findNeighbors-methods: Find spatial neighbors and spatial weightst

In Cardinal: A mass spectrometry imaging toolbox for statistical analysis

Description

Methods for calculating the spatial neighbors (pixels within a certain distance) or spatial weights for all pixels in a dataset.

Usage

#### Methods for Cardinal >= 2.x classes ####

## S4 method for signature 'ImagingExperiment'
findNeighbors(x, r, groups = run(x), ...)

## S4 method for signature 'PositionDataFrame'
findNeighbors(x, r, groups = run(x), dist = "chebyshev", 
    offsets = FALSE, matrix = FALSE, ...)

## S4 method for signature 'ImagingExperiment'
spatialWeights(x, r, method = c("gaussian", "adaptive"),
    dist = "chebyshev", matrix = FALSE, BPPARAM = getCardinalBPPARAM(), ...)

## S4 method for signature 'PositionDataFrame'
spatialWeights(x, r, matrix = FALSE, ...)

#### Methods for Cardinal 1.x classes ####

## S4 method for signature 'iSet'
findNeighbors(x, r, groups = x$sample, ...)

## S4 method for signature 'IAnnotatedDataFrame'
findNeighbors(x, r, groups = x$sample, dist = "chebyshev",
    offsets = FALSE, matrix = FALSE, ...)

## S4 method for signature 'iSet'
spatialWeights(x, r, method = c("gaussian", "adaptive"),
    matrix = FALSE, ...)

## S4 method for signature 'IAnnotatedDataFrame'
spatialWeights(x, r, matrix = FALSE, ...)

Arguments

x	An imaging dataset or data frame with spatial dimensions.
r	The spatial radius or distance.
groups	A factor giving which pixels should be treated as spatially-independent. Pixels in the same group are assumed to have a spatial relationship.
dist	The type of distance metric to use. The options are ‘radial’, ‘manhattan’, ‘minkowski’, and ‘chebyshev’ (the default).
offsets	Should the coordinate offsets from the center of each neighborhood be returned?
matrix	Should the result be returned as a sparse matrix instead of a list?
method	The method to use to calculate the spatial weights. The 'gaussian' method refers to Gaussian-distributed distance-based weights (alpha weights), and 'adaptive' refers to structurally-adaptive weights for bilaterla filtering (beta weights).
...	Addtional arguments to be passed to next method.
BPPARAM	An optional instance of BiocParallelParam. See documentation for bplapply.

Value

Either a list of neighbors/weights or a sparse matrix (sparse_mat) giving the neighbors and weights for each pixel.

For spatialWeights, two types of weights are calculated and returned as a list:

The alpha weights are distance-based, following a Gaussian distributed that produces smaller weights for larger distances. The beta weights are adaptive weights used for bilateral filtering, which are based on the difference in the feature-vectors between pixels.

If method="gaussian" only the alpha weights are calcualated and the beta weights are all set to 1. If matrix=TRUE, the alpha and beta weights are multiplied together to produce the weights for the matrix; otherwise, both are returned separately.

Author(s)

Kylie A. Bemis

See Also

image

Examples

coord <- expand.grid(x=1:9, y=1:9)
values <- rnorm(nrow(coord))
pdata <- PositionDataFrame(coord=coord, values=values)

# find spatial neighbors
findNeighbors(pdata, r=1)

# calculate distance-based weights
spatialWeights(pdata, r=1)

# visualize weight matrix
W <- spatialWeights(pdata, r=1, matrix=TRUE)
image(as.matrix(W), col=bw.colors(100))

Cardinal documentation built on Nov. 8, 2020, 11:10 p.m.