NA cannot be in an image

[muschellij2]

If you put an NA into an image, antsImage classes automatically turn this into a NaN output. I don't know if it is possible to put NA into the image, but this seems like an issue we should at least have a warning about. This is an issue if people use is.na. Should not really affect new stats as na.rm for @jeffduda really applies to NaN:

library(ANTsRCore)
#> 
#> Attaching package: 'ANTsRCore'
#> The following objects are masked from 'package:stats':
#> 
#>     sd, var
#> The following objects are masked from 'package:base':
#> 
#>     all, any, apply, max, min, prod, range, sum
img <- makeImage(c(10, 100, 3), rpois(n = 1000*3, lambda = 5)) 
img[1,1,1] = NA_real_
img[1,1,2] = Inf
img[1,1,3] = NaN
img[1,2,3] = -Inf
as.array(img)[1,1,]
#> [1] NaN Inf NaN

^{Created on 2019-06-19 by the reprex package (v0.3.0)}

[muschellij2]

Similarly, any comparison procedure should likely give an NaN for the value if the value is NaN. Or at least it should likely be 0, not 1:

library(ANTsRCore)
#> 
#> Attaching package: 'ANTsRCore'
#> The following objects are masked from 'package:stats':
#> 
#>     sd, var
#> The following objects are masked from 'package:base':
#> 
#>     all, any, apply, max, min, prod, range, sum
img <- makeImage(c(10, 100, 3), rpois(n = 1000*3, lambda = 5)) 
img[1,1,1] = NA_real_
img[1,1,2] = Inf
img[1,1,3] = NaN
img[1,2,3] = -Inf
mask =img > 0
arr = as.array(mask)
sum(is.nan(arr))
#> [1] 0
arr[1,1,1:3]
#> [1] 1 1 1
arr[1,2,3]
#> [1] 1

^{Created on 2019-06-19 by the reprex package (v0.3.0)}

[muschellij2]

Do you have any thoughts on NA @jeffduda? Should NaN > 0 produce NaN at least?