Identifies significant feature cells using Grad-CAM data either with visualization (gating mode) or as a lightweight operation using pre-converted images (base mode).
Usage
inverseGradCAM(
x = NULL,
results = NULL,
feature_matrix,
mode = c("gating", "base"),
percentile = 0.9,
n_resolution = 100,
transpose = TRUE,
filename = NULL,
ncol = 2,
nrow = 2
)Arguments
- x
TockyPrepData object (required for "gating" mode)
- results
convert_to_image output (required for "base" mode)
- feature_matrix
Feature intensity matrix from Grad-CAM analysis
- mode
Operation mode ("gating" for visualization, "base" for lightweight analysis)
- percentile
Significance threshold percentile (0-1). If NULL, grad-CAM values will be returned, instead of feature cell designation.
- n_resolution
Binning resolution (for "gating" mode only)
- transpose
Logical Whether to tranpose feature_matrix input. Note that TockyConvNetPy output Grad-CAM matrix for feature_matrix typically needs to be transposed. The default is TRUE.
- filename
Optional PDF output path (for "gating" mode only)
- ncol
The number of the columns for the output plot
- nrow
The number of the rows for the output plot
Value
A binary numeric vector (1/0) for feature and other cells. If `percentile = NULL`, grad-CAM values are returned as a numeric vector.
Examples
if (FALSE) { # \dontrun{
# Gating mode with visualization
out <- inverseGradCAM(mode = "gating", x = prep_data,
feature_matrix = cam_matrix, filename = "output.pdf")
# Base mode with pre-converted results
img_data <- convert_to_image(merged_data)
out <- inverseGradCAM(mode = "base", results = img_data,
feature_matrix = cam_matrix)
} # }