Polygon ROI Growing Region Function Removed?...Rationale?

[bioni...@gmail.com]

Dear all, we have noticed that Horos no longer allows users to perform growing region segmentation using Polygon ROIs. It is only currently possible with Brush ROIs. This is problematic for labs such as ours that take a 'semi-automated' approach to HU-united based volumetric segmentation of CT findings. We would use the growing region to first place a polygon ROI which we would then variably adjust using the repulser tool to create our final ROI. The repulser tool unfortunately doesn't work on Brush ROIs. I guess my first question would be....am I missing something? I've been an OsiriX user since 2014 and Horos user since its' inception and can't figure out a way to perform this polygon roi growing region function using the latest Horos version 3.1.2. The second question relates to the rationale for removal and if we should expect this function to return to Horos as either a native function or possibly a plug-in in the near future. I appreciate any and all responses to this and thanks!

[joao santinha]

Hi bionicatjhu,

Just to leave you a workaround!

You can still use the region growing from Horos to generate the brush rois and then you can convert them to Polygon ROIs by going to ROIs menu -> Brush ROIs -> Convert Between Polygon / Brush ROIs

Let me know if this helps you!

Cheers,

Joao

On Mon, Jul 23, 2018 at 3:19 PM, <bioni...@gmail.com> wrote:

Dear all, we have noticed that Horos no longer allows users to perform growing region segmentation using Polygon ROIs. It is only currently possible with Brush ROIs. This is problematic for labs such as ours that take a 'semi-automated' approach to HU-united based volumetric segmentation of CT findings. We would use the growing region to first place a polygon ROI which we would then variably adjust using the repulser tool to create our final ROI. The repulser tool unfortunately doesn't work on Brush ROIs. I guess my first question would be....am I missing something? I've been an OsiriX user since 2014 and Horos user since its' inception and can't figure out a way to perform this polygon roi growing region function using the latest Horos version 3.1.2. The second question relates to the rationale for removal and if we should expect this function to return to Horos as either a native function or possibly a plug-in in the near future. I appreciate any and all responses to this and thanks!

--
You received this message because you are subscribed to the Google Groups "Horos Project" group.
To unsubscribe from this group and stop receiving emails from it, send an email to horos-project+unsubscribe@googlegroups.com.
To post to this group, send email to horos-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/horos-project/31d3717b-85f3-4d33-85bc-56ad34e8850b%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[bioni...@gmail.com]

Hey Joao, 

     I appreciate the prompt response, it was helpful. Yes, I am very familiar with that function, as we've attempted to utilize this pathway in the past (perform ROI in brush then convert to polygon then adjust) but have found that the resultant polygon ROI isn't as well-defined or precise as that generated by the growing region given the same thresholding algorithm. A less precise ROI means more manual repulsion work. In general, the addition of another step in our workflow with the need to covert, the yield of less precise ROIs, and the yield of post-conversion ROIs that differ in volume from the original brush ROI make it an inconvenience to our process. I just wanted to know if this function, present on the legacy Horos and OsiriX versions, would ever return to Horos, and possibly why it was removed at all. Thanks

On Monday, July 23, 2018 at 10:30:52 AM UTC-4, joao santinha wrote:

Hi bionicatjhu,

Just to leave you a workaround!

You can still use the region growing from Horos to generate the brush rois and then you can convert them to Polygon ROIs by going to ROIs menu -> Brush ROIs -> Convert Between Polygon / Brush ROIs

Let me know if this helps you!

Cheers,

Joao

On Mon, Jul 23, 2018 at 3:19 PM, <bioni...@gmail.com> wrote:

Dear all, we have noticed that Horos no longer allows users to perform growing region segmentation using Polygon ROIs. It is only currently possible with Brush ROIs. This is problematic for labs such as ours that take a 'semi-automated' approach to HU-united based volumetric segmentation of CT findings. We would use the growing region to first place a polygon ROI which we would then variably adjust using the repulser tool to create our final ROI. The repulser tool unfortunately doesn't work on Brush ROIs. I guess my first question would be....am I missing something? I've been an OsiriX user since 2014 and Horos user since its' inception and can't figure out a way to perform this polygon roi growing region function using the latest Horos version 3.1.2. The second question relates to the rationale for removal and if we should expect this function to return to Horos as either a native function or possibly a plug-in in the near future. I appreciate any and all responses to this and thanks!

--
You received this message because you are subscribed to the Google Groups "Horos Project" group.

To unsubscribe from this group and stop receiving emails from it, send an email to horos-projec...@googlegroups.com.

[joao santinha]

You're welcome.

I think I was one of the last to work with that code (added the possibility to generate the 3D preview instead of only 2D) so I will try to check it but I don't know when I will have an answer. If nobody has additional input, I will let you know when I have something.

Joao

To unsubscribe from this group and stop receiving emails from it, send an email to horos-project+unsubscribe@googlegroups.com.

To post to this group, send email to horos-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/horos-project/270a71a5-7418-4b14-b046-c1e4c54295f4%40googlegroups.com.

[bioni...@gmail.com]

Thanks Joao, I'll be sure to check in and monitor for any response of yours. For some help, I've attached 2 screenshots of ROIs taken with the same HU thresholding algorithm/growing region that yield different areas & ROI segmentation using the Polygon ROI and Brush-to-Polygon Convert ROI. I've also attached a pdf of a publication that our team has previously published including some of the methodology mentioned above. Here is a link to the same paper: 

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5154627/

Let me know if I can be of any other help and updates are always welcomed...thanks!

[Soren Christensen]

Just out of interest - do you have multiple disjoint ROIs for any of your cases?

