Simulate background cells

Simulate cell locations. The 2D locations of the cells are simulated and plotted in a rectangular window. Users can specify the window size, cell number and the minimum distance between two cells. All cells have the same cell type, specified by the "Cell.Type" param.

simulate_background_cells(
  n_cells,
  width,
  height,
  method = "Hardcore",
  min_d,
  oversampling_rate = 1.2,
  jitter = 0.3,
  Cell.Type = "Others",
  plot_image = TRUE
)

Arguments

n_cells

Numeric. Number of cells to simulate in the background.

width, height

Numeric. The width and height of the image.

method

String. The distribution model for the background cells. Options are "Hardcore" for tumour tissues and "Even" for normal tissues. Default is "Hardcore".

min_d

(OPTIONAL) Numeric. Use when method is "Hardcore". The minimum distance between two cells.

oversampling_rate

(OPTIONAL) Numeric. Use when method is "Hardcore". The multiplier for oversampling. Without oversampling, the simulation deletes cells that are within min_d from each other, resulting in a less total number of cells than n_cells. Default is 1.2 (this should be set based on n_cells and min_d; should always be larger than 1).

jitter

(OPTIONAL) Numeric. Use when method is "Even". The uniform distribution parameter to generate the jitter distance for each cell from the vertices of the hexagon.

Cell.Type

(OPTIONAL) String. The name of the background cell type. Default is "Others" since there shouldn't be any identity of the background cells.

plot_image

(OPTIONAL) Boolean. Default is TRUE.

Value

A data.frame of the simulated background image

Details

There are two options for the background cell distribution model. 1) Hardcore model for tumour tissues. This model uses rHardcore spatstat.random. Since rHardcore deletes cells that are within the distance of min_d to another cell, resulting in fewer cell specified by users, we here introduce parameter oversampling_rate to generate more cells than specified. 2) Normal tissues use an evenly-spaced model where the cells are distributed approximately according to the vertices of a hexagon. The function accomplishes this by generating cells on a hexagonal grid and individually applying a bounded uniform random jitter. In our algorithm, jitter is the parameter to define the uniform distribution of the jitter of the cells from the hexagon vertices. If there is a reference image, jitter can be estimated by comparing the average minimum distance between cells of the simulated image and the reference image. If without a reference image, We suggest 0.3 as the default value of jitter as this gives a sensible outcome.

See also

simulate_mixing for mixed background simulation, simulate_clusters for cluster simulation, simulate_immune_rings/simulate_double_rings for immune ring simulation, and simulate_stripes for vessel simulation.

Other simulate pattern functions: simulate_clusters(), simulate_double_rings(), simulate_immune_rings(), simulate_mixing(), simulate_stripes()

Examples

set.seed(610) # set seed for this background image simulation for reproducibility
background_image <- simulate_background_cells(n_cells = 5000, width = 2000,
                                              height = 2000, method = "Hardcore",
                                              min_d = 10,
                                              oversampling_rate = 1.5,
                                              Cell.Type = "Others",
                                              plot_image = TRUE)