How Good Is Voluming To Critique Automatic Voluming
Name Answer
Santam I feel your own contouring concepts should be very clear before you do that. The edge detection brushes in TPS are infinitely more accurate than a human hand at defining the interfaces. In order to appreciate where the automatic contouring has gone wrong you have to appreciate how u actually contour that same site. A case in point is spinal cord contouring. If your method is the contour the entire spinal canal then automatic contouring will probably be better than ur hand drawn contours. On the other hand if you are drawing the spinal cord strictly then manual contouring is yet to be surpassed by any automatic image segmentation algorithm I have seen (Eclipse / Oncentra).
Andrew When achieving 'consultantship', I was amazed to find that I operated as a solo practitioner. I discussed approaches to treatment ("60Gy to the neck"). But when it came to putting lines on an image, no-one ever looked. except at the really complex case. But how does a complex case help you to assess your accuracy in a routine case? When I talked with colleagues about watching each other volume, I sensed panic and felt resistance. Which tells me that we may not be as good or as confident as we should be.
Indranil It is quite certain that there is a high level of interobserver variability for gross targets and even more for CTVs among most experienced operators. On top of that the 'staff' are extremely possessive about their contours. There is often no right or wrong in the CTVs at least. I think your own volumes have to be 'very good' to perform a robust assessment of the automatic output.
Andrew Indranil, the MGRT protocol has taught me that the GTV is largely irrelevant but easy to draw because it is anti-anatomical. We all think the CTV is harder to construct and it is harder initially until you hit on the idea that the CTV is anatomical. I mean ANATOMICAL, not some 'Picassoid' line drawn on in an artistic manner, but exactly anatomical - that means the CTV line must run along (as close as you can perceive and draw the interface) anatomical boundaries, ergo if we both think that a SCM is not invaded then the CTV will run along the internal muscle-fat plane, and if a SCM is at risk with invasion then the CTV will run along the external muscle-fat plane. If I think the SCM has been breached on the outer side, then the CTV will run along the platysma m. and so on. Although there is room for prediction of pathology in this, the anatomical drawing should match your anatomical risk predictions exactly. Shouldn't it?
Santam Bull's eye. Yes the contouring of the nodal CTV is definitely straightforward. Like Andrew says following the anatomical "highways" that the tumour may track along makes for a common sense approach. In my own limited experience I have found that contouring and re-contouring the nodal CTV is definitely not a big issue other than the learning curve. However contouring the CTV around the GTV is quite something else. I tend to follow an anatomical model for CTV delineation around the GTV too. This is one part that may have problems secondary to radiation induced edema. What do you say regarding this aspect Andrew?
Andrew Good stuff! The questions are getting to the nitty gritty of the anatomical logic. What you are asking about is what I call the CTVp. The technique that I have developed is to take the GTVp which has been drawn around the MASS LESION (not fat stranding) and routinely expand it by 3mm. This first margin is what I accept as the required margin for microscopic extension around a node or lesion thought on Imaging NOT to have any infiltration. If I think that there is infiltration, the CTVp margin then extends to the next unbreached tissue plane (or well into normal looking fat tissue - whole muscles are volumed if infiltrated). Furthermore the CTVp is clipped at any Air-Tissue boundary, all uninfiltrated bone/cartilage boundary where there is no infiltration. At least that is what I think I am doing!
Indranil Interesting! But why don't I agree with this? I understand that this is a 'logical' approach but I don't think this is a biological approach - I don't think that tumor infiltration into tissue is an all or none phenomenon - why should I contour till the next anatomical boundary always? If the node infiltrates the anteriormost aspect of the SCM, I don't see any reason to include the posterior most aspect in my CTV. We often do it and we are often advised to. I just don't think it is right - there is no pathological rationale behind it. There definitely needs to be a generous margin covered but not the whole structure - if you took out all the these SCM's surgically I bet you won't find the posterior edge involved in the vast majority.
Andrew I agree with you that the whole SCM may not be involved, BUT unless you undertake some imaging exercise to satisfy yourself about the EXTENT of involvement, your line is a haphazard guess, even if you add "a good margin".

We are not trained to add a "good" or "generous" margin, but rather a "good enough" or "adequate" margin, and when uninformed, they are not the same.

None of the surgical studies have looked at how far through the SCM nodal infiltration goes. So my point is two fold -
One - volume according to the imaging you have, not the imagining you have, and
Two - anatomical imaging is biological and when incorporating natural history it is also logical.

Tumour infiltration is an all or nothing phenomenon - up to leading edge of the infiltration. Our issue is whether you can identify it accurately on the simulation CT, and we both know you can't. So then you can cease voluming to the next anatomical boundary when we have studies to map these infiltration extents, and are confidently that we can include less than the whole organ.

I agree that including the whole SCM is not pathologically correct, but my imaging is anatomical and I can't confidently see the pathology (this is analogous to the change from fields-on-bones simulation to CT simulation becasue of improved anatomical imaging).

Lastly, how do I know I am right? If the node goes into the SCM, the surgeon removes it all!
CHALLENGE - start looking at all the SCMs removed by surgeons in a Roland Holland-type pathological study.
Santam I agree with Dr Andrew here. In the treatment of head and neck cancers with radiotherapy there is only ONE chance to get it right. You either cure it in the first chance or the patient dies. Secondly how much morbidity will it actually add for the patient. In my experience what I have learned is no matter how bad the side effects are in squamous cell cancers of head and neck post radiotherapy, QOL always better than a patient with recurrence at the same time point. Another point is there is no imaging modality at hand that can reliably detect microscopic extension. And if you cant see it there is no point in blindly guessing. Once a malignant cell breaches the muscle fascia it essentially has well oxygenated, well vascularised highway for further spread. As highlighted above a rigrous pathological analysis of the involved muscle is way to go for getting good data. Till that data is available over treating is far better than not treating. In this respect it is noteworthy in the years of experience with conventional radiotherapy the full muscle belly were treated so failure at this site due to missing subclinical disease will not be apparent on the analysis of these series.
Indranil I need a clarification - let's suppose you had a 2.5 cm LN with radiologic ECE into SCM in level 2 - how would the two of you draw your nodal GTV and radical and microscopic dose CTVs in the muscle - please tell me the exact anatomical extent - especially superoinferior extent of the muscle included in the CTVs.
Santam Good a practical question . Problem : 2.5 cm node level 2 (a/b?) with ECE in SCM . GTV : Node plus visible ECE forms GTV. High Risk CTV around the node will be : 1. Entire level II upto base of skull. 2. Entire SCM muscle at the levels of the node visible + 1 cm above and below (total 4.5 cm assuming 2.5 longitudinal extent too). 3. If node at interface of Level II/III include level III in HRCTV too. Since ECE is not likely to be on one direction only and if the paraspinal muscles are apparently close / abutting then will include the affected muscle in the same way (1 cm cranial and 1 cm caudal to the node + the levels of node also). I generally adopt a SIB technique where i will deliver a dose of 60 - 66 Gy to HRCTV with 66 - 70 Gy to the GTV (drawing PTVs around them ofcourse). I know this goes against the RTOG protocol but it always makes more sense to me to focus the highest dose to the greatest disease burden. The LRCTV will be determined by the site but will include the ipsilateral level III/IV/V at least. I will dose this to 54 Gy using SIB. this following recommendations given by Gregoire et al in their publication of nodal CTV delineation in node positive neck. I know the publication by Apisaranthanarax et al which recommends 1 cm margins around N1 node without gross invasion of the musculature but are AT high risk of ECE. Here you have specified however that the ECE extends into muscle
Indranil Even more interesting! You will include the entire muscle in the transverse axial sections (even areas 2-3 cm away) but in the sup-inf direction you will only give a margin of 1 cm. Why so? What makes you think that the tumor cells go further in the transverse direction rampaging across muscle fibres but stop short in the the longitudinal axis?.
Andrew …. let's suppose you had a 2.5 cm LN with radiologic ECE into SCM in level 2 - how would the two of you draw your nodal GTV and radical and microscopic dose CTVs in the muscle - please tell me the exact anatomical extent - especially superoinferior extent of the muscle included in the CTVs.
I am going to assume that by ECE you mean some fat stranding AND that it is NOT abutting/fixed to the SCM (if not then there will be another answer!). And that the ECE can be volumed in its entirely, that is to the border of normal looking fat (you may need to use multiple window setting to appreciate this.

There are two ways to draw the GTVn. Firstly, GTVn = mass. Secondly, GTVn = mass + 'ECE'. I am not concerned about this difference. In our MGRT protocol, we would just use the mass because the ECE would be imperceptible after 3-4 weeks.
But there is only one way to draw the CTVn. The CTVn = mass + 'ECE' + subclinical infiltration margin (debatable1), and then clipped at the tissue boundaries that are judged as not being breached (so I won't go into air, bone or surrounding muscle for this particular specification (NB - my CTV includes ALL of the tissue with abnormal imaging plus a margin to tissue boundary).
I would expand the CTVn by the movement margin (3mm in our department) and incorporate that into PTV7000. The surrounding areas of neck node distribution would be included in a CTVn0 and incorporated into a PTV6000 - but that process is a little more convoluted to explain as it requires exclusions and the like.

1 there is a some evidence on this, so numbers like 3-5mm seem OK (H&N articles). I have always thought that 1cm is excessive.
Indranil Andrew- I meant abutting or fixed into SCM - could you tell me your approach in that scenario as well.
Andrew Here is a picture to use for discussions. We need to be putting pencil to paper from here.
Andrew Here are my volumes.
PURPLE = submandibular gland
LIGHT_GREEN = CTVn (this image isn't the best, so the line should be 5mm from the GTVn except at the normal looking submandibular gland and long muscles of neck margin. On reflection my lower volume would also be different (I'll change if I get time)
ORANGE = CTV (this is not normally what I would draw, but this is the final CTVn0 + CTVn with expansion/clipping which would be used to generate PTV6000.)
Here is the SVG file with the contours.
Open it in Inkscape and pull in the HnNpicture.jpg above which should have the size 744x1050 and put the image behind the contours.
Indranil These are contours representative of what would be drawn at PMH.
The purple is what would be called a L2 or a lymph node at level left II. The red is L2CTV70 and the yellow is a LCTV56 (the dose fractionation is 70Gy/35Fr for gross disease and 56Gy/35Fr for microscopic disease. Note that we do not treat levels or regions to gross disease dose but a margin around the radiologically visible tumor. This margin is usually 5mm but constrained by anatomic boundaries. We do not treat the entire SCM in the axial plane to 70Gy. The microscopic volume (56GY) ecompasses the nodal regions at risk of subclinical spread (in this case would include level1b usually) and encompasses the 70Gy CTV with a margin (again 5mm with anatomic constraints).
Patterns of failure data for oropharynx cancer here show that:
Number of nodal failures outside 56Gy volume = 0
Number of nodal failures outside 70Gy volumes = 0
All nodal failures have occured in known enlarged nodes. No funky SCM failures in the uncontoured part of the muscle.
I have contoured roughly 200 HN patients in the last year - several with gross ECE into muscle - I have included the entire SCM in the axial section in only a very small number of cases with very bad ECE. In all others a margin of 5mm to 70Gy and another 5mm to 56Gy have been used.
Santam Excellent data Indranil and Andrew this is turning out to be a really good discussion. Here is my contour.

Red contour = GTV , Orange = HRCTV and Yellow = LRCTV. Doses 66 Gy to PTV around GTV
60 Gy to PTV around HRCTV and 54 Gy to PTV around LRCTV. BTW since I could not answer your question about cranio-caudal extent I forwarded it to Dr Gregoire lets see what he has to say about this

Santam Dr Vincent Gregoire's reply for this I quote verbatim "Sir, Thank you for your question.
I fully agree with what you suggested. Regarding your question, I proposed to only include the muscle in regards to the invaded level simply because I believe that there is a low (probably extremely low!) risk of muscle infiltration far away from the node with ECS. It is an arbitrary decision, I agree, but you needed to propose something making sense.
I hope this help.
V. Grégoire"
Santam I went ahead and send my pic to Dr Gregoire to see what he would delineate. His response is as follows: Thank you
In my practice, I would have done the following:
1) I agree on the volume delineation
2) I would have delivered 50 Gy (2 Gy/fraction) on the yellow and orange CTV (in tact on the PTV associated to the CTV!).
3) I would have delineated another CTV (called high risk nodal CTV), which is the GTV plus 8 mm in every directions, but included in the low risk CTV. I would have delivered 70 Gy on that CTV (in fact to the PTV associated to that CTV!).
So now the margin as per Dr Gregoire becomes 8 mm.
Andrew Interesting, isn't it? Some observations - the issue of the whole SCM or not leads to a small difference in volume - so we may be discussing different ways to reach the same profile. I find it is interesting that PMH incorporates a dose level into the CTV nomenclature, I don't mention dose until the PTV formation. Estimation of how much SCM to volume is not well characterised, is PMH using MRI fusion with planning, and does that impact on these margins? No doubt the results are excellent, and since the volume I use for disease is larger, then I can also expect similar results.

From my point of view, I am not impressed by local protocols that lack an anatomical correlate because I don't have their experience and I can't see how the decisions are made (Indranil, you need to tell us EXACTLY how you decide how far you need to go into the muscle so that we can reproduce it. Saying we do it with no failures generates a fascade of superiority and doesn't help any of the rest of us work out what to do. That's why I say 'go to the next fascial plane', and will until I hear of an image-based approach that works. Furthermore, how you change this margin when the SCM atrophies with weight loss is unknown. Are you undertaking a weekly rescanning protocol to prospectively document contours changes in normal tissues and disease, Indranil?

