I’m working on a CT image analysis project and am running into some issues with non-regular slice intervals in some of our CT data, and I’m stuck trying to find a solution.
In brief, some of the Full body (CAP) CT studies at our institution are sent to PACS reconstructed as 1.25/0.625mm slice thickness/interval in the Chest, and then 2.5/2.5 mm slice thickness/interval for the Abdomen/Pelvis, and all images are stacked in a combined CAP series. When we look at a coronal MPR of the data in a standard DICOM viewer (Osirix) we get a distorted image (screenshot attached), presumably because the slice interval is 4 times less for the Chest (0.625) compared to the Abdomen/Pelvis (2.5). Our project involves only the Chest and so we are hoping to remove the Abdomen/Pelvis image from the series altogether and return the Chest images to their normal geometry. However, when we create a new series in Osirix only including the Chest images (exported from the original CAP series with non-regular slice intervals), the chest images still appear “stretched” in the z-axis. We are stuck with the images from PACS and cannot reconstruct the raw data from the scanner (screenshot attached). There is likely a simple answer to correcting the image geometry, but we have not been successful thus far. I would greatly appreciate any guidance that others may have! Some limited metadata pasted below:
Combined Chest/Abdomen/Pelvis Series
SliceThickness (0018,0050) 1.250000
SpacingBetweenSlices (0018,0088) 0.625000
SamplesperPixel (0028,0002) 1
Rows (0028,0010) 512
Columns (0028,0011) 512
PixelSpacing (0028,0030) 0.703125\0.703125
Chest Only Series:
SliceThickness (0018,0050) 1.25
SpacingBetweenSlices (0018,0088) 0.625000
SamplesperPixel (0028,0002) 1
Rows (0028,0010) 512
Columns (0028,0011) 512
PixelSpacing (0028,0030) 0.703125\0.703125
It depends on how your viewer interprets the DICOM attributes. If the voxels are not cubic, than the viewing software has to take care of that fact when displaying coronal or sagittal slices . Another possibility is - but not very likely - that your PACS changes some of the attributes.
Nicholas S Burris try sharing your CT images to replicate the error.
Do y'all have sagittal and coronal recons of the chest and separately the abdomen/pelvis sagittal and coronal recons as part of the standard PACS export you do to Osirix or are you only exporting the axial series? I wonder if you do have those recons then the segmenting can be with the original recons made from the source data (if y'all do them).
I am interested if there is a simpler solution for the source data, but working with source data performed recons is the only solution I can think of.
use Slice Location (0020,1041) which is defined as the relative position of the image plane expressed in mm relative ot the frame of reference, if availalbe.