What do people expect from their physicians?

[epepke]

(1) The Nyquist criteria, except for radiologists possible, is basically irrelevant (and unknown) to most clinicians.

In the words of Judy Tenuta, "HellooOOO!"

Can I remind you that I specifically cited radiologists, in plain English?

If you seek an understanding about why some people have an issue with physicians, and I somewhat charitably operate under the assumption that you do, you need look no further than how you are responding to me.

 

[Wrath of the Swarm]

epepke: "But you need to know that in a CT image that resolves to 256 by 256 (which is fairly typical of the kind of GE scanners I used to work with), the size of any structure smaller than 128 of the size of the image is pure guesswork."

Hopkins: "Its pretty much common sense that your field of view needs to be larger than the object to get a good image for MRI."

Is it just me or does Hopkins' statement seem to have nothing to do with what epepke said?

epepke, would you mind clarifying your statement? What do you by "resolves to 256 by 256"? And what does "smaller than 128 of the size of the image" indicate? Your wording is somewhat unclear to me.

 

[HopkinsMedStudent]

epepke: "But you need to know that in a CT image that resolves to 256 by 256 (which is fairly typical of the kind of GE scanners I used to work with), the size of any structure smaller than 128 of the size of the image is pure guesswork."

Hopkins: "Its pretty much common sense that your field of view needs to be larger than the object to get a good image for MRI."

Is it just me or does Hopkins' statement seem to have nothing to do with what epepke said?

Uhh...where did you get the idea that my statement was a direct response to his quote? Read my post again. My statement was not meant as a direct response to that part of his post.

epepke stated that radiologists need to know Nyquist for both CT AND MR scanning. The statement you quoted from me was specifiically in regards to MRI.

 

[evildave]

Possibly what people expect from their physicians is something that resolves their problem immediately.

You take that car to the mechanic's shop, and pay a bunch of money, and the car works fine.

You take your body to the doctor's shop, and pay a bunch of money, and you generally walk out feeling about the same as you did when you came in.

People want a an instant adjustment or a magic pill that makes everything all better.

In short, people have very unrealistic expectations.

Possibly this is in part due to television programs where patients come into the fictional hospital and the tall and handsome guy who plays a 'doctor' on TV says something dramatic and unintelligible, and 25 minutes later, they're all better, with time for commercials and a preview for next week's show.

They see action heroes in movies take multiple gunshot hits and throw on a bandage and keep fighting like nothing happened.

You walk in with an aching leg (poor baby!) and you walk out with an aching leg and an admonishment to allow it to heal AND a fat bill, and people feel like they've been cheated. Never mind that most "achy" things do fix themselves because the body has marvelous self-repair mechanisms.

So then people go to Chriropractors or 'alternate' practitioners and have their achy leg "adjusted" and "wow", after a lot of bending and pain and trauma, the dopamine or whatever other natural painkillers or nervous threshholds kick in, and they feel better for a while, and keep aggravating the injury. When it gets worse, it's all the doctor's fault for not instantly fixing it, don't you know?

And when it gets SO bad that the doctor finally has to resort to surgery to fix it, then they blurt "Why didn't they just do this in the first place?"


Well, welcome to the club! People do not have the same level of technical knowledge and training that you do. Nor do they even tend to have the remotest bit of interest in finding out. This is an important lesson to learn for EVERY technical profession, medicine included.

People are internally offended when you speak jargon at them. If you can't describe the process in plain English, there is a feeling that you don't understand it well enough to explain it. They might even be right.

I find I literally have to stop myself when answering a computer question for someone and edit my response. Yes, the actual technical jargon is excellent short-hand for what is happening, and can be used between myself and other members of my profession and specialization, but it is absolute gibberish to everyone else.

Instead of saying "you fractured your fifth, left metacarpal", try "you broke this bone, right here", and then point, when talking to the patient. Most people don't appreciate the opportunity to extend their vocabularies more than find out exactly why their hand is throbbing and ballooning up. Try to use this clear (although less technically descriptive) language when speaking to most people (besides your peers), and be prepared to educate them when they ask questions, and they will appreciate it. Sure it's more work to keep track of the cool and educated sounding dead language version on the forms, and have to speak english (even simplified english) and explain things to people, but developing this skill is worth it, because you'll be the doctor that people want, and appreciate, rather than the suspicious weirdo who 'speaks in tongues'.

