Time-lapse Image Analysis

Check Visualizations First

All of the results from the image processing steps are saved in the zip file you downloaded. I would start by looking at some of the visualization results to ensure that your image set was properly analyzed. Please note that all the visualization images have been scaled so that the maximum intensities are truncated at the bottom and top 0.1% of pixel intensities for puncta and ECM images. In the 'visualizations' folder you will find several subfolders:

  • side_by_side: The input images placed side by side.
  • overlap: The input images overlayed on another, puncta in green, ECM in red.
  • tracking: Each identified puncta is outlined using a unique color, which stays with that puncta for the duration of the experiment. Colors are reused as old puncta die and new puncta form.
  • invado_and_not: Puncta are outlined in this time-lapse based on whether the computer has classified the puncta as an invadopodia (green), not an invadopodia (red) or as not being classifed (blue). The unclassified invadopodia are those that didn't pass the minimum lifetime filter. When a long-lived puncta is born, a number appears next to the puncta in the ECM channel. This number is a unique identifer for that puncta and can be used to lookup additional information about the puncta.
  • single_invadopodia: Small multiple visualization of each puncta classified as an invadopodia. The file names are unique numbers which can be used to lookup additional information about each puncta, see the following section about additional measurements. Two versions of each small multiple is available, one with the puncta outlined another without the puncta outlined.
  • single_invadopodia: Small multiple visualization of the puncta not classified as invadopodia.

Does it look like the software correctly found and tracked the puncta? You will also want to make sure the photobleaching and intensity normalization worked for the ECM intensity. Does the ECM in non-degraded regions appear consistant in intensity?

Invadopodia Properties

After verifying the processing results, you can take a look at the numerical results. Most of the results are in two files in the 'puncta_props' directory: 'invado_data.csv' and 'not_invado_data.csv'. Hopefully this is clear from the filenames, but 'invado_data.csv' has information about the puncta classifed as invadopodia, while 'not_invado_data.csv' has information about the puncta not classified as invadopodia. Each file can be read by nearly any statistical program (including excel) and contains the following data:

  • lineage_nums: A unique number which identifies each tracked puncta. The same identification numbers are used in the visualizations.
  • longevity: The number of minutes the puncta was present in the experiment. A longevity value of "NA" indicates that the software can't be certain of the actual puncta longevity because it was present at the beginning or end of the experiment.
  • hit_max_time: The number of minutes it took for the puncta to degrade 90% of it's maximum average intensity. This property is calculated for non-invadopodia, but probably isn't relevant.
  • mean_area: Average of the areas measured for the puncta. Assuming you specified the size of your pixels when submitting the experiment, the units will be µm².
  • mean_local_diff: The average of the local difference in the ECM intensity inside versus outside the puncta. The values used to find the average are calculated by subtracting the average ECM intensity under the puncta from the average intensity in a region surrounding the puncta. The units are arbitrary.
  • p_value: The p-value as determined by a T-test of the local difference values compared to zero. This is used as a filter to classify puncta as invadopodia or not.
  • mean_local_diff_corrected: The average of the local difference metric corrected by the intensity of the ECM before the puncta was born. If the puncta was present in the first image, then the correction values are taken from the first image. The units are arbitrary.
  • local_diff_corrected_p_value: The p-value for the corrected local diff values, using the same methods as the p_value column.

Everything Else

The software also produces a range of intermediate processing steps, I'll quickly walk through what you can expect to find. The 'individual_pictures' directory contains the results of the image processing pipeline. The 'tracking_matricies' directory holds the tracking matrix produced that allows the puncta to be tracked through time. The 'errors' directory has a record of the processing command executed and any associated errors thrown by the pipeline.

Fixed Image Analysis

The fixed image analysis results are all contained in the 'individual_pictures' folder. In this folder, you should find a seperate folder for each of the images in the image set you submitted. Each of the image folders contains the results of processing for that image. Take a look at 'puncta_highlight.png', you should see the puncta outlined in red and the cell edge in green. If not, you may want to go back and resubmit your images with different processing settings. All of the numerical results are in 'raw_data'.

Numerical Results

All of the numerical results are in units of pixels, even if you specified the size of your pixels in the image submission process. Also, the results for identified puncta are stored in the same row across each file. Thus, the data associated with puncta #1 is always in row #1 of the area, centroid, etc. data files. Each file is as follows:
  • Area.csv: The area in pixels of each puncta
  • Average_puncta_signal.csv: The average intensity of the pixels in each puncta
  • Centroid.csv: The location of each puncta's centroid
  • Centroid_dist_from_edge.csv: The distance between each puncta's centroid and the nearest cell edge in pixels. There might be some 'NaN' values, these indicate that the nearest "cell edge" can't be determined because the puncta is too near the edge of the field of view.
  • Eccentricity.csv: The eccentricity (a measurement of major versus minor axis) of the puncta
  • gel_background_intensity.csv: The average intensity of the ECM in the region within the local background region of the puncta. The local background region is defined as the area within five pixels of the puncta, that aren't another
  • gel_intensity_puncta.csv: The average intensity of the ECM underneath each puncta.
  • Local_gel_diff.csv: The difference between the background the puncta ECM intensities. Positive values indicate that the background region is of higher intensity than the region underneath the puncta.
You can match between the results in these files and the images using the 'gel_nums.jpg' and 'puncta_nums.jpg' files. These visualizations have the unique puncta numbers in blue next to each puncta. The image size is blown up 4 times (to make sure clustered puncta numbers can be identified), so you will probably have to zoom in to pick out each number.

Picking Out Invadopodia/Not Invadopodia

I would use the values in Local_gel_diff to determine a threshold at which to call a puncta an invadopodia or not invadopodia. Determining this threshold is a bit subjective, but I would attempt to find several examples of characteristic invadopodia. From there, I would lookup each puncta's identifiying number (see the 'gel_nums.jpg' and 'puncta_nums.jpg' files) and compile a set of verified puncta Local_gel_diff invadopodia values. This value for invadopodia should be positive, indicating that the ECM is more intense on average outside the puncta than immediately underneath it.