getRoots: Convert a matrix of semi-processed DICOM images to root...

In troyhill/coreCT: Programmatic Analysis of Sediment Cores Using Computed Tomography Imaging

Description

Calculates the number of root/rhizome particles, volumes, and surface areas, for different size classes. This version accommodates calibration curves with >4 calibrants, and uses density thresholds converted to Hounsfield Units using the calibration curve (rather than direct calibration rod values) to partition sediment components.

Usage

getRoots(mat.list, pixelA, diameter.classes = c(1, 2, 2.5, 10),
class.names = diameter.classes,
thickness = 0.625,
means     = c(-850.3233, 63.912, 271.7827, 1345.0696),
sds       = c(77.6953, 14.1728, 39.2814, 45.4129), 
densities = c(0.0012, 1, 1.23, 2.2),
pixel.minimum = 4)

Arguments

mat.list	list of DICOM images for a sediment core (values in Hounsfield Units)
pixelA	pixel area (mm2)
diameter.classes	an integer vector of diameter cut points. Units are mm (zero is added in automatically).
class.names	not used presently
thickness	slice thickness for computed tomography image series (mm)
means	mean values (units = Hounsfield Units) for calibration rods used.
sds	standard deviations (units = Hounsfield Units) for calibration rods used. Must be in the same order as means.
densities	numeric vector of known cal rod densities. Must be in the same order as means and sds.
pixel.minimum	minimum number of pixels needed for a clump to be identified as a root

Details

Calculates the number of root/rhizome particles, volumes, and surface areas, for different size classes. This function requires that values be Hounsfield Units (i.e., data must be semi-processed from the raw DICOM imagery).

Value

value getRoots returns a dataframe with one row per CT slice. Values returned are the number, volume (cm3), and surface area (cm2) of particles in each size class with an upper bound defined in diameter.classes.

See Also

convert

Examples

ct.slope <- unique(extractHeader(core_426$hdr, "RescaleSlope"))
ct.int   <- unique(extractHeader(core_426$hdr, "RescaleIntercept")) 
# convert raw units to Hounsfield units
HU_426 <- lapply(core_426$img, function(x) x*ct.slope + ct.int)

rootChars <- getRoots(HU_426, pixelA = 0.0596,
diameter.classes = c(2.5, 10))

## Not run: 
# plot using "ggplot" package after transforming with "reshape2" package
area.long <- reshape2::melt(rootChars, id.vars = c("depth"), 
   measure.vars = grep("Area", names(rootChars)))
ggplot2::ggplot(data = area.long, ggplot2::aes(y = -depth, x = value, 
   color = variable)) + ggplot2::geom_point() + ggplot2::theme_classic() + 
   ggplot2::xlab("root external surface area per slice (cm2)")

## End(Not run)

troyhill/coreCT documentation built on Feb. 6, 2021, 7:16 p.m.