When we were young, curiosity was our second nature. The world was not some mundane environment, but an exciting adventure. Slowly as we grew up, this sense of curiosity and wonder started to fade. Education became an automatic act, a way to add on more facts to the overstuffed cabinets of our minds.
I am in my mid-30’s now. I’ve completed my training to be a neurologist and instead of working hard to become a consultant, I chose to spend a few years pursuing a PhD. One of the reasons for going down this path is to discover — or perhaps re-discover — the joys of learning. Alhamdulillah, so far I feel like I’m doing well on that track.
You see, the good thing about being a student at this age is that you’ve done it before. That’s why, when you come across something challenging, or maybe some concept that you’ve never heard about before, you (hopefully) won’t panic too much. Just chill and learn, chill and learn. After all, nobody can ever hope to know every single thing, so why should we be afraid or frustrated when we encounter something new?
Just this week, I gave a brief presentation during my research group’s lab meeting. I talked about one of the things that’s been bugging me as someone studying cerebral small vessel disease — the question of how exactly to diagnose the condition. One can write at length about this topic, but suffice it is to say that the diagnosis of cerebral small vessel disease is still somewhat subjective, and nowhere near as straightforward as it should be ideally. While researching the topic, I came across this paper by Sundaresan et al. entitled:
Automated lesion segmentation with BIANCA: Impact of population-level features, classification algorithm and locally adaptive thresholding
https://doi.org/10.1016/j.neuroimage.2019.116056
You may wonder why I’m talking about this paper. Well, the reason is because I stumbled upon something new while reading the paper that reminded me of the joys of learning.
Those of you who are mathematicians or computer scientists may be laughing at me now, but I genuinely did not know of the existence of Voronoi diagrams before.
And yet, they are an elegant solution to many problems that occur around us, for example, if you had 10 post offices in a given area, how do you determine the optimal coverage area for each branch?
Similarly, when doing automated segmentation of lesions in the brain (in this case, attempting to segment and quantify white matter hyperintensities), one way to improve the technique is by using Voronoi tesselation to produce what are effectively Voronoi polygons in the brain. This allows you to apply local thresholding to generate a better binary mask of white matter hyperintensities.
Simple, right? Ha ha ha…
Anyway, despite struggling with the concept initially, I came away with something rare, a feeling that ‘Hey, I learnt something completely new today!’
Bonus: check out this Medium article on making ‘Artistic Voronoi Diagrams in Python‘ by Frank Ceballos. The colours are gorgeous!