 

[Wrath of the Swarm]

Uhh...where did you get the idea that my statement was a direct response to his quote? Read my post again. My statement was not meant as a direct response to that part of his post.

epepke stated that radiologists need to know Nyquist for both CT AND MR scanning. The statement you quoted from me was specifiically in regards to MRI.

Irrelevant. The relevant part of your statement suggests that as long as the "object" is within the field of view, there're no problems. epepke claims that physicians don't understanding the practical differences between CT and MRI.

epepke is suggesting that MRIs, as shown to physicians, appear to have a much greater resolution than they actually do. His complaint is that physicians don't seem to understand this - and as expected, you don't seem to either.

If I remember correctly, even the best MRIs don't have a resolution greater than roughly a cubic milimeter. That is the primary reason why microstrokes can't be directly detected by our current scanning technology - the affected areas are just too small.

I must concur with epepke: physicians not understanding the resolutions of their scans is a problem.

 

[HopkinsMedStudent]

Irrelevant. The relevant part of your statement suggests that as long as the "object" is within the field of view, there're no problems.

Well, there's no Nyquist ghosting artifact as long as that criteria is met. Nyquist aliasing is therefore extrapolated away from what the radiologist needs to know.

epepke claims that physicians don't understanding the practical differences between CT and MRI.

I dont recall him stating that. He claimed that because doctors dont know what Nyquist is, they cant properly interpret CT/MR images.

epepke is suggesting that MRIs, as shown to physicians, appear to have a much greater resolution than they actually do. His complaint is that physicians don't seem to understand this - and as expected, you don't seem to either.

You obviously have no clue as to how resolution is calculated for MRI, so please dont lecture me on what I do or dont understand.

BTW, here's a thread on how MRI works: http://www.randi.org/vbulletin/showthread.php?s=&threadid=24084&highlight=how+does+MRIs

For MRI, Nyquist limits prevent aliasing. If the FOV is set too small, the Fourier transform "wraps" the image outside the FOV into the image, thus creating a Nyquist ghost artifact.

The MINIMUM resolution is determined by a number of things, including FOV, gradient strength, and T2 signal decay.

When a radiologist runs an MR scan, the computer PRINTS OUT the minimum resolution on the image parameter list.

Therefore, as long as the radiologist doesnt try to "resolve" an object less than that minimum resolution, as calculated by the computer, then he doesnt have to worry about "overanalyzing" the image.

In conclusion, since the computer computes the nuts/bolts of the image, knowing Nyquist is moot. You dont need to know what Nyquist is to properly understand the minimum resolution of an MRI.

If I remember correctly, even the best MRIs don't have a resolution greater than roughly a cubic milimeter. That is the primary reason why microstrokes can't be directly detected by our current scanning technology - the affected areas are just too small.

Sure, but what does that have to do with Nyquist? Epepke's idea is that you MUST know Nyquist in order to understand concepts such as what you just said. My argument is thats a false statement.

You didnt even know what Nyquist was until I told you. My argument is that Nyquist is irrelevant from the radiologist's perspective, and yet they can still make judgments regarding minimum resolution such as the example you gave.

I must concur with epepke: physicians not understanding the resolutions of their scans is a problem.

the computer tells them what the minimum resolution is, so its extrapolated away from the radiologist.

 

[Wrath of the Swarm]

I am sincerely pleased that you understand so much about the technical issues of MRI interpretation.

Now, if I go talk to a random doctor who's just ordered such a scan, will they understand it as well?

 

[Bearguin]

You take that car to the mechanic's shop, and pay a bunch of money, and the car works fine.

I want the number for your mechanic.

