How about this:

convert input.png -colorspace rgb -fx 'if(u!=0 && u!=1, rgb(255,0,0), u)' output.png

This command preserves the black and white parts of the image and turns anything else red. This way, the red pixels can be seen in context.

Note: A solution using "global" operations like -negate, -opaque, etc. would be MUCH faster than applying a formula to every pixel. This command pegged my CPUs for about 10 minutes, but it worked!