I’m facing an issue with some DICOM files where the ImageOrientationPatient tag is either missing or empty. In such cases, I’m currently using the default orientation: imageOrientationPatient: [1, 0, 0, 0, 1, 0] // Axial identity However, this causes a problem for sagittal or coronal images because they are incorrectly treated as axial. As a result, the reference line functionality doesn’t work as expected.
How can I handle missing ImageOrientationPatient values in a way that correctly determines the image orientation and ensures reference lines work properly? this specially happein with philiphs or simens machine
Hello,
(You’ve got a knack for forgetting to greet people, Akash123 
)
ImageOrientationPatient is a type 1 tag, thus a mandatory one. These DICOM files are therefore invalid. You might be able to recreate the tags based on heuristics.
(If you’re paying for support for these modalities, you should also complain to the vendor and get the machine properly configured. That would be step #1)
You might try to use PatientOrientation, which is a textual version. For instance, L/P in it means that the rows are oriented from L to R and columns for A to P, thus meaning axial slices.
You can also use ImagePositionPatient and check how it evolves from slice to slice. If the second coordinate changes, then you have a coronal model.
The best you can do is work from heuristics, or from other similar series from the same modality.
How, for the technical side of things, you may install a callback on the stable series, download it from Orthanc, tweak it with Pydicom for instance, then re-upload it by making sure that your Orthanc is configured to overwrite existing instances.
This is how I would approach the problem
HTH