The thing I find most frustrating with Doctors is that I really feel I need to see them before I go. So I've had this lump for 2 weeks and it isn't going away. I hate being told to "come back in a week and we'll see how it is". If I'm in there, I'm either in friggin pain or I had something a while. Don't just shoo me off. Learn which patients are hypochondiacs and which aren't.

I've even had my current doctor (who I do like) send me away to soak in a tub and use a cream for a week. Two days later I'm back as the pain is intolerable. He said "I thought you may be back sooner" as he lanced a blood clot out of my @$$.

Then, once everything is healed nicely, he send me to get probed. Yuck.

 

[HopkinsMedStudent]

I am sincerely pleased that you understand so much about the technical issues of MRI interpretation.

Now, if I go talk to a random doctor who's just ordered such a scan, will they understand it as well?

Will they understand the nuts and bolts of how to compute resolution or other MR parameters? Probably not, although many radiologists do understand them.

Will they understand how to interpret it properly in a clinical setting, with the assistance of the computer's calculations? Absolutely.

 

[epepke]

Are you trying to suggest that radiologists routinely misinterpret CT scans?

I have no idea.

Are you saying that it is necessary to claim that radiologists routinely misinterpret CT scans in order to demonstrate that there is a source of error that they should be aware of?

Thats a critique of computer scientists, not doctors. Doctors dont develop the technology, they just look for applications. Your gripe is with teh computer scientists.

Not really, but thank you for giving me an opportunity to talk about what I omitted from the earlier posting.

Computer scientists have far better ways of dealing with the uncertainties of sampling; these go under the heading of "volume visualization." The best algorithms are the COD algorithms, although sometimes the D is omitted to accomodate current graphics hardware. These images are accurate. However, the way in which they are accurate is that, when the resolution of structures is mathematically undiscernable, those structures appear blurry.

I have found empirically, from working with radiologists, that the vast majority will simply not accept an image from a computer calculation that does not look clear and focused.

I've also found empirically that they demand to have a number for, say, the volume of an intestinal polyp. Even though it is possible to generate a minimum and maximum size, given the mathematical constraints, they don't want to hear about it.

Perhaps this will change at some point in the future, twenty years down the line, but I don't see it happening now. I'm talking about it now in the (possibly vain) hope that some of the med students here will eventually graduate with some understanding of the vagueries of generating a complete picture from sampled data.

This does not seem to be happening. What I see is a lot of defensiveness and hostility.

Again, your criticism is not with the doctors, its with the engineers/manufacturer who designed the protocol. Doctors (the vast majority anyways) dont design protocols, they just use built in "recipes" supplied by the manufacturer.

See above. I've tried, very hard, in an academic research setting that involved working with radiologists, to convey that volume visualization is a better way of approaching this problem. As of 1996 or so, which is the last time I did this, I didn't see any acceptance.

Radiologists dont need to know Nyquist to be able to recognize MR image problems. All they have to do is set the FOV larger than the object. As long as they do that, there is no Nyquist violation.

This is such an eye-watering misunderstanding of the problem that I don't even know what to say. I'm just boggled.

 

[epepke]

Is it just me or does Hopkins' statement seem to have nothing to do with what epepke said?

No, you got it right.

epepke, would you mind clarifying your statement? What do you by "resolves to 256 by 256"? And what does "smaller than 128 of the size of the image" indicate? Your wording is somewhat unclear to me.

I don't mind. The resolution of the GE CT and MRI scanners I worked with was 256 by 256. This means that there was a square of 256 by 256 samples over a field. It might be the brain, or it might be a gut, but it's still 256 by 256.

Now, fortunatley, for the purposes of the Nyquist theorem, a 256 by 256 square can be decomposed into linear combinations of a 256-point line. Actually, any number of dimensions can be decomposed that way.

Such a line has a spatial frequency. It's a good metaphor to say that it's like a sound that can be graphed as a line.

The good bit about the Nyquist theorem is that, if a Fourier analysis of a signal has no frequency components greater that N, a sampling at a rate of 2N can entirely replicate the signal.

The bad point is that, if a signal has frequency components greater than N, with a sampling of 2N, there will be noise. This noise will reduce the accuracy of sampling in some way such that the original signal cannot be reconstructed.

