Make Volume Fraction respect ROIs

[mdoube]

During the BoneJ1-2 port, Volume Fraction stopped respecting ROIs. This is a total pain for many users who want the BV/TV of e.g. the trabeculae in an irregular medullary cavity.

Reinstate ROI respect for BV/TV measurements.

[adavi0913]

I second this! I've been trying to run data on an irregular specimen, and my values are off from published data using similar modalities. It would be great to know when this problem will be fixed. Thanks!!!

[mdoube]

This macro gives a basic idea of what's needed:

nRois = RoiManager.size();
histogram = newArray(256);
bv = 0;
tv = 0;
setBatchMode(true);
for (i = 0; i < nRois; i++){
	RoiManager.select(i);
	getHistogram(values, histogram, 256);
	bv += histogram[255];
	tv += histogram[0] + histogram[255];
}
print("bv = "+bv+" pixels and tv = "+tv+" pixels. BV/TV = "+(bv/tv)+".");

Input is an open image (assumed that foreground (bone) is 255 and background (marrow space) is 0) and an ROI Manager populated with ROIs. Output looks like:

bv = 2969730 pixels and tv = 6240472 pixels. BV/TV = 0.4759.

The macro snippet doesn't do any sanity checking like making sure the ROI coordinates are inside the image, nor does it provide calibrated real-world volumes.

[mdoube]

This macro prototype adds calibration:

nRois = RoiManager.size();
histogram = newArray(256);
bv = 0;
tv = 0;
getVoxelSize(pw, ph, pd, unit);
pv = pw * ph * pd;

setBatchMode(true);
for (i = 0; i < nRois; i++){
	RoiManager.select(i);
	getHistogram(values, histogram, 256);
	bv += histogram[255];
	tv += histogram[0] + histogram[255];
}
print("BV = "+bv+" pixels ("+bv * pv+" "+unit+"³) and TV = "+tv+" pixels ("+tv * pv+" "+unit+"³). BV/TV = "+(bv/tv)+".");

TODO:

• Port to Java (but as modification to existing plugin or as new plugin?)
• Check headless operation, ROI Manager might not be headless-friendly
• Check that ROIs are inside the volume - what is the default behaviour when ROIs are completely or partially outside?
• Check the image is binary. Maybe just making sure that TV (histogram[0] + histogram[255]) = ROI area or ROI pixel count is enough. Quit with an "image is not binary" error if it's not true.
• Handle multiple ROIs per z-slice