[music]
Herbert Y. Kressel, MD Hi. This is Dr. Herb Kressel and welcome to the June Radiology podcast.
Today we have an interesting panel discussion of a pair of articles that we're publishing
in this month's issue that use a physiologically based pharmacokinetic model to look at contrast
dynamics in tissue and the focus of the first paper is to describe the accuracy of the phantom
and the second one looks at the impact of contrast medium on radiation dose in this
model.
I found the papers very interesting.
And we're joined also by the discussant of the articles and an editorial.
So let me introduce everybody.
Dr. Pooyan Sahbaee is a Senior Scientist at Siemens Healthcare and he was a doctoral student
at Duke when this study was written.
Welcome Dr. Sahbaee.
Pooyan Sahbaee, PhD Thanks so much Dr. Kressel.
It's my honor to be here.
Sure and Dr. Ehsan Samei, Professor of Radiology at Duke and a frequent contributor to Radiology.
We're delighted to have you join us as well.
Welcome.
Ehsan Samei, PhD Pleasure to be here.
Yes.
And Dr. Boone, Dr. John Boone, is Professor of Radiology and Vice Chair for Research at
the University of California Davis; and Dr. Boone I have to first thank you for your great
service to Radiology as a reviewer and for the provocative editorial that you have written.
I think it will help our readers place this study in the proper context.
Welcome.
John M. Boone, PhD Good morning.
It's a pleasure to be here.
So let's begin.
Dr. Sahbaee can you tell us about this notion of a physiologically based pharmacokinetic
model that you use to predict contrast dynamics?
Sure.
If that's okay I'll start with a short summary of the objective of the whole study
and then why we needed these physiologically based compartmental models and individual
clinical trials.
Sure.
From a – like if you studied – before we learned that there is a possibility of
increasing radiation dose due to the contrast agent.
We know that everyone is familiar with the phenomena that the presence of contrast agent
in vessels or organs enhances the attenuation and that's why we use the contrast agent.
We know that iodine increases the total radiation dose, meaning that the total radiation to
the tissue mix with iodine.
But the questions were how can we quantify this increase in different organs, and more
importantly like as a function of time and as a function of patients' attributes.
The second question that we really wanted to address was how much of this tissue, how
much of this increase in radiation dose can go beyond the mixture of iodine in the blood
or in other words how much of this increase is biologically relevant.
That was our objective of this study in some area, and we knew that we could only do it
with mutual clinical trial, the reliable mutual clinical trial.
We are aware of the complication of the process and things, but the first part of the study
we developed and incorporated a compartmental model of cardiovascular system which was indeed
a physiological based compartmental model and we in this model we basically modeled
the distribution of contrast agents or the dynamics of contrast agents through the whole
body and it consisted of a series of differential equations which take all the organs and vessel
parameters into consideration and as a result it delivers a contrast material concentration
time(inaudible)to calculate this (inaudible)rate was applied to each compartment and each compartment
was basically consisted of three subcompartments, intracellular, extravascular and intravascular,
three subcompartments.
And then based on as output again we were getting the iodine concentration in different
organs and vessels, but the good feature of this compartmental model was that we could
personalize it based on different patient attributes including like the organ volumes,
including the cardiac output, height, weight, sex, and some other factors we could basically
make them patient specific.
H.Y.K.
So what actual factors were used?
For example, did you include potential variations in renal function, the clearance rates of
the contrast?
In the current model we didn't include the renal function into account, but the other
factors were again like cardiac output, the organ volumes, and we extracted this information
from the real patient models created from the real CT images and all those like organ
volumes, the vascular volumes and all those high space and everything was basically real
lump or real values.
And you had 58 models but how did you choose the distribution of features within those
models?
Were they based on a population or how did you decide how many heavy people and how many
thin people and how many people in failure and whatnot?
This was a population; this wasn't basically first 58 patients created in our lab, in Duke's
lab, which called XCAT phantoms.