I believe the built in volume calculation in these cases is far from ideal. Just mentioning it in case you get bitten by it.

Soren

To unsubscribe from this group and stop receiving emails from it, send an email to horos-project+unsubscribe@googlegroups.com.

To post to this group, send email to horos-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/horos-project/7d2923cf-406c-4435-915a-872f9a21072b%40googlegroups.com.

[bioni...@gmail.com]

Hey Soren and yes, the presence of disjoint ROI (if I understand disjointed as requiring multiple ROIs on the same image to adequately segment an area of interest) is sometimes a part of our methodology. It's also another pitfall of the Brush to Polygon ROI conversion mechanism, as noncontiguous brush ROIs on the same image get reduced to a single polygon ROI that omits the smaller of the discontiguous areas in its' coverage. Am I understanding your conception of disjoint correctly here and thanks

[Soren Christensen]

Hi,

 Sorry,I was not very clear. What I meant was situations like this:

Eg, patient has two hemorrhages, one isolated to the top op the brain and one on the base of the brain. Say they are 8 cm apart and each one is 2mL. You'd expect the software to tell you there's 4mL of hemorrhage in total. However, the default volume calculation - and i just checked the newest version - will give you the volume of the "cylinder" that connects these two objects. This is not always what people want and I am not sure why it was written this way. Normally you'd just multiply voxelvolume with voxel count. 

Example below: I drew 2 ROIs, on on top slice, one on bottom. The real volume of these regions added up is 3.6 mL.  If you run the default volume calculation, I will get the volume for the object below (from which I only made the top and bottom ROI, rest are made by the application).

There may be situations where you want to interpolate this way, but I'd imagine in most situations this is not what the user assumes is going to happen.

I made a plugin to calculate it as the voxelvolume X voxelcount product. Happy to share and also happy to hear if there's a way to disable the interpolation taking place in the volume calculation.

This is not a Horos thing - has been this way in Osirix for years.

Soren

To unsubscribe from this group and stop receiving emails from it, send an email to horos-project+unsubscribe@googlegroups.com.

To post to this group, send email to horos-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/horos-project/129c0264-861f-4bfd-ab4a-08ce9dc63a79%40googlegroups.com.

[tbriz...@purview.net]

Hi,

Thanks for all the feedback. Just resurrected polygon ROIs on segmentation here... they got broken by changes in ITK/VTK.

This fix should be included in next Horos release!

All the best,

Tiago

[bioni...@gmail.com]

Hey Soren, I would love to take a look at your plugin, do you mind sharing it? In these instances, we have simply been assigning a different ROI name to discrete regions of hyperintensity (i.e. hemorrhage) that are noncontiguous in nature. See my attached screenshots. The 'true bleed' picture represents a hemorrhage with 2 noncontiguous components (seen in satellite bleeds or hemorrhage post-neurosurgical intervention) with a true volume of 6.4cc. When I assign 1 ROI name to all ROIs, as you've done, I get an incorrect volume of 16 cc. However, if I name one discrete unit ICH1, which has a volume of 3.2cc, and the other unit ICH2, with a volume of 3.2 cc, I simply add these ROI volumes to get to my correct ICH total. We perform a similar workaround for contiguous ICHs/areas of interest that span different slice thicknesses on variable slice CT scans. We often see these in neurology, where physicians sometimes prefer thinner slices of the cerebellum but thicker slices for the cortex. In these cases, we segregate the ROIs, via their names, at a given slice thickness (i.e. ICH_2.5mm and ICH_5mm). We've noticed that the ROI Manager/volume calculation will take the slice thickness of img 1 and assume its' uniformity for the rest of the CT. In these variable slice CT cases, we simply multiply the 5mm slices by 2 and add it to the ROI volumes obtained at 2.5mm (i.e. volume = { (ICH_5mm x 2) + (ICH_2.5mm) }). 

Again, would love to test your plugin in though.

Thiago, that's great! We'll look forward towards that new Horos update and will let you know if we run into any problems the moment we get to testing it. Great work

[Soren Christensen]

Ok good to know there's a workaround for the build in connectivity!

Here's another example of what can happen with the current algorithm:

Draw 2 ROIs in adjacent slices. Say each is 2.55 cm2 area. Thickness is 5mm, no gap

You'd expect a volume of 2.55cm3 then (at least I would). Current algorithm gives you 1.3cm3!   For 1 slice lesions it cannot calculate at all, because it is based on distance to the next slice. So overall problematic in my opinion.

Plugin is here and should build for horos - if not let me know:

https://github.com/sorenchr2011/voxvol

It also deals with dual thickness CT and will notify you that the slices were not "equidistant" in that case, but provide a volume that was calculated taking into account the correct interslice distance for each slice.

Let me know if you find any issues with it. It currently treats all ROIs as the same group, but this shouldn't be too hard to change should you need to calculate volume for multiple classes.

Soren 

To unsubscribe from this group and stop receiving emails from it, send an email to horos-project+unsubscribe@googlegroups.com.

To post to this group, send email to horos-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/horos-project/61f558e1-07a1-4cec-9a5a-17def2446358%40googlegroups.com.

[Fauze Polpeta]

Hi there –

 

Below a nightly build o version 3.2.0 that should have fixed the Polygon ROI Growing Region.

 

https://horosproject.org/horos-content/Horos3.2.0_Nightly20180731.dmg

 

Please, let us know your results.

 

Fauze

--

[bioni...@gmail.com]

Just realized I didn't get back to you. Thank you Fauze, the ROI growing region in now functional and we've been implementing Horos. I appreciate the link!