search

ncoudray / DeepPATH

Classification of Lung cancer slide images using deep-learning
492 stars 213 forks source link

Removal of ~empty slides #30

Closed bcli4d closed 5 years ago

bcli4d commented 5 years ago

In the preprint of your paper, you state that "The slides with a low amount of information were removed, that is all the tiles where more than 50% of the surface is covered by background (where all the values are below 220 in the RGB color space). " I don't see this happening in the tiling code (v0b_tileLoop_deepzoom4.py). Did I miss it, or did you determine that culling such images is not important?

Thanks for making this very interesting work available.

ncoudray commented 5 years ago

Hi.

See lines 84 to 87 of the code

Best, Nicolas

bcli4d commented 5 years ago

Nicolas, I interpreted the above quote to say that entire slides are removed if more the 50% of the slide is background. Should that sentence really have said "The tiles with a low amount of information were removed,..." ? Thanks, Bill

bcli4d commented 5 years ago

By the way, I suggest changing line 89 of v0b_tileLoop_deepzoom4.py: if avgBkg <= (self._Bkg / 100): to if avgBkg <= (self._Bkg / 100.0): It appears that the argument parser doesn't coerce the --Background argument to a float when the specified argument value "looks" like an integer, e.g. "50", and even though the argument type is declared to be float. This results in integer division, which produces a value of 0 or 1.

ncoudray commented 5 years ago

Thanks for the feedback. Indeed, only the tiles are removed