These XCAT phantoms were created based on a population of patients which had different
types of patients, body index size like – same number of the like males versus females
and it was like a very well distributed population.
Okay, well good.
How accurate are the models actually?
Dr. Samei?
Dr. Kressel, again great to be with you.
These models as Pooyan already pointed out are based on, most of them are based on whole
body CT scans that we have at our institutions.
We segment these models, we add additional features to the model, we do additional dynamic
components and we try to do a reasonable representation; sampling of the population if you will.
Not as comprehensive perhaps because there are only 58.
These are the first 58 that we have created, but they are representative of low BMI, high
BMI ranges of the patients that we have at our institution.
I think they are reasonably representative of the population at Duke University Health
System with not necessarily, we cannot necessarily say that represented individual patient that
walks into our hospital today.
(inaudible) matching.
So I would say population wise, they are representative, individual wise we are in the process of figuring
out how we can match individual patients to any of these models that we have.
On a population basis we have done a great job that actually has calibrated or validated
that this population contrast enhancement that we see in them are consistent with the
contrast enhancement that we see in a similar population that is reported elsewhere.
How accurate is the model actually?
Population wise it's very close.
I would say in the 5 to 10% range.
(Inaudible)it's difficult to say.
So I was sort of particularly excited with the notion of a virtual clinical trial where
in theory you could introduce a new agent and you could model in some lesions and you
could predict a degree of contrast, so where do we stand with this?
How can you actually use a model like this as a virtual clinical trial?
Yeah so it's a very exciting time actually.
As you know one of the challenges in medicine is that medicine being a scientific enterprise
we hope, you've got to do experiments to be able to validate your hypothesis.
But doing experiments is very difficult in the context of patients.
So virtual clinical trials really addresses that ethical dilemma that we have in that
we want to have evidence validated to experiment, at the same time we can't really do a whole
bunch of experiments that we want to do.
For example, there are a zillion different ways of imaging a patient on MR scanner or
a CT scanner and we cannot possibly entertain that kind of enterprise in the context of
a patient population.
So virtual clinical trials enable us to essentially create a population virtually and experiment
on them without ethical difficulties or pragmatic difficulties informed consenting patients
and variability across the patient.
Also the beauty of this model is that we do know the truth because we created them.
We know what's going on with them.
Now the challenge, the limitation, is that how realistic are these realizations?
If we come up with a contrast distribution or anatomical contrast that is not realistic,
of course we are gonna lead us to the unrepresentative conclusions.
But I think I would say in the last few years we have seen a greater deal of realism in
the formation of these virtual models.
In the latest ones that we are working on we are adding intra-organ heterogeneity that
would even make the images much more realistic looking not just for dose estimation.
I think the great promise of this technology is not just looking at dose.
We just happen to be using this for a dose evaluation, but the idea is that can we actually
bring all elements of the imaging, not only the risk of imaging regardless of how minute
that risk might be, but also the benefit in the construct that we know what the truth
is, and therefore we can have a much more informed way of optimizing the process of
medical care.
So the promise is that we go beyond CT, we definitely want to go beyond dose and risk.
These are just I would say low-hanging fruits if you will.
Dr. Boone are you familiar with these physiologic models?
Where do you think this technology is going to wind up?
I'm certainly familiar with the XCAT models the 58 that has been discussed.
They were sort of developed for Monte Carlo dose estimates.
What I think that the authors at Duke did which is amazing and very exciting and I congratulate
them for doing such an eloquent job with this, is they incorporated this kinetic model with
the vasculature.
Frankly it's something that I've wanted to do and just haven't had the bandwidth
to get around to doing that.
I read a lot of manuscripts from Radiology and other journals and I have to tell you
I think you know when – we just discussed this, I was very excited.
I mean more excited than in several years of reading manuscripts.
So congratulations to the authors for this spectacular work.
And I think where to go from here really is further validation.
