1. Find and Quantitate Bands
IQ-11 uses two simple parameters to detect bands: band thickness – top to bottom in a lane -- and band intensity – darkness. Many samples have bands with a wide range of thicknesses; using a “narrow band” value generally finds all of those, but often finds two bands on big dark objects. That’s what happened in the left lane above. But your training set always has one big band there, so IQ-11 will learn that, and create the one band like the right lane above. Your training set always has two bands on the shoulder below the big band. IQ-11 generally finds the lower of those two, but often not the upper one; again, it will learn to find that upper one.
Sections 4 and 5 below has a more in-depth presentation of IQ-11’s detection and quantitation algorithms.
IQ-11 uses 15 vertical cross-sections, or profiles, and 3 cross-sections on either end, to create a 36-sided outline of each band. The background-corrected darkness of each pixel inside the outline is summed up to yield the band’s quantitative volume. IQ-11 looks for inflection points in the profiles to determine the outline point. Frequently, our algorithm will result in an outline which doesn’t align with what your operators want, like the left and right bands above. So again, our AI algorithms will learn that, on a band-by-band basis.
Now we’ll back up to show the entire process from the beginning.