Brilliant application of machine learning and computer vision to point-of-care diagnostics! Way to go @dpcbod & @tanay_tandon !
Report
How is machine learning used in this when it can do an imaging-based approximate count of the individual WBCs in the bloodstream?
How far down can its resolution go down to? (can it have the resolution to even track individual proteins?)
@inquilinekea we can resolve particles down to a size of 5 microns with our imager. in order to classify out particulate matter, reticulocytes, and floating lymph (often appearing similar to WBCs) we trained a bunch of classifiers. Furthermore for 5-part differentials, the vision needs to distinguish between a bunch of nucleation patterns, a task requiring a well trained machine learning classifier.
Report
@tanay_tandon Particulate matter?!? So... can this literally measure the increase of particulate matter in people following exposure to air pollution? [not to mention inflammation] There really needs to be a faster feedback loop between air-pollution exposure and "bad thing happening", and maybe this imaging device can finally provide it to nudge people towards avoiding air pollution (and preserving their brains) more?
http://www.sciencemag.org/news/2...
I'm also fascinated if this can provide the fast-feedback loops that can measure increases in inflammation that follow consumption of unhealthy food (e.g. high glycemic index food, or heavily fried foods).
@inquilinekea by particulate matter we mean any particle in the blood smear image that's not a viable cell (lysed cell, dust from the strip or finger, etc.) interesting Science Mag article! One of the potential applications could definitely be recording such particles (assuming our model is trained to identify) and counting in an appropriate sample. Will read more into
Replies
Athelas
Athelas
Hone