You sort of used past data to validate this.
I think it would be fun and exciting to go and probably with a real clinical trial in
a limited number of patients and see if your organ enhancement curves, your time density
curves, really do match.
What can go on in real patients and further validation would give more utility to this
work down the line.
I think that you describe this work as hypothesis generating and I think that's very accurate.
This could be the beginning, the opening salable of a lot of research with the kinetics.
So further validation and that would lead to protocol optimization, maybe a different
timing regime for imaging the kidneys or whatever could change the dose profile and still give
the same physiologic information, things like that.
KB would have an impact here as well.
So it's very exciting and I'm sure that this will have long legs for the people at
Duke and other readers as well.
Yeah I was taken with how it might really affect the introduction of new agents because
you might be able to estimate the effect sizes and then tailor sort of clinical trials to
reduce the numbers because you have a better sense of sort of the magnitude of the effects.
I think it's very, very exciting work.
With that, let's move on to what was your primary goal in this actually which was not
just to validate the model but to look at the impact of the pharmacokinetics of contrasts
on tissue dose.
My question is actually why do this with a model?
I understand that for a lot of the other things, but actually looking at tissue dose, why use
a model as opposed to another experimental method?
That's a very good question.
Of course there are alternative ways of characterizing this phenomenon that we have characterized
here.
You can defer to the physical measurement for example as opposed to Monte Carlo simulations
or human modeling simulations and so on.
There are two reasons for doing it the way we did it.
Number one, is the issue of sampling in that if you even if you have a patient model let's
say an animal model whether a phantom model or a cadaver.
In either of those cases you could put probes inside your unit, whatever that might be,
and make an actual measurement.
The problem is that at best you will get one sample, one realization, two realization.
Doing that across 58 will be overwhelming if not impractical.
So we can't only have one or two conditions evaluated not do it across the population
that we have attempted to do so here.
Number two is that the whole if you're going after contrast agents then the dynamics of
the contrast agent needs to be incorporated.
At different time points, different organs, different tissues, get different amount of
contrasts going in and out of them, so you should do it in sort of a dynamic way across
a wide range of tissue sample.
So if you have it with the animal model, the number of probes that you need to put in and
do the temporal evaluation would be experimentally very, very challenging.
(inaudible) The other thing is you can also now change the cardiac output of your patient
and see what happens.
In the case of animal model you cannot do that as readily.
There's a lot of flexibility and sampling convenience that comes from this kind of model
that otherwise would not be possible.
Got it.
So what did the model show Dr. Sahbaee?
What is the effect on radiation tissue dose from using contrast in different organs?
Sure.
At this point it's important to clarify what we mean by radiation dose.
So the radiation dose we introduced in this study is the total radiation dose delivered
to the iodinated tissue.
Okay so it's tissue plus iodine alright?
But so with this model we showed like the administration of contrast material can increase
the total radiation dose up to like 54% in the kidney.
Okay and also in this study the second question that we wanted to address was we assume that
in highly perfused organs like liver and kidney or maybe lung and brain, the increasing radiation
dose can be approximated by normal distribution or Gaussian distribution reflecting the proximity
of iodine molecules, the organ as it distributes through the blood vessels.
Our results showed that for an individual patient with considering this proximity of
iodine to the organs, the anticipated biological relevant dose increased with respect to the
unenhanced CT can be in the range 0 to 18% increased for liver and same 0 to 27% for
kidney.
I see.
That's pretty substantial.
Now I guess the controversy in this is sort of the whole radiation dose may not be the
biologically relevant dose and there the question is the contrast that's actually intravascular
may not be contributing the same proportion to the tissue dose as the interstitial contrast.
Dr. Boone I know this is something that you've been concerned about, perhaps you could tell
us your thoughts on this issue.
Sure and let me briefly address a question to Dr. Samei, with respect to Monte Carlo
radiation dose assessment versus physical measurements you could only have so many physical
measurements because you use point detectors.
