Detect Cells
- Shrink the red channel (nearest area interpolation) to get the background gradient of the image
- Dim the image slightly and subtract it from the original red channel to increase contrast
- Threshold at the mode (assuming mostly blank space) and erode and dialate repeatedly to reduce noise
- Draw contours around the cells, and get the convex hull to maximize cell area (avoid missing off center beta gal coloration)
- Remove small "cells" to filter noise
- Fill in the convex hulls to get mask
Detect Beta Gal Expression
- Subtract the red channel from the blue channel to get areas which are bluer
- Get the average subtracted blue value and use it threshold the image
- Mask the image with the detected cells, so only detecting beta gal spots within detected cells
- erode with a small kernel to remove spot noise, and dilate with a larger one to fill in any gaps within beta gal spots
- Find Contours, and filter out any remaining noise contours
Required Parameters
-n --name: Sample name for the output file
Optional Parameters: Cell Counting
-c --CellSDMulti: Multiplier for cell counting threshold - higher values are stricter.
-C --ContrastMulti: Determines the magnitude of background image dimming.
-g --GResMulti: Determines how small the gradient image is - large numbers matches subtle gradients while small ones
will preserve large cells.
-l --loops: Number of times to repeat erosion and dilation for cleanup.
-m --minCell: Minimum cell size (square pixels) to be considered a cell.
Optional Parameters: Beta Gal Counting
-b --BGalSDMulti: Multiplier for beta gal threshold - higher values are stricter.
-M --minBGal: Minimum beta gal size (square pixels) to be considered true beta gal stain.
Optional Parameters: Generic Parameters
-i --InputDir: Directory to look for images. Defaults to current directory.
-t --testing: Flag to display testing images and cell/beta gal distribution histograms.
Python 3.7.5
- scipy v1.4.1
- matplotlib v3.1.3
- numpy v1.18.1
- opencv v3.4.2
- argparse v1.1
