R/segmentShapeFromImage.R

In stnava/ANTsR: ANTs in R: Quantification Tools for Biomedical Images

#' convolution-based shape identification
#'
#' takes an image and a shape and relates the 2nd with the first to yield a
#' feature image - the function is user modifiable but defaults to cor
#'
#' @param img antsImage
#' @param shape to define features
#' @param mask with values 1 or 0
#' @param rad max radius ( optional, not recommended )
#' @param scfun function to apply to create feature image (defaults to cor)
#' @param maskZeroes if TRUE (default), zeroes will not influence the result
#' @return feature image
#' @author Brian B. Avants
#' @keywords shape template
#' @examples
#'
#' set.seed(123)
#' fi <- makeImage(c(20, 20), rnorm(400, mean = 1, sd = 0.1))
#' mask <- getMask(fi, 0.8, Inf, 0)
#' segs <- kmeansSegmentation(fi, 3, mask)$segmentation
#' segs[segs != 1] <- 0
#' shp <- labelClusters(segs, 1)
#' shp[shp != 1] <- 0
#' fimg <- segmentShapeFromImage(fi, shp, mask)
#'
#' @export
segmentShapeFromImage <- function(
    img, shape, mask = NULL, rad = NA, scfun,
    maskZeroes = TRUE) {
  if (nargs() == 0) {
    print(args(segmentShapeFromImage))
    return(1)
  }
  dim <- img@dimension
  if (is.null(mask)) {
    mask <- getMask(img)
  } else {
    mask <- check_ants(mask)
  }
  shapemask <- getMask(shape)
  shape <- cropImage(shape, shapemask)
  shapemask <- cropImage(shapemask, shapemask)
  if (all(is.na(rad))) {
    rad <- round(((dim(shape) - 1) / 2))
  }
  eps <- 1.e-12 # this is an odd bug : round(dim(img)/2) gives wrong result
  temp <- iMath(as.antsImage(as.array(shape)), "PadImage", 1)
  shapevec <- getNeighborhoodAtVoxel(temp, round(dim(temp) / 2 + eps), rad)$values
  temp <- iMath(as.antsImage(as.array(shapemask)), "PadImage", 1)
  shpmskvec <- getNeighborhoodAtVoxel(temp, round(dim(temp) / 2 + eps), rad)$values
  selector <- (shpmskvec > eps)
  mat <- getNeighborhoodInMask(img, mask, rad, boundary.condition = "image")
  mat <- antsrimpute(mat)
  if (maskZeroes) {
    shapevec <- shapevec[selector]
    mat <- mat[selector, ]
  }
  if (!missing(scfun)) {
    shapecor <- rep(0.0, ncol(mat))
    for (kk in 1:ncol(mat))
    {
      shapecor[kk] <- scfun(mat[, kk], shapevec)
    }
  } else {
    shapecor <- cor(shapevec, mat)
  }
  # convolve(h2$xt,x2$xt,type="open") # could not get this to work
  shapecor <- antsrimpute(as.numeric(shapecor))
  featurei <- makeImage(mask, shapecor)
  return(featurei)
}