is now selecting the wound in all the planes of the image evenly. 

Measuring Scratch Area

1. Go to Analyse 🡪 Set Measurements… and configure it to measure the Area Fraction and make sure it is set to Limit to Threshold. Press OK 

2. Go to Analyse 🡪 Analyse Particles… 

3. In the Analyse Particles dialog Tick the Clear Results and Summarise results boxes. Also make sure the Show: value is set to Masks as we will use this later. Press OK. Once again agree to process all planes. 

4. You will get the result for all the planes in a summary table 

Creating Overview Movie of Result

1. To create an outline movie of the scratch select the output mask and go to Process 🡪 Binary 🡪 Outline as before. Agree to processing all planes. 

2. Invert the LUT 

3. Reopen Control.tif, convert it to 8 bit and merge the outline and 8 bit converted version 

4. Convert the Composite image to RGB and go to File 🡪 Save As 🡪 AVI… 

5. In the Save as AVI… dialog set the Compression to JPG and the Frame Rate to 5 

6. Choose somewhere to save the file and press Save 

7. You should end up with a movie that can be played in your favourite video player or embedded into a presentation