Histology pattern recognition -- zoom out!

Back to News

Experienced pathologists are very good at using minimum magnification to look at a slide. A pathologist can see more scanning an entire slide at 5x and then occasionally jumping to 20x or 40x in a couple of areas, than starting out at 40x. We all know this, and take it for granted as pathologists. However, when we do image analysis, we seem to immediately want to run analysis at 40x or the highest magnification possible. BAD IDEA. Why? Two reasons:

1) The time and computing cycles required to run at higher magnifications goes up with the square of the magnification. So running at 40x will be 4 times longer than at 20x. Running at 20x will be 16 times longer than running at 5x and so on. Waste of your time, waste of computer time.

2) You are distracting the computer with needless detail. This will come back to bite you, trust me.

Ask yourself when you are planning your analysis:  if I was just making these measurements myself with a plain old microscope, what magnification would I use? I bet if you wanted to look at necrotic areas of a xenograft, you would be fine at 2x, and if you wanted to look at pancreatic islets, 8x would be fine. Now, if you want to count vessels, yes, you would be using a 20x magnification, and occasionally switching to 40x.  If you run Genie or another pattern recognition tool at 20x in xenografts, you will be feeding the computer way too much information, it just needs to see the minimum detail differentiating necrotic from normal, nothing more. Pattern recognition is easily distracted, avoid these distractions by running at the lowest magnification you can.

Hope this helps, it helped me.

Share this Post: