PDA

View Full Version : Question re Margin of Error



mamandcrm
03-16-2009, 12:17 PM
I am wondering to what the +/- 5 degrees margin of error in angle is thought to be attributable. Is it attributable to potential (human) measurement error or to potential anomalies in the imaging (e.g., a slightly different posture) so that the curve looks better or worse on film than it really is. Or both.

Thanks

Pooka1
03-16-2009, 12:57 PM
I don't know the official answer or the relative magnitude of each source of variance but these are things that would seem to have to go into the precision (I could be wrong):

1. selection of which vertebrae to measure

2. selection of the exact surface used in the measurement (top of first tilted vertebrae or bottom of last untilted vertebrae or the middle in-between, etc.)

3. range of curvature just from how the patient is standing. Large changes can be made based on how the patient is standing. Just shooting a radiograph and having the patient walk around and then shooting another might produce a few degree variation it seems.

4. if a different person is measuring the curves, the variation can be expected to be as large as +/- 10%, I assume due mostly to selection of different vertebrae thought to be affected.

5. rotation seems like it can affect the measured curve depending on if the rotated vertebrae reveal a different topography of the vertebrae surface to the radiograph

There may be other reasons.

I'd like to hear what others think.

concerned dad
03-16-2009, 01:54 PM
see about half way down this page for a description of measuring Cobb angle

scroll down to "learning point" (http://www.pediatriceducation.org/2006/12/11/)

It is the first and last maximally tilted verterbrae, not the first and last affected verterbrae.

This would imply to me that in a series of measurements of the same curve, the largest would more likely be the correct one. (the other measurements didnt use the maximally tilted verterbrae).

regarding margin of error, I would think that it is also a function of curve type (perhaps S curves are harder to measure than C curves).

Finally, on that figure in the link I gave, if you extend the two lines to the right, when they intersect they will form the same angle as the intersecting perpendiculars. IE, it is the Cobb angle.

Pooka1
03-16-2009, 02:09 PM
That is very interesting,CD! I see I should have looked it up first.

I am looking at 4 radiographs, two for reach of my daughters:

1. The top line is drawn EXACTLY through the middle of the disc and the bottom line is drawn at the bottom of the vertebrae. Neither one are obviously the "maximally tilted" ones to my eye.

2. Same as #1 though now one the vertebrae appear maximally tilted.

3. The top line is drawn at the bottom of what is arguably a "maximally tilted" vertebrae and the bottom line is drawn THROUGH the lower portion of the last arguably "maximally-tilted" vertebrae.

4. Same as #3.

ETA: As long as the lines are parallel (that is, through the disc or at the top of the vertebrae, etc.), it doesn't matter to the measurement.

LindaRacine
03-17-2009, 12:23 AM
My understanding is that it's based on 1) time of day (curves worsen as the day progresses) and 2) preciseness in measurement. With the exception of super sharp xrays, it can be very difficult to get an accurate measurement. I've seen one expert measure curves and another expert change the measurements. It's actually surprising that it's only +/- 5 degrees.

--Linda

Pooka1
03-17-2009, 07:24 AM
My understanding is that it's based on [...] 2) preciseness in measurement.

The +/- 5 is the precision. You seem to be saying the precision is a function of the precision.

I thing the OP was asking what the reasons are for why the precision is +/-5 and not something else.

mamandcrm
03-17-2009, 10:00 AM
I'm not up on the abbreviations so I'm not sure if I'm the "OP" or not, but actually what I am wanting to understand is the reason for the deviation (whatever the variation is)--is it primarily due to the human element of measurement (skill, choice of vertebrae, what have you), or to the image that is captured (something like an x-ray later in the day, or a different posture, or the positioning of the x-ray tech)....From the browsing I've done on-line, I am getting the impression that it's more the measurement end of things but I haven't found a discussion of it that lays it out clearly.

Pooka1
03-17-2009, 10:11 AM
I am guessing the main source of variation is not the actual measurement but rather everything else (time of day, not being able to stand the same way twice, choice of which vertebra is maximally tilted, etc.).

I suspect if you ask five surgeons to measure from SPECIFIC designated vertebra, the precision would be better than +/- 5 degrees, maybe +/- 2 or 3 degrees. It's just using a ruler and protractor at that point; I would expect a very small range.

concerned dad
03-17-2009, 06:09 PM
I am guessing the main source of variation is not the actual measurement but rather everything else (time of day, not being able to stand the same way twice, choice of which vertebra is maximally tilted, etc.).

I suspect if you ask five surgeons to measure from SPECIFIC designated vertebra, the precision would be better than +/- 5 degrees, maybe +/- 2 or 3 degrees. It's just using a ruler and protractor at that point; I would expect a very small range.

I agree, although complex curves may be more tricky perhaps.
There are a couple of online papers available that discuss this. One was in JBJS from 1990 (JBJS makes their old papers available) and another from an asian journal just a year or two old. The newer paper compared the results from the old fashion way (ruler and protractor) to the new way (digital xrays, ability to zoom/pan, ability to adjust contrast, precise angle measurements once the lines are chosen).
Pretty interesting papers, sorry I dont have time to look up the link but google should find them both.

mamandcrm
03-17-2009, 06:18 PM
Sorry to harass you, CD, but when you have the time (no hurry here) could you post links if you have them? I had trouble turning up papers that I could read in full that related to idiopathic scoliosis that did anything more than compare measurement differences...also, any insight into inter v. intra comparisons (that came up a LOT within the material I was reading and I did not understand it)? I am not a statistician and will never pretend to be one....

mamandcrm
03-17-2009, 06:26 PM
OK, maybe the intra v. inter thing is dawning on me. If the same subject measurer took multiple stabs at the measurement that would be the intra, and then the inter would be a comparison among all the subjects?

titaniumed
03-18-2009, 12:06 PM
Focal distance is a factor here. Its like viewing a spring in a microscope, the angular measurement is not guaranteed, and changes with focal distance. I've experienced these problems in the past, and had to prove angles with other methods, either with trig, or the use of a sine bar etc. 2 dimensional optical viewing is acceptable with the same focus. 3 dimensional optical viewing creates problems with measurement.

