A new page has been setup to discuss issues with contouring in general. It promises to be an interesting and free for all discussion. Do join in at http://www.isocentre.org/headneck:contour-changes-in-head-and-neck-cancer
Hello Dr Miller,
I felt it was probably more polite to address you through the forum rather than send a PM. I went through your profile and am really happy to find a fellow open source software user and linux advocate. What open source projects are u associated with sir?
Santam
Hello Santam,
firstly, it is quite OK to use my personal name, Andrew. The only time that I am "Professor" or "Doctor" Miller is when I want something done. I might do this once a month!
I am an avid linux user and I impose this choice on all my children. The only windows machine is an Acer netbook which has gone to Germany for the year and which I couldn't troubleshoot from here. There are 7 machines at home and all run linux.
I am also a FOSS advocate as I see the restrictions and dumbing-down that results from the use of proprietary software. If each unit stopped purchasing software and employed one developer to collaborate with all the other developers/coders, we would have some excellent systems. The par excellence example of this is ClearCanvas, which is a FOSS PACS.
I am not associated with any projects at the present. I am hampered by the inability to spend the necessary time needed to learn to code.
Like you I am a Mendeley user, having been a Zotero user before this.
In terms of H&N Contouring we are undertaking a weekly re-scan project at present (complete replan each week) to quantify answers to some of the questions that I posed. The most interesting thing has been the effect on my voluming. The ONLY thing transferrable from this weeks scan to next weeks scan is the anatomical constraints of you lines (e.g., above or below the SCM), which means you have to know the anatomy, trust your knowledge of the anatomy, and draw the line precisely where it must be (in some ways reproducing what the surgeon does). The pt rescanned today starts today but it has been 3 weeks since original sim (teeth problems) and the covering isodose has fallen from 98.5% of the 95% line to 90% of the 95% line! Bet you didn't know that was happening!!
A
Interesting study. Couple of question since Iam also involved in Head Neck treatments at Tom Baker in Calgary.
How frequently do you have to change/re-plan? Do you have prophylactic Ryles Tube or G-Tube inserted for all Radical Chemo+RT patients ? If not then how much weight loss do you accept ?
Nikhilesh
for patients on the protocol, we are treated for 5th fraction (8am) then re-scanning (@8.30am), fused, re-volumed and then re-calculated on old intensity maps, reassessed for coverage and if unacceptable then re-planned & re-optimised, irrespective of perceived contour changes. All new dose intensity maps are re-QA'ed also. This is all done in one day. we are calling the MGRT - Morphology Guided Radiation Therapy
Patients who are receiving concurrent chemotherapy and can't have half their oropharyngeal mucosa spared get a PEG tube (most!).
We are recording weight, and dietary measures to try to get a predictor for 'significant' changes.
Major problem with the protocol is working out how to volume accurately and reproducibly from week to week.
Yes Andrew,
I have attempted to evaluate the same thing in a small way using KV CT in the CT sim. Yes you are right there is significant deformation. Due to logistical constraints we adopted a compromise and attempted to see if doing a repeat CT at the 15th fraction helps. The first and foremost change we notice is the significant radiation edema in the masticator, parapharyngeal space, tongue musculature, larynx that obscures the fat planes significantly.I am concerned about this as I contour the CTV around GTV based on the anatomy itself not on a generic margin expansion around the GTV. The second observation is the shrinkage of the parotid and its inner movement. The third and most irritating observation with our TPS is that plan sum creation becomes a challenge. The fourth point is the accuracy of the fusion itself. I am yet to be satisfied with the rigid image registration algorithm results that eclipse provides. Specifically my concern is the poor image registration in the region of the shoulder and lower neck. How do you circumvent these Andrew?
Is there any published evidence that suggests that there are failures because of contour changes - I know it is theoretically possible, but it is my perception that iin the majority of cases you get in in your PTV. What has been your specific experience so far? I am interested in knowing how often you find that you are undersoing the gross tumor due to contour change.
I am more inclined to believe that contour changes follow a certain pattern that can be recognized and quantified. These should then be used to incorporate a deformation margin that makes sure that predicatble deformations are accounted for and do not require a replan. In the day to day scenario, I would let my observations guide which patients may require a replan based on their pretreatment characteristics (tumor bulk) and on-treatment regression (on CBCT or review clinics) etc. I would not blindly rescan every patient. In most situations it is the perception of increased dose to normal structures like the cord/plexus/mandible that will guide a replan rather than the tumor going out of field. In these situations replans are often guided quite well by regularly observing CBCTs. My experience is limited, but I am not a great believer in replanning someone multiple times. With automatic replanning - we bring in an extra diension of error from the whole replan process and that may be equally harmful.
Hi Indranil as such data of failure patterns with IMRT is rather scarce. Which is something worrisome really are we getting so good control rates that failures are non existent or something else is at work. I feel data regarding contour change correlating with failure will only come after rigorous analysis of the CBCT data acquired by such works.
Having treated IMRT patients for the last 3 years I have become a strong advocate of replanning on the other hand. Once weight loss exceeds 5 kgs you will notice a significant reduction in the neck contour with the result that your carefully planned dose distributions can shift inside. While its true that the oropharyngeal / laryngeal mucosa dont undergo that significant change and have an acceptable coverage nonetheless it is also true that unless your tumor is shrinking in a centripetal fashion uniformly target underdose will result in IMRT. In particular nodal coverage can become problematic as the contour shrinks. I have seen nodes getting covered by 98% - 100% isodoses at the planning stage and deteriorating to 95 -94 % by the third week. Since we dont do weekly CBCT and active image guided set-up correction I feel the underdose can be more though I have no way of confirming it with the tools in hand.