The nice thing about a Monte Carlo program is that every voxel in your phantom becomes
a measurement tool, a measurement probe, so you really have thousands, hundreds of thousands
to millions of individual dose measuring probes in a virtual environment of course.
And as the authors of this great paper recognize as well as many who are involved in Monte
Carlo studies, there are limitations of that and modeling itself has limitations.
The example I gave in the editorial are related to what if a patient has a metal implant.
So obviously most people would recognize that the dose deposited into a titanium or other
metal implant is not biologic in nature and that's a little bit true to iodine flowing
through the vessels temporarily if you – if they absorb some radiation dose they're
going to be urinated out of the body within 30 minutes after the scan.
So that dose may not be as physiologically relevant or contribute to risk as Ehsan was
saying.
The other thing is because most of the contrast agent during a CT scan is imaged early phase
so it's really intravascular, it turns out that the vessels contain of course blood and
blood is plasma and cells and the like and largely there's not a lot of DNA.
There is some, but compared to the tissues in your liver and kidney and brain, there's
really a large void of DNA in the vascular system and DNA is the target for radiation
risks.
The other thing is, is that size matters.
When you're talking about dose supposition you're really talking about, not to get
too technical here, but the x-rays interact with atoms and they kick off electrons and
it's the range of those electrons that sort of determine the resolution requirements of
your phantom and because that range is on the order of 20 to 100 micrometers you would
need a very, very high resolution phantom beyond even modern, impressive computer capabilities
now to fully simulate that.
I'll give you other examples so radiation dose via electrons to blood to feces to urine
in the bladder to implants, none of those really have a biological effect or have the
same biological effect as the regular tissues.
So the tone of my editorial was really that scale matters and we need to get to the point
and this was a great starting discussion for this important issue and then further down
I imagine that we'll refine these models, make them more accurate and get to the bottom
of whether or not this radiation dose really does contribute to risk or not.
So if I may, first of all I think I agree with you for most of the things that you said,
John.
I think we recognize the limitations of our modeling and we have acknowledged that in
the publications.
So none of these are really new to us, I agree.
It would be crazy to compute the dose on implants and say this is the dose of the patient.
Physiologically it's meaningless.
With that being said, these implants, these so-called implants iodinated implants that
we are talking about are much, much smaller admittedly than would be seen in a large metal
implants and they do get mixed in with not only the blood but also in the extra-cellular
blood and extra-cellular space.
So iodine eventually needs to make it from the arteries to the vein.
In order to do so it needs to go to the tissue.
A number of our protocols are essentially orchestrated designed time so that we can
get maximum enhancements for example in kidney or the liver; and during that time in an organ
that is highly vascularized, let's say the liver which is like 60% of – most of the
blood is capillary, you have iodine that is in a very close proximity to the tissue that
has lots of DNA that can be potentially damaged.
We have tried to do a little bit of that modelling in a few, not all of the iodine ends up in
the tissue, there was a secondary analysis that was added to the paper towards the end
assuming some sort of Gaussian distribution but the peak in the tissue would never exceed
51% that goes to the blood at Gaussian distribution.
And even that had a measurable impact on radiation dose that I think that analysis is perhaps
more physiologically relevant which actually we reflect that in the abstract as well.
And actually if I may just interject here as I'm hearing this I'm recalling that
we've published some papers on looking at double strand DNA breaks following CT and
then following CT plus contrast and the - not that double strand breaks are the equivalent
to measuring tissue dose, but the magnitude of the change was sort of more or less in
the same magnitude as I recall, so I guess the issue is it's not hard to understand
that there will be an effect but just sort of how best to express it.
I guess some of the concern is that if you over interpret the data that you'd come
away thinking that there's a lot more effect on tissue dose and I think that the subsequent
experiments that you did I think were very, very helpful at least in scaling the magnitude
of the change.
