Hi Andreas

Andreas Schröter wrote:

> I have some DICOM images of modality type "DX". Inside these images I have a
> tag "(0018,1164) ImagerPixelSpacing" which tells me the pixel spacing of one
> side of the image (which may change on other positions in the image).
> Also I have tags like (0018,7030) FieldOfViewOrigin, (0018,70220) Detector
> Element Physical Size, (0018,7022) Detector Element Spacing or (0018,1720)
> VerticesOfThePolygonalCollimator and some further tags.
>
> So, how can I calculate the real pixel spacing inside the image or better:
> How can I calculate a distance of two points inside that image? I guess I
> cannot use the Imager Pixel Spacing as a "global" Pixel Spacing.

There is no simple answer to this.

The question is really one of what (and where) is meant by the "real"
pixel spacing.

In cross-sectional modalities like CT and MR, the size of a pixel is
unequivocally known, since it is the size of that pixel within the
patient's body at wherever that pixel is located. DICOM assumes
uniform sampling throughout the cross-sectional image (which may not
be entirely true in the MR case due to susceptibility and homogeneity
factors that the scanner may not correct for, but is close enough
for most purposes).

However, in the case of projection radiography, where the image is created
on a sensor or film on one side of the patient and a diverging x-ray
beam is passed through from the other side, a problem of geometry arises.
Please excuse me if I labor the obvious.

The Pixel Spacing attribute itself is not used in this case, because for
any particular pixel, the intensity is dependent on the path that the x-ray
beam took whilst passing through one side of the patient to another,
diverging from the central ray as it traveled, before impinging on the
sensor.

All that the acquisition device knows is where on the sensor the beam
arrived, and what the size of the pixel sample at that location is.

Hence the use of the attribute Imager Pixel Spacing, which is defined to
be the spacing at a particular known location relative to the sensor
plane (check the standard for the exact wording for each SOP Class).

If the patient had negligible thickness relative to the distance between
the x-ray source and the sensor, the patient was positioned directly
adjacent to the sensor, and the x-ray source was from a point of
negligible size, then the Imager Pixel Spacing could be used to measure
the size of objects in that hypothetical extremely thin patient.

As a patient's thickness increases to more realistic proportions and
becomes significant relative to the distance between the x-ray source
and the sensor, or if the patient is positioned farther from the sensor,
then geometric magnification comes into play.

If one knows the patient's thickness and the various distances involved,
and wants to measure an object in some place within the patient known
relative to their thickness, e.g. assuming the heart is in the "middle"
of the patient, then one can make some transformations based on these
assumptions.

To do this, you need attributes like Distance Source to Detector (0018,1110)
and Distance Source to Patient (0018,1111), but those alone are insufficient
because the machine has no knowledge of the size of the patient, nor of
where within the patient you want to measure.

In practice, one measures the size of something of known size (e.g. a catheter
or stent) in the same plane as what is of interest (e.g. a coronary artery)
and computes the size on a relative basis, rather than assuming anything
about the geometry.

Alternatively, one treats the image as one would an x-ray film (which has
exactly the same issues), and Image Pixel Spacing is a good analog of
what one would measure as the distance on a sheet of film. I.e., the
numbers are equally meaningless in terms of correlation with the
patient, but follow conventional assumptions if acquired in conventional
ways (e.g. for matching against orthopedic templates).

David

Hi Andreas,

Imager pixel spacing was introduced because the original planar X-Ray
IODs contained no measurement information and some of the
manufacturers started using elements from the image plane module of
other IODs to include measurement information. The problem with
planar X-Rays is that they have a geometric magnification dependent on
the ratios of the source to detector and source to target distances.
Without those measurements present there is no way to make accurate
measurements at the intended target. Considering the average size of
the human anatomy there can be significant magnification difference at
the front and at the back of the subject. Some DX/CR/DR devices
provide the magnification factor calculated as the ratio of source to
detector/source to target. For orthopedic studies it is common
practice to use feudatory markers

dee
;-D


"Andreas Schröter" wrote in message news:<2vrvjqF2pc5e9U1@uni-berlin.de>...
> Hallo,
>
> I have some DICOM images of modality type "DX". Inside these images I have a
> tag "(0018,1164) ImagerPixelSpacing" which tells me the pixel spacing of one
> side of the image (which may change on other positions in the image).
> Also I have tags like (0018,7030) FieldOfViewOrigin, (0018,70220) Detector
> Element Physical Size, (0018,7022) Detector Element Spacing or (0018,1720)
> VerticesOfThePolygonalCollimator and some further tags.
>
> So, how can I calculate the real pixel spacing inside the image or better:
> How can I calculate a distance of two points inside that image? I guess I
> cannot use the Imager Pixel Spacing as a "global" Pixel Spacing.
>
> Is there some reference source code?
>
> Thank you in advance for some help!
>
> Best regards,
> Andreas