I feel what Andrew is trying to do is to do a reproducible contouring from plan to plan- which is really not easy as it sounds despite using anatomical guidelines. It does indeed introduce an element of error as do all interventions but I feel unless you do it you will not learn and improve.
Firstly I am finding this a wonderful discussion and hope that I have and do not offend anyone. A
"Is there any published evidence that suggests that there are failures because of contour changes"
no, but that doesn't mean it is non-existent, non-contributory or insignificant. IGRT is a big deal because every time we look we see lots more movement than we previously realised.
"… it is my perception that iin the majority of cases you get in in your PTV."
This is likely if your PTVs are 1-1.5cm larger than they need to be. Given that movement in a shell is +/- 0.5cm and smaller with IGRT, there is going to be a problem when the countour changes by 0.5cm. Can we pick a 0.5cm change in contour - I don't know becasue we have no way of reliably measuring this (we tried and failed!). We use an IGRT protocol with a 0.3cm CTV>PTV margin, and degradation in delivered dose is frequent. Would it be less with a 0.5cm margin? No. It would only be less if I planned on a 1.0cm expansion when I only needed to cover a 0.5cm margin (that is deliberately irradaited tissue not at risk).
"What has been your specific experience so far? "
Our experience is that contour changes are common, and more than expected.
"I am interested in knowing how often you find that you are undersoing the gross tumor due to contour change."
Our experience is that AFTER the 20th fraction there is little change, but before that there can be very large changes (-15 to 20% in covering isodose) but there are also patients who require NO change throughout the course of treatment.
"I am more inclined to believe that contour changes follow a certain pattern that can be recognized and quantified."
So do we, but there is no data systematically analysing what the pattern is, how to quantify it and how to find it easily. The only way to get an answer to this is to prospectivly rescan everyone at fixed intervals and do additional measures. When you do this, you still have the problem of producing the CTV that has to be covered on each image (because it changes with the patient's weight loss. At the start of the patient's treatment the fat between the SCM to long neck muscles might be 2cm, but 0.5cm at the end. If you keep the volume at 2cm you will still not cover the CTV because the contour changes are not symmetrical and can be compounded by muscle 'atrophy' in the catabolic state.
"incorporate a deformation margin that makes sure that predicatble deformations are accounted for and do not require a replan"
This approach is well known! It's called POP! As soon as you start to confine dose to a PTV and carve out critical structures the problem of contour change is active. If you decided to use 3-4cm margins on CTV>PTV, and then just take out critical structures, that will reduce the impact of contour change. But then you can kiss your mucosa and pharyngeal muscles goodbye! And most of your parotids. Only your central spinal cord would be missed.
"In the day to day scenario, I would let my observations guide which patients may require a replan based on their pretreatment characteristics (tumor bulk) and on-treatment regression …"
we have had patients drop 15% in the 2 weeks BETWEEN SIMULATION and FRACTION #5! The isodose coverage are dropping off WITHOUT any discernable external changes. Conversely we have a patient with lots of weight loss and consistent coverage - go figure!
"In most situations it is the perception of increased dose to normal structures like the cord/plexus/mandible that will guide a replan rather than the tumor going out of field."
I know it is an off the cuff remark, but why would you not replan when you think the tumour is being missed? Isn't this a denial of the Clinical Justification principles inherent in radiation protection. If I don't cover the tumour, then application of any amount of radiation is unnecessary, and harmful. Also what do you look at to get the perception that doses to critical structures are changing. How gross are these indications? Is it possible that these modifications are too late? (I know there aren't answers but my research has demonstrated that if you can't get these answers quantified somewhere, you really should start measuring)
"My experience is limited, but I am not a great believer in replanning someone multiple times. With automatic replanning - we bring in an extra diension of error from the whole replan process and that may be equally harmful."
I agree that "automatic" re-planning is an extra dimension of error, but I would assess the problem as being much greater than this. It brings a new level of danger where automated changes are just accepted (Dr Software is very believable!) My issue with replanning is that there is LESS error if I draw my CTV for the patient's anatomy today rather than use the old anatomy from 5 weeks ago when the patient was 4cm thicker and 30 pounds heavier.
"I feel data regarding contour change correlating with failure will only come after rigorous analysis of the CBCT data acquired by such works."
This is part of my point - failure analysis in the setting of contour change can only be quantified if the altering CTV is changing, not by quantifying contour change.
Santam is reflecting on the shock factor which occurs when you start routinely re-scanning patients and seeing unappreciated changes. As I say, these changes are enough to have all ROs assess the plan as unacceptable. We have also noticed some benefits with the 'local minima' phenomena in inverse planning. Rescans can result in significantly better coverage and OAR exclusion.
My ultimate aim is to develop a coherent and safe strategy for deliberate high dose inhomogeneity to improve control of the cancers that currently fail INSIDE the GTV where the dose is obviously inadequate. With movement from day to day better controlled (IGRT), the week to week changes are now relevant. Then it will be possible to dose escalate hypoxic areas and such.