There are various ways of munging the signal to change the noise, but there are no ways of eliminating the noise.

Therefore, if a CAT or MRI scan only resolves to 256 by 256, there is no way of accurately resolving structures smaller than 128 by 128.

I don't have an X-ray for reference, but from memory, they're about a cubit (18 inches) wide. So any structures that they can resolve are limited to 128th of a cubit. Which is a bit more than a third of a centimeter. Anything beyond a third of a centimeter is pure guesswork.

This doesn't seem like much, except that these results are sometimes enlarged to give a "better" picture.

 

[HopkinsMedStudent]

Are you saying that it is necessary to claim that radiologists routinely misinterpret CT scans in order to demonstrate that there is a source of error that they should be aware of?

I'm saying that if the computer extrapolates it away from teh radiologist (which it DOES for MRI) then its irrelevant.

Like I said, the MR computer TELLS YOU what the minimum resolution is. The radiologist doenst need to know all the math behind MRI because the computer abstracts it away from what the radiologist needs to know.

Not really, but thank you for giving me an opportunity to talk about what I omitted from the earlier posting.

Computer scientists have far better ways of dealing with the uncertainties of sampling; these go under the heading of "volume visualization." The best algorithms are the COD algorithms, although sometimes the D is omitted to accomodate current graphics hardware. These images are accurate. However, the way in which they are accurate is that, when the resolution of structures is mathematically undiscernable, those structures appear blurry.

I have found empirically, from working with radiologists, that the vast majority will simply not accept an image from a computer calculation that does not look clear and focused.

I've also found empirically that they demand to have a number for, say, the volume of an intestinal polyp. Even though it is possible to generate a minimum and maximum size, given the mathematical constraints, they don't want to hear about it.

Perhaps this will change at some point in the future, twenty years down the line, but I don't see it happening now. I'm talking about it now in the (possibly vain) hope that some of the med students here will eventually graduate with some understanding of the vagueries of generating a complete picture from sampled data.

See above. I've tried, very hard, in an academic research setting that involved working with radiologists, to convey that volume visualization is a better way of approaching this problem. As of 1996 or so, which is the last time I did this, I didn't see any acceptance.

I dont use virtual colonoscopies or GI imaging in general other than MRI, so I'll have to take your word on this. But you are flat out wrong when it comes to MR imaging.

This is such an eye-watering misunderstanding of the problem that I don't even know what to say. I'm just boggled.

Oh really? Then please enlighten me. In MR imaging, Nyquist ghosting is a result of a FOV smaller than the object size. In terms of minimum resolution, Nyquist is irrelevant because it sets a lower minimum than other MR parameters. You are obviously ignorant of MR physics. Let me enlighten you.

The minimum resolution on MRI depends on T2*, Gx/Gy/Gz (gradient strengths in each imaging direction), pi, and the gyromagnetic constant for the species being imaged (approx 42 MHz/T for hydrogen).

You mistakenly assume that Nyquist sets limits on the lowest resolution for MRI. THIS IS NOT ACCURATE. T2* sets a higher resolution floor than Nyquist, so T2* along with those other factors represents the lowest achievable resolution.

I don't mind. The resolution of the GE CT and MRI scanners I worked with was 256 by 256. This means that there was a square of 256 by 256 samples over a field. It might be the brain, or it might be a gut, but it's still 256 by 256.

Therefore, if a CAT or MRI scan only resolves to 256 by 256, there is no way of accurately resolving structures smaller than 128 by 128.

First off this is inaccurate. What you MEANT to say was that the minimum resolution is FOV / (1/2 *# samples). Your statement above would lead someone to believe that an object up to 1/4 the size of hte image is unresolvable, which is obviously false.

Lets look at some REAL data sets instead of just posturing with hypotheticals.

512 x 256 data set, 22 cm FOV, spin echo axial head image:

Minimum resolution = 22 cm/(1/2*256) = 0.17 cm, according to Nyquist.

