splitBone: Bone segmentation

In AlessioVeneziano/IndianaBones: Analysis of bone 3D morphology

Bone segmentation

Description

Performs the digital separation of cancellous and compact bone in a micro-CT image stack.

Usage

splitBone(
  Stack,
  strel,
  iterMask = 3,
  iterFill = 3,
  iterCompact = 3,
  cleanVoid = NULL
)

Arguments

Stack	binary micro-CT image stack of bone specimen (in 3D array form)
strel	KxK structuring element (with odd K) for dilation/erosion operators
iterMask	number of dilation/erosion iterations for generating the bone mask
iterFill	number of dilation/erosion iterations for filling the inside of the bone (cancellous and empty spaces)
iterCompact	number of dilation/erosion iterations for refining the compact bone
cleanVoid	maximum number of 3D pixels (voxels) considered as isolated islands. This argument is used to clean the image stack during the processing (see Details).

Details

splitBone uses an iterative process to separate cancellous and compact bone in a micro-CT image stack. The protocol consists of five sequential operations alternating dilation/erosion (using dilate and erode from the package 'EBImage') to subtractions between images. Dilations and erosions are applied along the Z direciton of the image stack (along the third dimension of Stack). The structuring element (strel) is the binarized kernel (of odd dimensions) used for 'probing' the shapes in the input image. The arguments iterMask, iterFill and iterCompact specify the number of dilation/erosion iterations to be performed in each step of the protocol. The function splitBone implements the following workflow: Step 1: the white pixels of the binary image stack undergo dilations (iterMask) that fill the empty spaces (voids) within the bone; the same amount of erosions (iterMask) shrinks the bone back to its original size and external contours. The result is a mask of the whole bone volume. Step 2: the mask is substracted from the binary image, forming a new stack with only empty spaces preserved. Step 3: multiple dilations (iterFill) of the voids fill the spaces occupied by the cancellous bone and erosions (iterFill) restore its contours. The result is a stack preserving the internal region of the bone (cancellous + voids). Step 4: the internal region is then subtracted from the mask, thus isolating the compact bone. The compact bone is refined using dilation/erosion (iterCompact). Step 5: the cancellous bone is finally obtained by subtracting the voids and the compact bone from the mask. The argument cleanVoid defines the type of cleaning applied to the image stack at Step 3. Cleaning may be useful to remove small pixel leftovers of incomplete dilation/erosion. The function cleanStack is used to identify and delete isolated regions of 3D pixels (voxels). When cleanVoid has value '0', all but the largest pixel island are removed; when cleanVoid is NULL, no cleaning is performed; when cleanVoid > 0, islands consisting of a number of 3D pixels lower or equal to cleanVoid are deleted.

Value

mask	an array of the bone mask of the same dimension as Stack.
void	an array of the empty spaces within the bone (voids) of the same dimension as Stack.
fill	an array of the internal region of the bone (cancellous + voids) of the same dimension as Stack.
compact	an array of the compact bone of the same dimension as Stack.
trab	an array of the cancellous bone of the same dimension as Stack.

Author(s)

Alessio Veneziano

References

Veneziano A, Cazenave M, Alfieri F, Panetta D, Marchi D. 2021. Novel strategies for the characterization of cancellous bone morphology: Virtual isolation and analysis. American Journal of Physical Anthropology. Urbach ER, Wilkinson MH. 2007. Efficient 2-D grayscale morphological transformations with arbitrary flat structuring elements. IEEE Transactions on image processing, 17(1), 1-8.

Examples

#Segment cancellous and compact bone from 3D image stack
data(exampleStack)
require(EBImage)
strel<-makeBrush(5,"disc")
seg<-splitBone(exampleStack,strel,3,2,3,0)
  image(exampleStack[,,30],col=grey(0:32/32))
  image(seg$trab[,,30],col=grey(0:32/32))
  image(seg$comp[,,30],col=grey(0:32/32))

AlessioVeneziano/IndianaBones documentation built on Dec. 2, 2022, 7:35 a.m.