Aaron Uy - Week 2 and 3

Week 2 and 3

My second week largely involved further familiarizing myself with the lab, its members, and the cellular analytic work I was tasked with. I continued my work with the image processing of brain slices to highlight the contrast between cell bodies and the “background noise” (while maintaining the original pixel data of image). This was in hopes that we could use some sort of automated way to count all the cells. Although quantifying the relative intensity of the pixels of cells was a somewhat viable method to “count” the cells, the ultimate goal was to actually count the cells.

I took a stab at counting the cells in one brain slice – 1181 red cells and 381 green cells – definitely not fun and this only motivated me further to find an alternative method. After some research, I then turned my attention towards a machine-learning program called Trainable Weka Segmentation. By “training” the program to identify cells, specifically where they start/end, the program could possibly segment cells from the background as well as overlapping cells automatically. Afterwards, the new image could be run under a different program to count the cells automatically. At first, I tried to run the program on my laptop, only to realize that 1. It either crashes my laptop or 2. Each training session takes upwards of 2 hours (and you need several training sessions to “teach” the program what a cell is). Thankfully, I was able to use one of the analytic computers in the lab, which dropped the session time to under 10 minutes. This allowed me to play around with the parameters and determine the best way to “teach” the program to segment the overlapping cell bodies.

Although by the end of the week,  I was pretty successful at segmenting the cell bodies in each slice, the resulting image was still not good enough to automatically and correctly count more than 80% of the cells. L

Meanwhile, my PI was in contact with another PI who conducts similar research at the National Institute of Health (NIH). In discussing ways to quantify the cells in the brain slices, they decided the quickest way would be to send the brain slices to their lab, and have them do the imaging and processing. Afterwards, they would send us back the data and images, teach us how to replicate it, and it would be up to us to familiarize ourselves with their proven method. Their method involves “processing the slices, reconstructing a 3D stack, counting the cells, aligning to the allen atlas and spitting out cell numbers by layer or within a layer”. By creating a 3D reconstruction, they can align it to an atlas, or an anatomic reference map of the whole mouse brain. This way, the cells can be binned into regions of interest in a 3D plane and counted appropriately.

Week 3 consisted of waiting for the slices to be shipped out to the other PI and receiving the processed slice data. In the meantime, I continued messing around with the machine learning program and also shadowed my Postdoc in the afternoons. Given some freetime, I was also able to create 3D reconstructions of the mouse brain using images consecutive stained brain slices. I used this 3D image along with the 3D segmentation component of the program. This was somewhat unsuccessful as it yielded similar results to the original version.

In the afternoons, I shadowed my postdoc. I learned how to perfuse a mouse, which involves anaesthetizing a mouse, injecting a solution into it heart, pumping this solution throughout its circulatory system, and dissecting out the brain. This process fixes and preserves the brain so that it can be sliced and later studied. Among other things, I assisted in his mouse behavior tests, and observed my postdoc prepare brain tissue so that individual cell counts can be obtained from a Flow Cytometry machine.

Although I was somewhat unsuccessful, my experience in the lab in the past two weeks has been very fulfilling. The imaging knowledge I learned should prove to be useful in my future at the lab. I look forward to my next few weeks!