Now, what you fail to realize is that Nyquist is irrelevant, because the T2* limit sets a higher threshold than Nyquist does. In addition, the computer prints out what the minimum resolution is, so the radiologist DOES NOT NEED TO KNOW WHAT NYQUIST IS in order to properly interpret the image.

 

[pgwenthold]

To get back to TT's original question:

So a couple of years ago, I was struggling with a persistent sore throat. I went to the doctor to test for strep, and it came back negative, and we went with the option of waiting it out for a couple of more days (note I didn't ask for any treatment if all it was going to be was a cold). Well, a couple of days later I developed a lump somewhere in the back of my throat that was really tender so went back to the doc and this time we figured it was probably bacterial and so antibiotics would be the right treatment. I agreed. So he gave me a bunch of samples he had sitting in his office of a new antibiotic. Pretty strong, he says, but don't worry, they shouldn't make you loopy or mess with your head of anything.

I'm like, dude, why not? Hey, if I'm going to get some strong prescription medication you might as well get me something that will give me a buzz. Come on...

So that's what I want from my doctor. If you are going to prescribe me something, make it worth my while.

 

[popsy]

I would like to have a medical professional actually listen without jumping to conclusions before they hear or know all of what I'm saying. It would also be good if they knew the right questions to ask, instead of assuming they know what's wrong and asking questions that lead only to that.

 

[Virgil]

be human, be polite, don't be offended because I'm cranky or upset when I'm sick because that is normal.

dont walk in and say "hi virgil I'm dr X." like i'm a nobody. I respond what a small world I'm a Dr. too. I worked just as hard for my PHD as you did for your MD. just because I sick I'm not a little child. Mr Virgil is OK because you don't know my status or virgil I'm steve i'm your doctor today.


give me my treatment options and let give input. if applicable. give me real odds.


speak english. if you don't know the material well enough to translate it into english you don't know it well enough.

if you don't know what caused it it say so... don't tell me it's an idopathic condition and expect that i won't look it up and feel like your laughing at me.

if you make a mistake and admit to it, I not going to sue you (unless I've suffered real harm). if you screwed up offer to fix it for free.
if you want me to see a second opinion because you're not sure thats fine.

show my the break or whatever on the x-ray. teach me what is wrong with me. give me the 1 minute lecture

be honest and treat me with basic human respect.

don't be cheap on the pain killers. if something hurts I want it fixed, painlessly. If I were a drug addict and MD shopping for drugs you'd never know about it. (I'm talking about tylenol with codene not powerful stuff like oxycotin)

IF I TELL YOU IT HURTS WHEN I GO LIKE THIS DON'T EVER SAY 'THEN DON'T GO LIKE THAT'

remember I'm a scared person and I don't know if it is just a cold or liver cancer.

hook up with a quality advice nurse. to let me know if I need to come in or not.

have your nurse call in three days and see if I'm feeling better. if applicable.


don't forget I'm a person not a case or a number

if I have to wait on you past my appoint time, be honest about it. if you had more important things say so. don't keep saying dr. will be right with you. when I call to confirm my appt say your running 30 min late. so I don't have to be on time.

when I make an appt it is like a business meeting if you late and you don't let me know it looks bad and wrecks my plans.

the best MD Was my grandfathers, he drank. when he went to the home he couldn't drink so the MD wrote a script for a shot of whiskey a day (we had to provide his medicine). that was the only way he could get it. was it a medical benifit no, but he was 95 and it wasn't hurting anyone and it made him happy.


Virgil

 

[epepke]

Sure, but what does that have to do with Nyquist? Epepke's idea is that you MUST know Nyquist in order to understand concepts such as what you just said. My argument is thats a false statement.

It's strange to me that you're saying this, because, first of all, I pointed out that you don't have to know what the theorem is in detail, simply the effects. Namely, that the resolution of structures is one-half as fine as the resolution of the scan, and that this is irreducable.

Second of all, what I was really commenting on was the hostility of physicans toward the possibility that there are things outside their traditional schooling that affect what they do. Your reactions appear to be confirming this.

