Hey everyone,
This week we have a research update blog, I know right it has been a while. Took some time to get back into the swing of things, has been a long few weeks but I have now started compiling new data from my Mistastin impact melt rock samples. I am looking to quantify the vesicularity, clast content, and melt matrix composition of the impact melt rocks from different locations across the impact crater. The data was collected using a semi-automatic digital image analysis method described and tested by Chanou et al. (2014), and energy dispersive spectroscopy (EDS) from the electron micro-probe lab at Western University.
Chanou, A., Osinski, G.R. and Grieve, R.A.F., 2014. A methodology for the semi‐automatic digital image analysis of fragmental impactites. Meteoritics & Planetary Science, 49(4), pp.621-635.
Let us get into the details about the steps of the semi-automatic digital image analysis method.
Image Analysis Method
The purpose of the image analysis method is to quantify the vesicularity, and the shape, size and distribution of the clasts in the melt rocks. Quantifying these details will tell us more about the physical properties of impact melt rocks, which can provide insight into the cooling processes of impact melt. Now let us look into the method.
Chanou et al (2014) used a semi-automatic digital image analysis method to highlight, separate, and measure the features of interest (FOI) on polished rock faces, scanned thin sections, and backscattered electron (BSE) images. Her method uses the program ImageJ and Adobe Photoshop, and involves a series of steps to obtain quantified results.
How about we go through the steps one at a time, it is easier to show some visual aid.
1.
The first step involves collecting high resolution images of samples, whether it is a thin section, BSE image, or polished rock face. Either three can work for this method but depending on the FOI's you are studying, some are more effective. For example, if you are wanting to study plagioclase microlites you would want to use thin section or BSE images. In my study I am using polished rock faces images since they show the vesicles and clasts clearly over the scanner.
A few of the Mistastin impact melt rock samples I am studying are in the slideshow above. The slideshow includes clast-rich melt rocks overlying impact melt-bearing breccia units, clast-poor melt rocks, and vesicular melt rocks.
2.
The high resolution images are opened onto ImageJ to begin highlighting the FOI's. First step, the brightness and contrast is adjusted to make the FOI's stand out and the background noise is removed from the image.
Second step, the channels of the image are split into the three colour channels red, blue, and green. We have to split the channels because some FOI's stand out a lot better under different channels. For vesicles, the red channel image works best while the red channel image works best for clasts. I will use the blue channel image as the example for this process. To highlight the FOI's or vesicles in this case, a pixel threshold needs to be established. Before that can happen, a scale needs to be set. After the samples were scanned each one was assigned a scale to make measurements a lot easier. The scale was added to ImageJ, so when I would begin the quantitative analysis I will be able to calculate values such as the area and aspect ratio of the vesicles, and total area of the sample.
Setting a threshold involved adjusting the pixel histogram. By doing this, I was able to highlight the pixels specific to the vesicles (red in the image below) and separate them from the rest of the image. This created a black and white image. In the imaages below, you will notice that the vesicles are highlighted black after the threshold is set.
3.
The final step is using the analyse particles tool in ImageJ, which can measure the dimensions of the vesicles. I am interested in the aspect ratio and area so I set ImageJ to calculate the length, width and area of the highlighted vesicles. However, sometimes during the threshold stage other features can be mistaken as vesicles. Voids, fractures, and weathered clasts leave holes and spaces in impact melt rocks. In the images, their pixel values are very similar to that of vesicles. So before I can begin calculations I need to remove these features. Originally, I was going to open the images onto Adobe Photoshop and manually remove them. I was later informed by a friend and colleague that I can manually remove features on ImageJ! This made everything a little bit easier since it meant I could still work on the same computer and not have to transfer all of my images to a computer in a different lab. Removing features on ImageJ is pretty straight forward. Using the brush and fill tool I can cover features I do not want in white and fill in features I want to analyze later in black (any colour will work, I just chose black and white for ease). I will go over this step in more detail in my next blog as I have only just begun going through the samples and editing them.
The gif above summarizes the steps taken to get an image where the vesicles are highlighted. Starting from the original scanned colour image we move to the red channel image which highlights the vesicles relatively well (better than the green and blue channel). We set a threshold to create an image where only the pixels in the vesicles are highlighted black. With the threshold image we can then set the measurements in ImageJ to tell it to calculate a certain size range in mm. The measurements give us the area of each highlighted vesicle, perimeter, circularity, etc.
Over the past month, I have developed a new found respect for researchers who apply any form of image analysis to their research. It requires a lot of dedicated time going over the samples to ensure you have all FOI's you are wanting to analyze, making sure the pixel threshold is set correctly as to not miss any FOI's, and remaining fixed to a computer screen for days since the editing process is very tedious and long.
I think that is everything from me this week. It is going to be a quiet week next week as unfortunately I will not be presenting at the Lunar and Planetary Science Conference (LPSC) this year. I am a little upset about missing this year because it is the 50th anniversary of LPSC. However, I am in the stage in my research where I need to collect new data because at this moment I have nothing new to present at conferences so it would in the end be a wasted trip. I have other conferences to consider this year. In September, there is the annual Geological Society of America meeting in Phoenix, and the Large Meteorite Impacts meeting held in Brasilia, Brazil.
What I have to do right now is focus on my research and think about conferences once I have something new to present, and the way the summer and fall 2019 terms are looking I may have three different research topics to work on!