This is a big factor in the viewing of a rotational scoliotic spine.(spring shape)

See page 70

http://books.google.com/books?id=2ompn48QmG0C&pg=PA68&lpg=PA68&dq=x-ray+focal+point&source=bl&ots=gLImM-p8VQ&sig=gaRpgKbzejRUNDihjt0H-t_dUio&hl=en&ei=6gzBSez7IYnOsAOghZAw&sa=X&oi=book_result&resnum=5&ct=result#PPA70,M1

It really isn't all that important if a movement of a spine is under a few degrees. Its nice to know if there is any movement, and comparative x-rays will tell you that to a certain extent. In other words, the difference between a 30 degree measurement and 33 degree measurement is insignificant. Now a 30 to 40 degree change is of concern.

A way of measurement without optics would be to go in and touch probe on a coordinate measuring machine. We use these machines in manufacturing and are accurate. This would not be feasible due to the invasiveness of the idea.!

For accurate measurement of anything, a minimum of 2 different systems need to be employed and they have to match. If they do not match, its inconclusive. Accurate measurment of anything has multiple parameters.

Ed

Haleysmom
03-18-2009, 03:44 PM
From what I understand, it is a combination of factors, yes the presison of the locations for the measurments is one, but I think a huge factor is how the person is standing. A slight body shift could result in a difference - lets consider handing them from their heads!:)

concerned dad
03-19-2009, 09:30 AM
Sorry to harass you, CD, but when you have the time (no hurry here) could you post links if you have them?

No problem, These are the two papers I was talking about:

On the upper right on the following page is a link to the “Full PDF”. I believe this is available free due to the age of the paper. More recent JBJS papers require payment or subscription.

1990 JBJS Paper (http://www.ejbjs.org/cgi/content/abstract/72/3/320)

and from Asian Spine Journal – paper discussing digital radiograph measurements…

more recent Asian Spine Journal paper (http://pdf.medrang.co.kr/ASJ/002/ASJ002-02-05.pdf)

concerned dad
03-21-2009, 12:21 PM
OK, here is a, ahem <cough>, hypothetical situation.
Suppose you were told by a doctor that a curve reduced from an initial, oh say, 38 degrees to 26 degrees in brace. And then, when you carefully looked at the xrays as an amateur yourself you saw little or no reduction (0 to 3 degrees). Just suppose you then asked a different (reputable SRS) doctor to look at your measurements and he said he didnt see any issues with your amateur measurement. Note, in this hypothetical situation there is NO difference in measurement on initial angle, just inbrace angle. And further suppose, there are 3 inbrace xrays to look at.

To what would you attribute the error?

Now suppose you asked the first doctor whats up and he replied that different vertebrae were measured.

Would it not seem apparent that since you are supposed to measure the most tilted involved vertebrae that if you picked this wrong you would incorrectly get a lower value.

Pooka1
03-21-2009, 02:41 PM
OK, here is a, ahem <cough>, hypothetical situation.
Suppose you were told by a doctor that a curve reduced from an initial, oh say, 38 degrees to 26 degrees in brace. And then, when you carefully looked at the xrays as an amateur yourself you saw little or no reduction (0 to 3 degrees). Just suppose you then asked a different (reputable SRS) doctor to look at your measurements and he said he didnt see any issues with your amateur measurement. Note, in this hypothetical situation there is NO difference in measurement on initial angle, just inbrace angle. And further suppose, there are 3 inbrace xrays to look at.

To what would you attribute the error?

Now suppose you asked the first doctor whats up and he replied that different vertebrae were measured.

Would it not seem apparent that since you are supposed to measure the most tilted involved vertebrae that if you picked this wrong you would incorrectly get a lower value.

That's a very interesting hypothetical situation, CD! Very interesting indeed.

I think a way forward is to ask if the most tilted vertebra changes in brace. Then, the next question is, if it does, do you still measure from the most tilted out of brace vertebra or do you just use the most tilted in brace one?

That is a question on which reasonable people might disagree, I don't know.

Now if you are saying measurements were NOT taken from the most tilted vertebra in brace then I'm going to wonder about the intellectual honesty of that brace inventor, er, I mean orthopod.

I can tell you that these types of issues seem to be "in the air" with respect to certain brace inventors. It's why some dismiss their papers.

concerned dad
03-21-2009, 03:14 PM
Now if you are saying measurements were NOT taken from the most tilted vertebra in brace then I'm going to wonder about the intellectual honesty of that brace inventor, er, I mean orthopod.

I suppose optimistic measurements would encourage compliance, hypothetically.

Of course, it would also encourage the sale of more braces.


I can tell you that these types of issues seem to be "in the air" with respect to certain brace inventors. It's why some dismiss their papers.

if you cant believe the numbers, how can you begin to analyze the data