Right.
That was those papers that you refer to, there was a paper by (inaudible) in Radiology January
of last year and there's a paper that just came out by Leon Wang in European Radiology.
Same strategy, they measured, they used biomarker protein of double strand break, before iodine,
after iodine, before CT, after CT; and their estimation of this biomarker increase following
the administration of contrast and CT was higher than what we estimated which – it's
not what we modeled of course and we were sort baffled almost as to why they estimated
a higher percentage of double strand break than we anticipated.
So in some ways when I read those articles especially Wang's article that just came
out, I mean in some ways I fully relate to what John is objecting because I have the
same objection.
Is some ways you can say we are over estimating.
Yeah.
In fact we said that ourselves in our paper.
We said the range can be between zero to whatever percent you said and you even admit that might
be zero.
But these papers say that it's even more and I feel that it doesn't leave room to
investigate the mechanisms behind this enhanced double strand break.
I also want to add here that, I mean our objective here is really not to discourage the use of
contrast agents or discourage the use of CT in any way; it's just the fact that 60%
of the CT is done with contrast agent and somehow that needs to be incorporated in our
broad scheme of observation.
So Dr. Boone where do we go from here?
How are we going to sort through this?
Well I think that this was an excellent dosimetry exercise.
I think what we need to do is actually recycle some research that was done in the fifties
and sixties on microdosimetry because that will get to the bottom of it and really what
we need to do is better evaluate risks which I think is your ultimate goal here.
So microdosimetry will you probably won't do a whole body, but you could do where the
CT beam is hitting in a phantom and tease out you know the proximity of the iodine atoms
within a large vessel and in a median scale vessel and down to the capillary scale in
all of that.
So lots to do.
We just got five new computers here that are very high tech so we're excited about doing
some follow-up work on this work as well.
Great.
Well this has been a stimulating discussion.
I want to thank you all.
Dr. Sahbaee congratulations on your work and Dr. Samei thanks for sharing your insights
and Dr. Boone thank you again for a fine editorial.
It's been a pleasure speaking with you about this.
Thank you sir.
Thank you.
Bye
Bye-Bye.
For more infomation >> Sources: Lieberman is FBI director frontrunner - Duration: 1:35. 
For more infomation >> 3ABN Today Family Worship - "Scripture Affirmations" (TDYFW017007) - Duration: 58:30. 
For more infomation >> Summer Lookbook (Sub. Spanish) with Janiushka's & others - Duration: 3:29. 
For more infomation >> Feeling Lost? - Figure out what to do with your life! (Life & Career Self-help) - Duration: 4:22. 
For more infomation >> Ford B-MAX 100pk Titanium €4.000 korting! - Duration: 1:50. 
For more infomation >> Ford B-MAX 100pk Titanium €4.000 korting - Duration: 1:41. 
For more infomation >> Patente del OVNI triángular es ahora de dominio público - Duration: 2:57. 
For more infomation >> vlc record 2017 05 18 20h35m54s TV 2 Danmark HD N Regionalprogram 20170518 193040 0000 ts - Duration: 2:39. 
For more infomation >> Nissan Micra !! AIRCO 5 DEURS !! 1.3 SE N-CVT - Duration: 1:01. 
For more infomation >> BMW X1 2.0I SDRIVE, AUTOMAAT , M SPORT PAKKET,NAVI,XENON,17'' LM VELGEN,CRUISE C,4 ELEK RAMEN - Duration: 1:01. 
For more infomation >> 30 LIFE ADVICE TO LULU ON HER 30TH (2 OF 3) | lulusmiles - Duration: 4:23. 
For more infomation >> Subaru Forester - Duration: 1:01. 
For more infomation >> Mercedes-Benz M-Klasse 280 CDI Sportpakket Navi Clima Cruise Leer Pdc Lmv Trekhaak - Duration: 1:04. 
Không có nhận xét nào:
Đăng nhận xét