Regulatory Landscape Exercise Solution

Summary

The algorithm would use digital pathology data and would be a segmentation algorithm used to segment and count red vs. white blood cells from a blood smear.

This would have to occur in the workflow after digitization of the images, but before the pathologist got to the image. I would use this to help pathologists quickly know the exact ratio of red: white blood cells. If this value is high, it could indicate leukemia. Since the algorithm would give an exact ratio, rather than a pathologist’s estimate from just looking at the smear without counting it, this could be useful as early detection of low white blood cells, before the count gets so low that it is visually obvious.

I am going to guess that this is either Class I or Class II, because I don’t think that something like this would be high risk. If it is wrong, the pathologist is still always going to be looking at the image anyways, so they will probably be able to correct any incorrect output that my algorithm gives in their report.