We were visited by the H&N RO from a very large USA centre who laid out his excellent control results. He then shared the couple of failures. He added in his presentation about one patient - "you would not have had this failure here." We questioned him, and the failure turned out to be a 'benign' node which we would not have missed because we volume all visible nodes that are larger than 2mm into CTVn (irrespective of PET/MRI). Another local custom which we have no reason for other than it aids the IMRT algorithm to conform dose.
Santam Yes you are right it is exceedingly difficult to replicate the guidelines without an anatomical or a fixed radial distance as a correlate. Indranil essentially has a radial distance of 5 mm around his GTV to which a dose of 70 Gy will be prescribed and the rest of level II/IB/V in the cut will receive 56 Gy in 35 fractions. This amounts to a gradient of 14 Gy (in 1.6 - 2 Gy = 0.4 Gy per fraction) from the posterior edge of the SCM to the anterior. My dose prescription will result in a gradient of 6 Gy (60 to 66 Gy) in 2.2 - 2 Gy per fraction or .2 Gy per fraction in the same region. Will this make a big difference in the loco-regional control. Its doubtful really. Again in our institution we have not seen nodal failures out side the original involved nodal levels except in two patients. These two patients are perhaps the reason why I am not conservative with volumes any more. One was treated with 3DCRT and another with IMRT. Both had nodal failures in the contralateral level V at the junction of the trapezius with clavicle. And none of them had a pretreatment node at this level (we went back and looked really hard). Both were locally controlled and had no metastatic disease. In both the decision to not treat this nodal level was made on the basis of a well lateralized leison in tonsil in early stages and following guidelines. I feel those two patients have given me an object lesson on the penalty of being too conservative. That said I still dont have a good answer to your question. Andrew can you please clarify the superior and inferior extent of the disease?
Indranil Guys, I am not asking you to replicate my guidelines. I am telling you what we do, and in the absence of any path data, my practice is as good as any other arbitrary guideline. It seems from Dr Gregoire's comment that he doesn't treat the entire muscle to high-risk dose, but includes it in his microscopic volume. If you ask the 4 other co-authors of his Green journal paper, they probably do their individual different things. There is in fact ridiculously little patterns of failure data in in HN IMRT. I realize that both of you have experienced or heard of some unusual failures. But for 2/100 unusual failures are you going to overtreat the other 98/100 patients. I will not. What are your numbers needed to treat (NNT) here? Unusual failures have occurred since time immemorial - this is not just an IMRT issue.
It is my personal opinion that overtreatment based on paranoia is unscientific. What if this was lung cancer or breast cancer? What are the anatomic boundaries around a parenchymal lung lesion or breast lesion. Where are your margins in those circumstances? High risk CTV to entire lung and breast? Is SBRT in lung or APBI some lunatic's idea? The CTV is the most empirical object in radiation oncology. We have thrashed the PTV issue to death, but there is NEGLIGIBLE high quality evidence on CTV margins. So till someone actually gives us some facts - we all follow our individual rationale. My rationale is that there is no biological reason to believe that tumor cells move further in muscle tissue than in soft-tissue fat spaces in the neck and my CTV70 expansion around a grossly positive node is the same in muscle or fat. You can give 5 or 8 or 10 mm - do as you please - but why do you want to give a wider margin in muscle to include the whole muscle? If I find out that there is evidence to suggest that the extension in tissues are different - I will change my practice. Till then, I am very happy to give a 5mm expansion for 70Gy and 5mm more for microscopic dose. I feel that including the whole muscle will increase the skin dose especially in IMRT planning and I have no business increasing a patient's skin toxicity with less than robust scientific evidence.
To answer Andrew's question - we do an MRI fusion for many but not all of our patients. Neck volumes are usually guided more by CT than MRI although we do seriously look at our MRIs. Andrew, you may not be impressed by my local guidelines - but the idea of extending the HRCTV to the tissue interface feels more like convenience-based than evidence-based to me. We have evidence to suggest that >95% of tumor ECE extension is limited to 5mm and 100% to 1cm in fat spaces of the neck. I don't have any evidence from any site to suggest that (non sarcoma) tumor cells move further in muscle than in fat spaces. I therefore extend the HRCTV out exactly the same way in muscle as I do in fat - by 5mm. Our miscroscopic dose extends 5mm more in muscle and I deem this perfectly adequate to control 'microscopic extension' in the remaining 5% odd situations. Finally in response to an earlier comment - I am pretty sure that surgeons remove the entire SCM primarily because its difficult to take a large chunk of muscle and leave the rest hanging there - not because they feel that the entire muscle is infested with tumor cells. You still haven't told us your supero-inferior extent of HRCTV in the muscle and what you do about tissue interfaces there.
Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License