Auto-contouring continues to advance

[GI_RadOnc]

@seper @Palex80 - And any others... PM me and I can give the details on how this is done. Frankly, I can talk about this for hours... even more than my love of short course radiation for rectal cancer

The dose and planning scan is basically written as an image file like a fused PET/CT; and therefore does not need any special viewer or software. The key to implementation is 1) having physics/dosimetry agree to do it at the end of treatment and run the software (10-15 minutes) and 2)Radiology IT agree to create an accession number for this and accept it into their PACS.

It's not custom software. We are using MIM, but other TP systems should be able to do it. In fact, it's so easy to do it's not even patentable. I tried, rejected due to the 'obvious nature' of the solution.

I was not able to pass 1) and 2) at my prior workplace; but my current workplace saw the utility straight away; and were kind enough to let me proceed. I now get reports like this from radiology; who are even starting to look at the radiation plan prior to reading the report! It shows up in the film strip along with all the other scans.

If we (rad onc) could generalize this; we can much easier show others what we do; what else is possible etc...

---

Members don't see this ad.

 

[salubalu]

Wanted to bump this thread. What auto-contouring software have yall been using and how has your experience been with it. Looking to implement a better option in our clinic. Thanks.

Some I am aware of include:

Radformation
MIM ProtegeAI
MVision.AI

 

[elementaryschooleconomics]

My main personal experience is with Radformation's product, which I would definitely recommend.

 

[salubalu]

My main personal experience is with Radformation's product, which I would definitely recommend.

Could you elaborate on strength/weakness of the program?

 

[elementaryschooleconomics]

It's significantly cheaper than Varian's (at least it was a year or two ago - not sure if that changed). It's an Eclipse API so if that's what you use for an EMR/treatment planning, it easily integrates. It can work with other platforms but I suspect it's not as smooth, but I don't have direct experience with that.

It's relatively fast, and many of the structures were surprisingly well contoured.

Prior to using it, my main "auto contour" experience was with MIM's atlas-based segmentation (and other atlas stuff, but mostly MIM). I basically stopped using atlas segmentation because editing the computer contours took longer than drawing my own.

Not the case with the Radformation AI contouring.

Now, big caveat being I am NOT like, uh, how do I say this...an academic Boomer faculty attending who can't see the forest for the trees and will move single MLCs 5mm thinking it matters. I know that those type of people don't like this technology in it's current form.

For me, it definitely gets in the "good enough" zone. I will personally draw/edit the ones that "matter" and leave the AI contours with minimal edits depending on the case. Meaning like, in a prostate-only case I will be nit-picky about rectum and bladder but leave femur heads. Or in breast I'll tweak the heart if needed but leave lungs - stuff like that.

Bottom line is - contact them for a demo and see if it could work for you.

 

[sirspamalot]

How much?

 

[thecarbonionangle]

Add Limbus AI. People i know who have it rave about it

 

[maxxor]

Wanted to bump this thread. What auto-contouring software have yall been using and how has your experience been with it. Looking to implement a better option in our clinic. Thanks.

Some I am aware of include:

Radformation
MIM ProtegeAI
MVision.AI

I use MIM ContourProtegeAI for MRT (LuPSMA and Lutathera) dosimetry as well as Y90 planning. It's much much better than the old Atlas methods. Not sure how fancy it can get for the stuff you guys contour.

 

[sirspamalot]

here I be like "hey if you can't fuse the PET/MRI/contrast CT for (insert technical glitch) reason just send it over and I'll eyeball it in side by side and figure it out"

/s : because real men (er, and women, and everyone in-between) contour their own volumes

 

[Palex80]

Add Limbus AI. People i know who have it rave about it

LIMBUS AI is excellent!

 

[metallica81788]

here I be like "hey if you can't fuse the PET/MRI/contrast CT for (insert technical glitch) reason just send it over and I'll eyeball it in side by side and figure it out"

/s : because real men (er, and women, and everyone in-between) contour their own volumes

Cognitive fusion

Can you bill for that?

 

[elementaryschooleconomics]

Cognitive fusion

Can you bill for that?

Only for proton plans.

 

[BobbyHeenan]

It's significantly cheaper than Varian's (at least it was a year or two ago - not sure if that changed). It's an Eclipse API so if that's what you use for an EMR/treatment planning, it easily integrates. It can work with other platforms but I suspect it's not as smooth, but I don't have direct experience with that.

It's relatively fast, and many of the structures were surprisingly well contoured.

Prior to using it, my main "auto contour" experience was with MIM's atlas-based segmentation (and other atlas stuff, but mostly MIM). I basically stopped using atlas segmentation because editing the computer contours took longer than drawing my own.

Not the case with the Radformation AI contouring.

Now, big caveat being I am NOT like, uh, how do I say this...an academic Boomer faculty attending who can't see the forest for the trees and will move single MLCs 5mm thinking it matters. I know that those type of people don't like this technology in it's current form.

For me, it definitely gets in the "good enough" zone. I will personally draw/edit the ones that "matter" and leave the AI contours with minimal edits depending on the case. Meaning like, in a prostate-only case I will be nit-picky about rectum and bladder but leave femur heads. Or in breast I'll tweak the heart if needed but leave lungs - stuff like that.

Bottom line is - contact them for a demo and see if it could work for you.

I echo this.

I've been happy with Radformation.

In my experience it shines in definitive head/neck cases (esp if patient has nice fat planes). It does elective nodal volumes incredibly well for head and neck and I like it to draw in breast nodal levels as well if treating comprehensive nodes.

Bowel it struggles with (as I'm sure we all do, it can be tricky) and pelvis nodes it's so so in my experience.

If you have a busy head neck or lung service it can really save you time in my experience.

 

[thesauce]

Contouring is not a bottleneck in my clinic’s processes. The contouring packages out there have may be saved me 10 minutes a week. I like having that extra 10 minutes, but it is not a game changer.

 

[BobbyHeenan]

Contouring is not a bottleneck in my clinic’s processes. The contouring packages out there have may be saved me 10 minutes a week. I like having that extra 10 minutes, but it is not a game changer.

I think it is more helpful for physic/dosimetry than it is me.

I have heard from my physics team it is especially helpful at some proton centers that require more frequent re-sims and re-plans to account for anatomy changes.

 

[DoctwoB]

Cognitive fusion

Can you bill for that?

As ridiculous as the term "cognitive fusion" is, we actually are pretty decent at it (or software fusion is pretty bad, or both).

Cognitive Versus Software Fusion for MRI-targeted Biopsy: Experience Before and After Implementation of Fusion - PubMed

In our institutional experience, no significant difference was observed between the diagnostic ability of COG-TB vs FUS-TB for detecting clinically significant PCa. Greater evidence demonstrating an advantage of FUS-TB over COG-TB would be required for clear recommendations in favor of FUS-TB.

pubmed.ncbi.nlm.nih.gov

 

[sirspamalot]

CMS:

';' Vee don't pay for your vorthless cognitive abilitiz.. vee only pay for procedurz...