3D Counter segmentation

ImageJ has a built in function called Counter3D. This is used as an alternative to the existing Auto-segmentation.

To use it click on use 3D OC (OC is short for Object Counter). Checking the box will make the grown mask unavailable. The 3D Counter is much faster than the existing auto-segmentation.

The minimum SUV in 3D Counter is 2.5. While you cannot go below this value you can certainly use use SUV in the Brown fat dialog to set the SUV to 4, or to 41%, or anything else.

The next step is to eliminate false positives. There can be a large number of ROIs and manually stepping through them all can be tedious. A helper function has been introduced to reduce tiny ROIs which have less than the minimum number of pixels. The idea is to use use SUV and set the lower SUV to something higher than 2.5. A good choice seems to 4. (Note: do not choose something like 41% as this will give a lower threshold on tiny ROIs, guaranteeing that they pass the number of pixels test. This would mean that you would not eliminate any ROIs.)

After setting the elevated SUV in the low SUV box, switch back to the nifti tab. You will see a button . Pressing this will eliminate tiny ROIs from the list which needs to checked. Switch back to the ROI tab. You may want to lower your SUV or simply remove the check from use SUV.

Then you need to step through the found ROIs and eliminate false positives using and perhaps add false negatives by manually defining additional ROIs. I saw a case where part of a ROI was a tumor but it extended up through the neck to the head where it was not a tumor. To eliminate part of the ROI, I manually added a ROI in the neck and head part and marked the ROI as "exclude ROI head".

Options

As indicated above the minimum SUV is set at 2.5. There is also a minimum number of pixels which need to be included in the ROI which is set at 10. These can be changed in the extended tab of Options.

If you choose to change these values, make sure all members of your group change them as well, or you will face different members of the group getting non identical results for the same input study. (The TMTV values will be close but not identical.)

Continue to: Brown Fat, or to: Help