 

[epepke]

Oh really? Then please enlighten me. In MR imaging, Nyquist ghosting is a result of a FOV smaller than the object size. In terms of minimum resolution, Nyquist is irrelevant because it sets a lower minimum than other MR parameters. You are obviously ignorant of MR physics. Let me enlighten you.

Sigh. I figured it would be something like this. You have glommed onto one Nyquist effect and declared it to be everything you need to know.

And then, you try to "show" me by presenting an image of a 256 by 512 dataset as a 512 by 512 image. And you don't even get the joke! Just the fact of expanding the image introduces artifacts. Since it's a power of two, it only produces one variation in the frequency domain, which is easy to filter out, but it still affects the image.

Nor is that the only source of error. Some of it comes from the rectangular resampling. I can even see features in that image that are likely to be sampling artifacts.

So, how's about you, since you know all you need to know, show me where they are, hmm?

 

[Wrath of the Swarm]

Hopkins seems to have confused "one-half the resolution of the scan" with "one-half the size of the overall image". It's the only way I can interpret several of his statements and have them make sense.

And now, like a typical physician, he'll deny that he has any problems understanding and insist that we're the ones who don't comprehend.

 

[HopkinsMedStudent]

It's strange to me that you're saying this, because, first of all, I pointed out that you don't have to know what the theorem is in detail, simply the effects. Namely, that the resolution of structures is one-half as fine as the resolution of the scan, and that this is irreducable.

Nyquist does NOT set the floor on minimum resolution. T2* effect and gradient strength sets a higher threshold than Nyquist does.

When are you going to get this?

Second of all, what I was really commenting on was the hostility of physicans toward the possibility that there are things outside their traditional schooling that affect what they do. Your reactions appear to be confirming this.

Thats not what you said originally. Your quote was that radiologists are "stupid" because they dont know what Nyquist is. I'm pointing out that knowing Nyquist has no impact in terms of CLINICAL IMAGE INTERPRETATION, as well as the fact that your understanding of Nyquist as applied to MRI is deficient and flawed.

You still havent explained why a radiologist needs to know what Nyquist is, given the fact that the computer TELLS the rads doc what the minimum resolution is, and that this minimum resolution for MRI is dominated by T2* effects and Nyquist is moot.

 

[HopkinsMedStudent]

Hopkins seems to have confused "one-half the resolution of the scan" with "one-half the size of the overall image". It's the only way I can interpret several of his statements and have them make sense.

No, I simply pointed out epepke's faulty statement. Lets look at his quote:

The resolution of the GE CT and MRI scanners I worked with was 256 by 256. This means that there was a square of 256 by 256 samples over a field. It might be the brain, or it might be a gut, but it's still 256 by 256.

Therefore, if a CAT or MRI scan only resolves to 256 by 256, there is no way of accurately resolving structures smaller than 128 by 128.

Thats a false statement. Epepke didnt describe what he was trying to say correctly, which I pointed out. What he MEANT to say was that the minimum Nyquist resolution is FOV/(1/2 * # samples), or FOV/(1/2 * 256) = FOV/128

You obviously didnt read the comments closely. I didnt misinterpret him, I corrected his erroneous statement. If you dont think his quote above is erroneous, then I dont know what to say to that.

And now, like a typical physician, he'll deny that he has any problems understanding and insist that we're the ones who don't comprehend.

This coming from someone who has no clue how MRI works, other than what little of it I've spoon fed to you like an infant.

1) Nyquist determines a resolution limit and an FOV/aliasing limit. Only the Nyquist ghosting (FOV wrapping) is relevant to the rads doc, due to #2 below.

2) Nyquist resolution limit is a LOWER threshold than T2* effects. Going by Nyquist would give you a false level of confidence. MR resolution is effectively determined by T2* and has NOTHING to do with Nyquist.

3) The computer tells the radiologist what the minimum resolution is, based on T2* and gradient field strength.

4) Therefore, since the computer abstracts this level of detail away from the radiologist, knowning Nyquist is NOT NECESSARY to interpret the image correctly, as long as you go by what the computer tells you.