Week 7: Image Processing and Tile Scanning

Week 7: Image Processing and Tile Scanning

Welcome back to my Senior Project blog! As a quick reminder, my research focuses on optimizing the production of Multilamellar Vesicles. These are microscopic bubbles that can deliver drugs and cosmetics through the skin and potentially replace painful injections. Over the past six weeks, I completed all my wet lab work, which consisted of testing the effect of different salt solutions on the formation and yield of these vesicles. 

This week marks the next phase of my project. I am officially moving away from the lab to my computer to begin Phase 3 of my methodology. Now that I have captured hundreds of fluorescence microscopy images, I need to digitally process them to quantify my data.

My primary task this week was creating tile scans. When looking through a fluorescence microscope, you only see a tiny fraction of the sample at one time. To get a complete picture of the overall vesicle yield, I used a software program called ImageJ to stitch nine separate images together into one giant map. This gives me a comprehensive view of the entire lipid film and ensures I am not biased towards the most populated spots to analyze.

After I made some of the tile scans, I had to prep them for actual counting. Computers are not great at analyzing glowing colored circles on a dark background, so I converted every stitched image into an 8-bit grayscale format. I then applied a standardized brightness and contrast filter to the images. This step removes all the background noise and turns the vesicles into quantifiable shapes, making them incredibly easy for the software to identify accurately. 

By standardizing this digital process, I am removing any personal bias from the results. Tomorrow, my goal is to finish applying these grayscale filters to all my control group images. Once every single file is perfectly prepped, I will dedicate all of next week to actually counting the vesicles and running the final statistical analysis. I might also try to train a YOLO (image detection model) to count for me. Looking forward to sharing my final results. 

Thank you for reading!

• Samahith