Re: background subtraction in x-rays
- From: ImageAnalyst <imageanalyst@xxxxxxxxxxxxxx>
- Date: Sun, 9 Mar 2008 10:59:41 -0700 (PDT)
On Mar 9, 9:46 am, anujan <anujan...@xxxxxxxxx> wrote:
hello all,
i am working on an assignment which is background subtraction in
digital x-rays. i am working on MATLAB and am restricted to MATLAB as
my platform.
i am having quite a few issues & am actually stuck in the assignment.
i am enlisting them -->
1. though background subtraction ( BS ) seems a very common topic
there is no alogrithm that is clearly mentioned. most of them are
vague and strictly mathematical. i would be really greatful if some1
could point out an alogrithm which will help me is this respect.
2. the main problem is that BS is primarily for surveillance and hence
used for motion pictures. my images is however totally static. i am
using a reference image as a x-ray widout a patient and then another
image as my operating image which is a digital x-ray. as a result d
subtraction isnt actually givin me any noticeable results.
by reading the related papers i have tried to implement an algorithm
viz. HEIKKILA & OLLI on my images. i have just started out wid matlab
programming and am a beginner in this respect. treshold value has been
assumed on a trail & error basis as ther is no standard mentioned
value in the algo. the problem is that i am not gettin any results.
perhaps possibly coz my image is monochrome or then the algorithm is
not suited for such images. i am actually stuck wid it and am at my
wits end.
3. also wat quality measures should i consider to show that i am
actually enhancing the x-ray iamge? are there ne standard parameters
or then do i work on some of the statistical properties as my quality
measures?
hopin that all you experienced people could perhaps have some
guidelines for me or show me the way.
d worst part is that the submission time is just around d corner.
cheers,
ANUJAN.
ANUJAN:
1. Yes it is mathematical. It involves such mathematical
complexities as subtractedImage = dataImage - backgroundImage. But
before you do that you need to make sure the images are calibrated
(e.g. same kVp, dose, etc.) And it can become more complex if you
take into account the intensity linearity and spatial non-uniformity
of the detector(s). But for your case, you might just gloss over
that. Write out the equations for x-ray transmission (Beer's Law) and
assume a linear detector. Then assume a log detector (like x-ray film
with it's H-D curve). You will understand better how to do the math
once you've done some simple cases for yourself.
2. Background subtaction is used for serveillance. That doesn't in
any way affect you or make your problem more difficult. Also, being
monochrome does not introduce any additional difficulties into the
problem. In fact, quite the opposite: it makes it easier. Think what
additional problems might occur if it were a color image. Just
subtract and manually threshold. This might be fine for your class
project. You can some up with some standard (Otsu, etc.) or ad-hoc
automatic thresholding method if you must have an automatic method.
3. You would look at dynamic range. After subtraction, you should
have a better dynamic range than before and you should be able to see
your blood vessels easier. You could image a phantom with a known
object embedded and then show how the contrast between the object and
its surrounding neighborhood has increased. Perhaps you could also
look at detectability of the lesions or whatever you're looking for
and create some ROC curves (http://en.wikipedia.org/wiki/
Receiver_operating_characteristic). I can't do anything about your
submission time but don't fret. It's not too difficult to make a book
report on this subject in a day, just doing internet research,
including doing a simple MATLAB macro to subtract images and plot
histograms. This doesn't have to be a huge project. Don't worry,
you'll get it done because it's not that difficult (at least as far as
a class project goes because of course some people make a career out
of digital subtraction angiography).
Regards,
ImageAnalyst
.
- Follow-Ups:
- Re: background subtraction in x-rays
- From: anujan
- Re: background subtraction in x-rays
- References:
- background subtraction in x-rays
- From: anujan
- background subtraction in x-rays
- Prev by Date: Re: Is there anynone who knows a chatting site or someting like this? - Face Matching
- Next by Date: Merging many microscopic images taken from optical microscope
- Previous by thread: background subtraction in x-rays
- Next by thread: Re: background subtraction in x-rays
- Index(es):
Relevant Pages
|
