In atusy/qntmap: Enhance quantitative analysis of EPMA map

wd <- tempdir()
knitr::opts_chunk$set(
  root.dir = wd,
  fig.width = 5,
  fig.height = 5,
  collapse = TRUE,
  comment = "#>"
)
library(qntmap)
library(dplyr)
library(pipeR)
library(ggplot2)
library(stringr)
library(purrr)

formals(read_xmap)$renew <-
  formals(read_qnt)$renew <- 
  TRUE

Prerequisite

Install qntmap package on R and prepare EPMA data following "Read me".

Preparation

Load package

library(qntmap)

Go to working directory

wd <- tempdir() # Arbitrary
setwd(wd)

Copy example data

file.copy(
  from = system.file('extdata', 'minimal', package = 'qntmap'),
  to = wd,
  recursive = TRUE
)

TRUE indicates copying is successful. Check copied files by dir().

dir(file.path(wd, 'minimal'), recursive = TRUE, all.files = TRUE)

Analysis

Specify directories containing data

dir_map <- file.path(wd, 'minimal/.map/1')
dir_qnt <- file.path(wd, 'minimal/.qnt')

dir_map <- 'minimal/.map/1'
dir_qnt <- 'minimal/.qnt'

Load data

# Read X-ray mapping data
xmap <- read_xmap(dir_map)

# Read spot analysis data
qnt <- read_qnt(dir_qnt, renew = TRUE)

The example data contains olivine and quartz in the mapping area. The phases are quantified 20 points each.

epma <- tidy_epma(qnt, xmap) %>>%
  filter(elm == elm[[1]]) %>>%
  select(x_px, y_px, phase)
plot(xmap, 'Si', interactive = FALSE, colors = "gray") +
  geom_point(
    aes(y_px, x_px, colour = phase), data = epma, inherit.aes = FALSE
  ) +
  guides(
    fill = guide_colourbar(barheight = grid::unit(1, "npc") - unit(10, "line"))
  )

Plot X-ray maps

Plots are interactive with Web UI. Elements can be chosen by mouse actions.

plot(xmap)
# A following plot is non-interactive mode run by
# plot(xmap, 'Mg', interactive = FALSE)

plot(xmap, 'Mg', interactive = FALSE)

Cluster analysis

Initialize cluster centers

This step guesses initial cluster centers (i.e., mean mapping intensities of each phase). Check if results are adequate by comparing them with plots. If inadequate, consider modifying values

centers <- find_centers(xmap, qnt)
print(centers)

Run cluster analysis {#run-cls}

A result is saved as a binary file and PNG images in clustering directory below the directory containing mapping data.

For a quick look, plot(){.r} function is also available.

cluster <- cluster_xmap(xmap, centers)
plot(cluster, interactive = FALSE)

Summarize cluster analysis

summary() function gives abundance ratios of phases in the map.

summary(cluster)

Quantify

qmap <- quantify(xmap = xmap, qnt = qnt, cluster = cluster)

Plot quantified map

Just like X-ray map.

plot(qmap, 'SiO2', interactive = FALSE)

Summarize quantified map

summary(qmap)

Note also that this summary does not correct densities of phases.

V <- unclass(table(cluster$cluster))
V <- V / sum(V)
E <- mean(qmap)
E <- setNames(round(E[['Whole area']], 2), E[['Element']])

According to compositions of olivine (57.29% MgO and 42.71 wt% SiO~2~), and of quartz (100% SiO~2~),

$$ E(MgO^{\mathrm{bulk}}) = 57.29 \times r V['Ol'] + 0 \times r V['Qtz'] = r 57.29 * V['Ol'] \simeq r E['MgO'] \mathrm{,} $$

and

$$ E(SiO^{\mathrm{bulk}}_{2}) = 42.71 \times r V['Ol'] + 100 \times r V['Qtz'] = r 42.71 * V['Ol'] + 100 * V['Qtz'] \simeq r E['SiO2'] \mathrm{,} $$

which are concordant to the above summary.

Summary based on mask images

A simple use of mean() function returns mean() values of each elements.

mean(qmap)

Further, by using PNG format mask image width and height are same as mapping data, mean() values of each element are calculated for each area with same colors in the mask image.

Let's use an image from cluster analysis above.

Input path to the image to segment function.

img <- dir(
    file.path(dir_map, "clustering"), pattern = "_map.png", full.names = TRUE
  )[[1]]
i <- segment(img)

i <- segment(dir("`r img`"))

Then, input the result of segment() to index parameter of mean().

mean(qmap, index = i)

Alternatively, index parameter can be given as a character vector such as name of phases (i.e. . Thus, giving index = cluster$cluster returns a prettier result than the above.

mean(qmap, index = cluster$cluster)

The above two mean() values are equivalent as black (#000000) corresponds to olivine cluster, and white (#FFFFFF) to quartz cluster.

This feature is useful to find average compositions of each minearls, local bulk compositions of domains (layer, symplectite), and so on.

unlink(c('centers0.csv', 'phase_list0.csv'))

atusy/qntmap documentation built on April 11, 2021, 4:45 p.m.